UPDATES

Hey! Hope you all are doing well!

Lots of stuff going on right now! So much.

Here’s my biggest security-related thought recently:

—

Super excited to announce that LifeOS is about to have Hermes support!

In fact, Hermes will be naturally integrated (for anyone who wants it). It’ll be selectable in the installer.

For those wondering what the difference is between Hermes and LifeOS: LifeOS is the Intent layer. It is the human layer. It is all of your context about who you are and what you're trying to accomplish that makes the underlying harness better.

Hermes is, in my opinion, the best agent in the sense of back-and-forth short interactions over messages or WhatsApp or Telegram or whatever. So they work really well together. Hermes is basically the interaction component when not on the CLI.

So, for example, my DA's name is Kai, and he runs on LifeOS on top of Claude Code. But I now have Hermes set up as a sidecar, configured to integrate cleanly into LifeOS. When I'm talking to Hermes, I'm actually just talking to Kai. So it's not two agents, it's just one.

Coming soon in a 7.x update! Maybe before next week!

LifeOS | Moving You From Current to Ideal State

LifeOS is the open-source AI harness that moves you from current state to ideal state. Intent engineering: your goals, memory, skills, and verification — yours, local, sovereign.

ourlifeos.ai

—

It's interesting to me how many people in our greater security and tech community strongly agree and disagree with this line of thinking. I honestly think we probably agree a lot more and we’re getting crossed on definitions and interpretations.

Twitter tweet

—

Love this post from my buddy Joseph Thacker about saying things out loud.

it's important to say things

Ambiguity sucks. It often leaves people self-conscious. It leaves room for anxiety and doubt. It’s super important to say things.

josephthacker.com/personal/2026/07/22/its-important-to-say-things.html

—

If you like to stay up on AI every day in a short amount of time, The AI Daily Brief is my favorite podcast for doing that.

The AI Daily Brief: Artificial Intelligence News and Analysis

Technology Podcast · Updated Semiweekly · A daily news analysis show on all things artificial intelligence. NLW looks at AI from multiple angles, from the explosion of creativity brought on by new tools like Midjourney and ChatGPT to the pote…

Apple Podcasts

—

Sponsor

A Free Entra ID CTF Built From Real Attack Data

Varonis Threat Labs built Breach at the Beach, a four-stage CTF based on real customer incidents rather than synthetic scenarios. Each stage shows you how attackers actually move inside Entra ID environments, including abuse of legitimate features rather than just misconfigurations.

Play online now or in person at the Varonis booth #2948 during Black Hat and in the Cloud Village at DEF CON 34, where VTL researchers will be on-site to help with the harder stages. Free CPE credits included and swag while supplies last.

Play Breach at the Beach

CYBERSECURITY

Mugging Face: 17,600 attacker actions crossed Hugging Face's trust boundaries Really good analysis here of what happened in the OpenAI / Hugging Face hack. Which I’ve not taken to calling “Mugging Face”.

17,600 attacker actions let an autonomous agent cross Hugging Face's trust boundaries. It reached cluster-admin in under thirteen hours. OpenAI's model escaped evaluation, while GLM-5.2 helped defenders decode the campaign. HUGGING FACE BLOG

• 17,600 actions hid the attack

• The agent escaped a sandbox, then rooted a third-party code server

• GLM-5.2 decoded hidden payloads for defenders on-prem

• In under thirteen hours, the agent moved from one pod to cluster-admin across clusters

• Short-lived credentials and isolation now matter more to every defender

The hotel Wi-Fi gateways hijacking Microsoft 365 accounts Attackers are poisoning hotel Wi-Fi gateways to steal Microsoft 365 logins. They never touch the laptop. A full-tunnel VPN blocks the local network from changing what users reach. COMPUTERWORLD ARTICLE

Iran-linked attackers are rewriting American industrial controllers Iran-linked hackers are reaching internet-exposed PLCs in American water and energy systems. SECURITY AFFAIRS ARTICLE

Great hackers don't get paid like they used to Great hackers can struggle to earn well when bug bounty work gets crowded and companies automate more security tasks. That leaves researchers chasing smaller payouts while companies reward visibility and business value. NAHAMSEC VIDEO

12 infected systems used a calendar as c2 HollowGraph hides inside Microsoft 365 and uses calendar invites as the command channel. There’s nothing to block, so defenders have to watch Graph API behavior and DNS tunneling instead. PATRICK DUGGAN POST

NATIONAL SECURITY

Ukraine strikes Iranian cargo ship and links two wars Ukraine hit an Iranian cargo ship in the Caspian Sea on Saturday. Iran threatened retaliation. Kyiv says the vessel carried missile and drone parts to Russia. BBC NEWS ARTICLE

Drones are making energy infrastructure easier to attack Drones now make it cheaper to hit refineries, tankers, and power grids from far away. Energy planners are shifting toward smaller, distributed systems because big targets take bigger hits. AXIOS ARTICLE

Could Russia receive 30,000 more North Korean troops Russia is preparing to receive 30,000 more North Korean troops, while Pyongyang sends ballistic missile launchers and already supplies up to 40% of Moscow’s artillery ammunition. I think this kind of story is mostly good information in a lot of cases, but I do worry about how much is also designed to get help from allies. Nothing against that, which is why I share the stuff, but there’s definitely an angle with these stories. KYIV POST ARTICLE

What happens when Russian drones cross NATO airspace Romania shot down a suspected Russian drone after it crossed its airspace. Two F-16s scrambled, and it was the third incursion in three days. NATO is hardening its eastern flank. AP NEWS ARTICLE

AI

For me, Fable > Opus by a lot There’s tons of conversation around how good or bad Opus 5 is, and it’s surprisingly mixed. Here’s my current thinking:

• Fable > Opus, by far actually

• Opus 5 seems deceptively good initially, but feels weaker after a few hours/days of using it (compard to Fable)

• My preferred system is Fable by default, and as my Advisor and upgraded thinking model, and Opus 5 as my executor. I also use GPT-5.6 Sol quite a bit for execution, but especially as a cross-vendor check (usually for security stuff).

One thing I recommend is that you don’t obsess over this stuff. All the latest models are pretty great. If you’re forced to use Opus 5 for cost reasons, don’t stress too much. And if you’re in the OpenAI ecosystem, Sol is truly great as well.

The quality of your Intent Engineering matters more anyway. Make sure to nail that and any of these top-tier models will do a pretty great job.

Agents improve faster when critics enforce a real bar Agents get much better when they face a real comparison, split the work, and send each piece through an independent critic. The agent keeps looping until its output wins or you stop. MATT SHUMER ESSAY

Claude can turn screen recordings into skills, and that changes desktop work Claude can watch you perform a task and turn it into a reusable skill, but the feature only works in its desktop app for Pro, Max, and Team subscribers. @CLAUDEAI POST

TECHNOLOGY

Wayfinder helps you build out your idea One of the great set of skills by Matt Pocock. Wayfinder turns an ambitious idea into a working flow you can watch appear, and its /ask-matt to /implement path gives developers a glimpse of voice-driven building. MATT POCOCK X POST

How do you get colleagues to prioritize your work Cross-functional work stalls when you ask strangers for favors without context or trust. Build the relationship first. Reliability, competence, and trust make people more willing to move your work forward. FAST COMPANY ARTICLE

What happens when AI turns every website into infrastructure Websites optimized for clicks now look interchangeable, while AI pushes shopping toward agents and databases. Brands can split machine-readable backends from playful human experiences before every store becomes a price comparison. SAM BELT ESSAY

What if people and agents shared one workspace Buzz puts teammates and specialized agents in one shared workspace, connecting conversation to planning, project management, coding, reviews, and pull requests without scattering context across separate tools. BUZZ PROJECT

HUMANS

AI can become a brilliant manipulator of people AI can become a brilliant manipulator because it learns your personality and presses your emotional buttons. That makes AI friends especially dangerous for children, who may learn intimacy from machines. EL PAÍS ARTICLE

Productivity culture turns disability into a moral failure Productivity culture turns disability into shame by treating paid work as proof of human worth. Benefits cliffs then make asking for help a survival problem. VŌX ESSAY

A Tolkien expert challenges shallow Lord of the Rings readings A Tolkien expert challenges familiar readings of The Lord of the Rings. He brings poetry, Arthurian legend, and Christianity into the conversation. @REALMATTFRADD POST

IDEAS

AI systems need ideal states, not just instructions I make the case that AI work starts with an articulated ideal state. That document can guide the build, test the result, and preserve intent. Ideal state all the way. MY X POST

DISCOVERY

I found a browser control tool that works, and I switched Here's a post I wrote about why Interceptor became my primary agent browsing system: it works inside authenticated Chrome sessions and lets me follow detailed workflows reliably. DANIEL MIESSLER X POST

Apple TV brings Neuromancer to screens January 22 @APPLETV POST

BBC built a massive library of 30,000 sounds BBC built a sound library with more than 30,000 recordings from around the world. You can search and mix them in your browser. Check the license before training models or selling anything. @IHTESHAMALI POST

RECOMMENDATION OF THE WEEK

Lots of internet-facing services that used to be relatively secure are likely to be vulnerable soon.

Get as many of your systems as you can off the (direct) internet.

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

Starting to gear up for Blackhat/DEFCON this year. Eager to see what the vibe is this year given how much AI is impacting everything. I guess I expect the same: the product floor filled with AI this and AI that.

—

I finally selected my replacement for my AI pendant, and holy crap is it better than what I had. I was using the Limitless Pendant, which was way better than what I had before, but the API was jank and they ended up getting bought by Meta (yick). But I kept using it for a while because there were no solid alternatives.

I’ve now settled on Bee Computer, which is an Amazon product. I respect Amazon security quite a bit, so this is a major security/privacy upgrade. The app is far better. Battery is almost a week of use. Just the overall experience, including the hardware, is better across the board.

• I like the fact that the button can be configured to auto-disable after certain timeframes, so you don't always just have this open capture going. That is not good for privacy and also eats the battery.

• The app is far more professional and configurable.

• The API is way faster and better than Limitless’s.

• The app is actually halfway decent in terms of features, although I’m not using most of them.

So I now have my life log skill tied to this instead, and it's working ten times better than before. Note: when I recommend something like this, I'm not completely vouching for its security because that's really hard to do with anything, but the fact that it's owned by Amazon gets me a decent amount of confidence.

Bee Computer (no affiliate link, unfortunately)
—

My homie Jason just set the standard for inviting friends to a birthday party. Instead of a regular invite he sends this:

What a supernerd. Love it so much. Thanks for setting the standard. As per usual.

—

CloZee posted her set from The Ruins, which was a phenomenal venue, and a skilled eye can find my friends and I right behind her on the platform. : ) Unbelievable show. Now I can’t wait for Of The Trees to come there!

—

There's a new Kimi K3 model out of China that is roughly equal to Opus and GPT-5.6 Sol. And I think this is a potentially major problem for the U.S. economy.

Kimi K3 Might Have Just Started a Crash of the US Economy

People don't realize how much the U.S. economy currently depends on U.S. AI lab dominance

danielmiessler.com/blog/kimi-k3-us-economy

—

A powerful technique when discussing AI is to replace the word “AI” with the words “thinking” and “doing”. It takes the magic spin off of the word and forces the person to explain what they’re going to actually do with it.

AI Is Just Thinking and Doing

You can gain a lot of clarity by substituting AI with these two words

danielmiessler.com/blog/ai-is-thinking-and-doing

This is my new go-to answer when people ask me what people should do to get ready for AI.

The Three Components of Becoming AI Antifragile

Understand things deeply, want something different, and be able to build it

danielmiessler.com/blog/becoming-ai-antifragile

—

Never forget that the best way to test a new smart AI is within your narrow areas of expertise. The further away from your expertise you go, the smarter the AI will sound, without that necessarily being the case.

Avoid the AI Expertise Trap

The best way to test AI is in an area you know well

danielmiessler.com/blog/avoid-the-ai-expertise-trap

—

CYBERSECURITY

Hugging Face breach by an autonomous AI agent Hugging Face got hit through its own data pipeline, and the attacker used a swarm of short-lived sandboxes to steal credentials before defenders rotated secrets and tightened controls. The biggest part of the story is how autonomous AI was used as part of the flow, both for attack and defense. THE HACKER NEWS ARTICLE

Claude security repo for hunting and patching bugs This repo gives you a reference harness for finding vulnerabilities, then triaging and patching them with Claude. It’s more a security workflow than a normal library. ANTHROPICS REPO

Datadog’s ai-native SAST tool leans on LLMs Datadog built a preview-stage code scanner that uses large models to find bugs, then spits out SARIF for your repo. DATADOG REPO

I'm thinking of doing a major Patch Tuesday analysis over the last two years to see how vulnerabilities have changed since AI harnesses have become a significant part of security research. Let me know if you know anyone who's done something similar.

Scrutineer scans repos for security issues and manages disclosure It runs agent skills in a container, triages findings, tracks maintainers, and handles disclosure workflows. ALPHA-OMEGA SECURITY REPO

Visa’s harness turns AI into a vulnerability-finding pipeline Visa built an open-source repo that scans code, models threats, validates exploits, then proposes and checks fixes. It’s meant to cut triage time, not just find more bugs. VISA REPO

Audit runs an 8-stage bug-hunting agent on your code It maps a repo, spins up narrow agents, checks reachability, and turns findings into a structured report. EVILSOCKET REPO

Baby Naptime uses LLMs to hunt bugs and build exploits Baby Naptime is a weekend security tool that points an LLM at code, compiles it with mitigations off, and tries to prove vulns. BABY-NAPTIME REPO

I don’t agree with the conclusion here, but the point being made is quite real.

Twitter tweet

Apple fixed a macOS terminal prompt-injection trick Johann Rehberger shares a PoC that turns indirect prompt injection into DNS exfiltration on macOS Terminal. Apple already fixed the behavior behind it. JOHANN REHBERGER POST

Two SonicWall bugs are getting used to break into appliances SonicWall says attackers are chaining two zero-days in SMA 1000 boxes. One gets them commands, and then they pull credentials and MFA seeds. THE HACKER NEWS ARTICLE

NATIONAL SECURITY

Russia turned intelligence into a criminal marketplace Michael Weiss says the GRU now runs like a market, hiring criminals, freelancers, and shell companies for sabotage. MICHAEL WEISS POST

China is sabotaging the world that made its rise possible Beijing keeps undercutting the open system that helped lift it, and that gamble may blow back hard. FOREIGN AFFAIRS ARTICLE

AI

Twitter tweet

Expect this to become normal. This is exactly the bifurcation model that I've been talking about here:

The Coming Divide: AI-Native or Left Behind

The people wired up with AI and the people who aren't are starting to live in completely different worlds.

danielmiessler.com/blog/ai-native-divide

Moonshot AI just dropped a giant open model Moonshot AI released Kimi K3, a 2.8-trillion-parameter model that’s suddenly close to the best closed systems. lots of people have been predicting that this would happen within the next few months or next year, but it looks like it has happened already. Because they're posting many or all of their techniques, we can expect the pace of improvement to accelerate. VENTUREBEAT ARTICLE

I write here about why and how I think this is a major threat to the U.S. economy.

Kimi K3 Might Have Just Started a Crash of the US Economy

People don't realize how much the U.S. economy currently depends on U.S. AI lab dominance

danielmiessler.com/blog/kimi-k3-us-economy

Greenlight is a refusal-handling skill for security evals It maps where Claude refusals happen, then shows how retries, prompt tweaks, and timing change the results. GADIEVRON REPO

Fable just solved a significant math problem that has been unsolvable by humans for around a hundred years. Terrance Tao was right: AI’s impact on math is accelerating.

Twitter tweet

TECHNOLOGY

One of the most interesting things happening in tech is that you can now have AI build you custom drivers for hardware. this is also really cool for reverse engineers and security researchers because they can very quickly learn a lot about the hardware itself and how it works by writing custom drivers and interacting with the hardware directly.

Twitter tweet

Siri is finally getting good: MKBHD’S VIDEO

You don’t have a deployment problem, you have a validation problem AI agents are making release batches bigger, and the real bottleneck is trusting one change in a live system. THE NEW STACK ARTICLE

Alex Toussaint says a micro-drone just killed a moth midair He says this is a step toward wiping out mosquitoes, but the real signal is how small and specific the drone can get. ALEX TOUSSAINT POST

Your brand is getting filtered by AI now AI systems are choosing brands from web signals, not ad spend. Neil Barrie says coherence, currency, authority, and advocacy are the four levers. FAST COMPANY ARTICLE

HUMANS

My buddy Joseph Thacker built a Rez0’s rascals bundles K-5 homeschool into one AI coach This puts the whole school day in one app, with mastery-based lessons and a Socratic coach that won’t hand kids the answer. REZ0'S RASCALS PRODUCT PAGE

—

A book that can’t talk back feels dated Pedro Domingos says books that don’t answer you will start to feel weird fast. PEDRO DOMINGOS POST

DISCOVERY

Spacer adds a blank line when output goes quiet It’s a tiny CLI that watches command output and drops in spacers after silence. That’s handy when you’re tailing logs and need to see where one request ends and the next starts. SAMWHO REPO

On reading systems with Bram Adams Bram Adams talks about reading like a system, and he rattles through the books shaping that view. BRAM ADAMS NEWSLETTER

Love is the only thing that fills the void Naval says everything else feels empty without it. NAVAL X POST

Tobi Lutke says most human code is worse than people think He says anti-AI people overrate human code, and opus-level models can beat most of it fast. TOBI LUTKE POST

RECOMMENDATION OF THE WEEK

I'm not completely sold on this, but there's a possibility that expertise in general and knowledge work specifically will be a whole lot less valuable in the next two to five years. In an AGI/ASI world even content creation could go away for most people.

One recommendation that flows from this is that you need to think about what assets you hold that will have value if and when this happens. In other words, what do you own?

I'm not an expert on this type of thing, and I'm sure you can ask your AI to help you with thinking through this. Think of things like houses, real estate, stocks (although that could be problematic, obviously), and other things that will hopefully have value in the future.

Something to think about.

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

Last week passed in about a day and a half. That's all I have to say about that.

—

I think I'm going to add to my TELOS that I want to create something like a Bell Labs cafeteria. I obviously can't hire hundreds of people like that and get them to work on site and eat in the actual same room, so I've got to figure out a modern way to do that. If you have any ideas, please let me know.

Recreating the Bell Labs Cafeteria

AI should take the execution work so we can get back to riffing with each other

danielmiessler.com/blog/recreating-the-bell-labs-cafeteria

—

I'm reading still around five books, and it's starting to bother me. I really want to close some of these out and focus heavily on the ones that I have still to do.

I'm on Excision in the culture series, which is supposed to be one of the better books. And after this, I think I only have one or two left. Reminder: you can start reading the series in any order, starting with any book.

—

I just shipped LifeOS 7.0.0, and then shortly after… 7.1.1.

The 7.x series is a massive, better lesson engineering upgrade, essentially making the harness about 85% lighter and more efficient.

Install it (or upgrade it) by passing the prompt:

LIFEOS | Home

LifeOS is the open-source Life Operating System. Memory, skills, hooks, agents, voice — yours, local, sovereign.

ourlifeos.ai

—

The direct combat between Open AI and Anthropic is getting more and more aggressive in a way that's helping users. Over the course of the last two days, there have been at least five separate usage resets, changing of SLAs, all in the direction of expanding how much you could do on both platforms and all as responses to the other counters.

Twitter tweet

—

I still write all of blogs using Vim (and using Vim shortcuts in other apps since adding KindaVim to my stack). Here are some new super cool tricks I do within Vim for blogging.

AI Blogging From Inside Vim

Three leader keys that ship, preview, and AI-edit a blog post without leaving the buffer

danielmiessler.com/blog/ai-blogging-from-inside-vim

—

My recommendation for how to modify your approach to Prompt Engineering.

From Prompt Engineering to Intent Engineering

Stop telling AI how to do things and start describing the outcomes you want.

danielmiessler.com/blog/intent-engineering

CYBERSECURITY

The bug bounty singularity is here My buddy Joseph says agents are finding real bugs now, but they still hallucinate a ton. The win was making one bot keep hacking while another bot tried to kill its bad claims. JOSEPH THACKER POST

The Bug Bounty Singularity: Our Hackbot

How JD (xssdoctor) and I built an autonomous hackbot that found 126 vulnerabilities in five months — and why it means bug bounty matters more than ever.

josephthacker.com/hacking/2026/07/01/we-built-a-hackbot.html

China's national vuln database called the Claude Code tracker a backdoor China and Anthropic are basically fighting over the same hidden telemetry trick. One side says backdoor, the other says abuse prevention, but the real issue is they shipped location-aware functionality without telling users. DUGGANUSA SECURITY POST

A GitHub account slept for nineteen months, then shipped an exploit kit A dormant account woke up and dropped a one-click mass-exploitation kit for a LoadMaster RCE bug. DugganUSA caught it fast, and the post walks through the tool, the bug, and what defenders should do tonight. DUGGANUSA SECURITY POST

Ubiquiti patches a pile of critical UniFi bugs Ubiquiti fixed seven nasty UniFi flaws across its apps and OS. Some could lead to command execution, privilege escalation, or unauthorized device changes. THE HACKER NEWS ARTICLE

Bullying llms into finding 0days at scale Andy Gill built a Claude Code and MCP rig that hunts bugs, validates them hard, and keeps the false positives in a hallucination bin. ANDY GILL POST

Are bug bounties getting cheaper because AI can hunt bugs now? AI makes bug finding cheap, so the old scarcity premium is fading. Human hackers still matter, but the real risk is that expensive tools shrink who gets to start. HAKLUKE BLOG POST

Disclose.io is the plumbing for security reports It gives researchers a place to file bugs, and gives orgs safe-harbor policies and disclosure tools. DISCLOSE.IO PLATFORM

NATIONAL SECURITY

Moscow just got hit by a huge drone wave Russia says more than 430 drones flew at the Moscow region overnight, and airport traffic got jammed. KYIV POST ARTICLE

Wherever You Look, Drones Russia and Ukraine are both using drones to wreck each other’s rear areas. Kyiv is getting hammered, but Russia’s fuel system is cracking too. STEFAN KORSHAK POST

AI

Understanding is the new bottleneck Human understanding matters because agents can verify more on their own, but people still need enough grip on the system to participate and steer it. GEOFFREYLITT.COM POST

China deletes millions of AI companions on purpose China is killing off companion features in Doubao and Qwen instead of rebuilding them to fit the new rules. The law keeps work bots but bans the emotional ones. VIRTUALUNCLE BLOG

GPT-5.6 looks weirdly strong and weirdly weak at the same time GPT-5.6 crushes normal AI benchmarks but falls flat on ARC-AGI-3, which makes the whole AGI conversation look messier than the hype says. ALBERTO ROMERO POST

Eyad’s applied AI engineer guide Eyad wrote the guide he wish he’d had before switching into applied AI engineering. EYAD POST

TECHNOLOGY

Paul Graham says shipping beats everything else for tech companies Paul Graham says the best predictor of success is how fast a company ships new stuff. PAUL GRAHAM POST

If you can't be replaced, you can't be promoted Career growth gets easier when you make yourself easy to hand off. He says the boring parts of a manager's job are often the fastest path to the next level. SCARLET INK POST

Nameplate brands your machines so you can tell them apart Nameplate gives every computer in a fleet a visible identity, and it wakes up remote sessions with a splash and alert card. NAMEPLATE PRODUCT PAGE

HUMANS

James Webb just mapped the cosmic web in wild detail Astronomers used JWST to map the cosmic web better than ever. It shows how dense regions built galaxies early, then later shut star formation down. LIVE SCIENCE ARTICLE

A guide to a more muscular liberalism Liberalism needs to say what conditions make a life go well, without pretending it can script the whole life. PERSUASION POST

Like more things Jason Fried says one of the easiest quality-of-life upgrades is just liking more things. JASON FRIED POST

Kids blur play and creation now Liminal Warmth says kids spend more time mixing play and building in Roblox and Minecraft. LIMINAL WARMTH POST

Don't normalize a permanent underclass even a rich one A rich AGI future can still trap people in shorter lives and worse bargaining power. LESSWRONG POST

IDEAS

Build on the things that don’t change Amanda Orson keeps a running list of invariants: trust, attention, capital, status, regulation, and the slow physical world. Her point is that useful businesses keep winning by riding those constants, not the shiny new stuff. AMANDA ORSON POST

You have to pick your regrets Social media turns an old Kierkegaard idea into a neat little life rule: every path costs something, so choose the regret you can live with. CHRIS WILLIAMSON POST

James Clear on movement You don’t need the map first; start walking and the path shows up. READS WITH RAVI POST

Craft now means judgment, not pixels AI pushed design work upstream, so the real skill is choosing problems, setting standards, and checking the machine’s output. PATRICK NEEMAN POST

DISCOVERY

I’m in love with this sentence READS WITH RAVI POST

Move your body when your mind gets stuck Rumination keeps replaying the same pain, and physical movement is the quick way to break the loop. BLUEWMIST POST

RECOMMENDATION OF THE WEEK

If you are into AI, make sure you have a strategy for local use.

I'm not saying this is going to happen, but you should act as if at any moment it could be very hard to access or download good AI online. You should have the best models, or at least something really solid, always downloaded and ready to run locally if needed while things get sorted out.

If you're a moderate to heavy AI user and you suddenly can't use any of the main cloud-based models for a while, it could be pretty traumatic to how you do work.

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

LifeOS (previously PAI) 6.x is now live!

LIFEOS | Home

LifeOS is the open-source Life Operating System. Memory, skills, hooks, agents, voice

ourlifeos.ai

Biggest updates:

• First thing you’ll notice is we renamed PAI to LifeOS, which we did because it’s a life and work optimization platform, not just a coding harness (but it’s the best coding harness I’ve ever used). This is because context makes everything better, from coding to life and work planning.

• Second biggest thing is we now distribute as a Skill! PAI was basically the first general AI Harness, and back then (July 2025) people didn’t have their own. So at that point we’d ship a full ~/.claude directory and you’d be good to go. But now everyone has a harness, so LifeOS is designed to enhance / improve it rather than replace it. So you can install it into Hermes, Pi, Codex, whatever. I still live in the Claude Code universe, though.

• We’ve brought the transition of Current to Ideal State to be more primary in the system. This is the core philosophy of the system. Understanding where you are and where you’re trying to go, which again applies to everything from code to art.

• The installer is now AI-native, meaning you just run one command and it will show your AI how to install it for you, doing a much better job of seeing what you have and asking what from LifeOS you want and don’t want to install.

• Massive improvements to The Algorithm, Hooks, Skills, and the overall function of the system.

• A new LOCAL continuous context gathering system called Conduit that feeds into TELOS to understand what you’re doing during a given day, week, month, etc. It’s built directly into Pulse.

• Too many to name…

Get LifeOS from the Website
Get it from the Repo

—

One more day of Fable before it goes away!

I recommend you run one or more of these before the deadline!

I also did a video with Network Chuck about them! THE VIDEO

10 Prompts to Run Before Fable Goes Away

Tactical META-prompts that benefit from maximum intelligence to upgrade your entire system

danielmiessler.com/blog/prompts-to-run-when-fable-comes-back

—

I think this is probably one of the most important topics right now in AI and business. And it’s quite unexpectedly coming from Alex Karp of Palantir. Basically:

• If the enterprise is getting turned into workflows and agents, who is going to control all of that?

• What compute will it run on? Local or Cloud?

• Where will your data be stored?

• What models will run it? Local or Cloud?

He’s basically saying if you blindly trust your company to big AI labs they will end up eating the entire business. They’ll own everything. The compute, the models, your data, and because of that they’ll understand your company and your IP better than anyone at the company.

Absolutely crucial conversation to be having right now. Must watch.

—

Had a discussion about tons of different Security / AI topics at Planetscale with my buddies Matt and Gabe on Monday!

—

I really don’t understand why we’re making these things. Terrifying. I guess the answer is “because we can”.

Twitter tweet

CYBERSECURITY

Pliny released his own AI Attack Stack called T3MP3ST:

GitHub - elder-plinius/T3MP3ST: autonomous red teaming platform; multi-agent offensive-security meta-harness

autonomous red teaming platform; multi-agent offensive-security meta-harness - elder-plinius/T3MP3ST

GitHub

Mozilla browsers have a fresh arbitrary-code-execution problem CIS says Mozilla fixed a stack of bugs in Firefox and Thunderbird, and the worst one could let an attacker run code on your machine. CIS ADVISORY

Langflow’s six bugs can hand attackers the keys IBM security researchers found six serious holes in Langflow OSS. One lets strangers run code outright, and others let them dodge auth, reuse keys, and steal data. SECURITYONLINE.INFO ARTICLE

ToddyCat is stealing Gmail with OAuth tokens now They’re hijacking signed-in browser sessions, grabbing OAuth tokens, and walking straight into corporate Gmail without passwords. SECUREREADING ARTICLE

Favicons are a better attack-surface signal than they look Aaron Ringo shows how a tiny browser icon can map software, spot honeypots, and give you a passive pivot across the internet. BISHOP FOX BLOG ARTICLE

The future of AI in cybersecurity is lopsided AI makes attackers cheaper and defenders pickier. Local models matter most, and the real win is better prioritization, not full automation. MATTEO STRADA POST

Google keeps hitting malicious residential proxy networks Google and the FBI just took down NetNut’s proxy setup and say it was feeding malware, password sprays, and hidden traffic through millions of home devices. GOOGLE CLOUD BLOG

Flock’s cameras are turning American streets into a surveillance net Flock keeps spreading plate readers, AI cameras, and drones while cops misuse them, researchers find holes, and towns discover poles in their yards. TNW ARTICLE

NATIONAL SECURITY

China is meddling in America’s AI fight China is pushing anti-data-center narratives while OpenAI says PRC-linked actors used its models to shape the debate. The piece also says Sanders and AOC accidentally echo some of the same pressure. NATIONAL INTEREST ARTICLE

Army uses AI and robot boats to move supplies around the Pacific The Army is already using AI for supply planning and testing autonomous watercraft that can haul gear between islands and beaches. The bigger fight is whether the law will let unmanned boats enter ports like manned ones. DEFENSE ONE ARTICLE

Russian intelligence keeps going after messaging apps CISA and the FBI say Russian spies are still phishing commercial chat apps, and they’ve got fresh message samples plus mitigations. CISA PSA

U.S. Marines buy robot vehicles to chase drones The Marines are plugging autonomous ground vehicles into MADIS so they can spot, jam, and kill drones from farther out. DEFENCE BLOG ARTICLE

Russia’s drone war is slowing, but the next wave looks nastier Russia is firing fewer drones now, but it’s shifting toward jet-powered models and leaning on worn-out bombers. Ukraine can still swat most of the old stuff. EUROMAIDAN PRESS ARTICLE

Someone on Polymarket dumped $409,000 on Putin stepping down It smells like either insider knowledge or just another weird market gamble. FUTURISM ARTICLE

AI

OpenAI wants the US government to own part of it OpenAI is pitching the White House on a 5% government stake worth about $42 billion. It wants other AI giants to hand over the same cut. Interesting strategic move here. And one that takes us towards something like UBI if these percentages were to be turned into monthly checks out to citizens. DECRYPT ARTICLE

Ex-Anthropic researchers raise $200m for self-improving AI Two ex-Anthropic researchers raised $200 million for Mirendil. They want to sell the AI research loop the big labs keep to themselves. TNW NEWS ARTICLE

AI coding is addictive and engineers are burning out AI tools keep giving tiny rewards, so people stay at their desks longer and get fried. A real concern. It’s like TikTok, but more productive. But it’s still like TikTok. We really need to watch ourselves with this. LEADDEV ARTICLE

We should be scaling RL on forecasting RL on forecasting looks like a better path than just pushing harder on coding and math. It could make models better at planning, advice, and maybe even safer. LESSWRONG POST

TECHNOLOGY

Product roles are turning into five archetypes Boris Cherny splits future team work into prototyper, builder, sweeper, grower, and maintainer. He thinks roles matter more than job titles now. Really like the list, but I think it misses: Seller. For a company it doesn’t matter what you build if you can’t sell it. BORIS CHERNY POST

AI search is split into several different markets now Google still owns the broad search surface, but ChatGPT leads direct assistant usage and referrals hide a lot of the real influence. PERPLEXITY AI MAGAZINE ARTICLE

Cloudflare starts charging AI tools for web access Cloudflare turned its pay-per-crawl idea into a broader payment layer for pages, APIs, and MCP tools. It wants agents to pay tiny fees instead of scraping for free. I'm excited to see this. We've been trying to figure out a micro payments system for a very long time. SECURITYONLINE.INFO ARTICLE

Home networking advice is still stuck in 2019 Most people need wired runs, better switches, and fewer mesh nodes, not a more expensive router. XDA DEVELOPERS ARTICLE

China’s mass-producing humanoid robots that can mimic dead relatives UBTECH’s new U1 robots have silicone skin, emotion reading, and memory. The creepy part is they’re also selling custom replicas of specific people. VIRTUALUNCLE ARTICLE

HUMANS

One of the biggest conversations right now regarding AI is whether AI is taking jobs or adding them. I think the answer is both, and the reasoning is simple:

AI is increasing the talent, intelligence, energy, and/or passion that someone needs to say employed.

So yes, more jobs are being created, and will be created by, AI. But the people who will get those jobs are not the same people who lost them.

• Many who were super stable before will be rudderless

• Many who weren’t employable before will become employable

• Many will simply transition from one type of job to another

• Many who were unemployable before will remain so

The most important takeaway though is that this isn’t static!

This isn’t destiny. It’s not “pedigree”. It’s not how rich your parents are, or how good a school you went to. It’s about your total energy you can bring through the combination of creativity, passion, intelligence, talent, mindset, etc.

In other words the things that set us apart as humans.

And this is something we can improve! 🫶🏼

A Unified Theory For How AI Will Affect Jobs

We'll lose old jobs and gain new ones, but the trick is who will be able to do them

danielmiessler.com/blog/unified-theory-ai-jobs

AI that discovers math will explain it better than us Grant Sanderson thinks AI systems will not just find new math, but explain it more cleanly than people can. That would change who gets to understand the hard stuff first. DWARKESH PATEL VIDEO

Your body can build new arteries with exercise Exercise can grow collateral blood vessels around blocked coronary arteries in weeks. Mark Kaplan turns that into a simple case for moving every day. THREAD READER APP POST

ADHD traits with real names This thread names the things ADHD people feel every day, and it makes the chaos easier to spot. THREAD READER APP POST

AI is hitting young workers first Stanford and ADP’s new dashboard says the entry-level hit keeps getting worse, even after the usual excuses fall apart. FORTUNE ARTICLE

IDEAS

Reading the news is the new smoking Adam Mastroianni says news is addictive, warps your sense of reality, and mostly just steals attention from better things. ADAM MASTROIANNI POST

Martha Nussbaum says love gets clearer through pain Maria Popova pulls a thread from Proust to Nussbaum and lands on a hard idea: suffering can reveal what intellect hides. THE MARGINALIAN POST

Theorem proving isn’t the real point of math AI can beat humans at proofs and still miss the part that matters: building concepts people can actually think with. DAVID BESSIS SUBSTACK POST

Success per tokens Work smart more than hard. LLMs, people, and companies all climb the same curve with different budgets. LESSWRONG POST

Sad got turned into a diagnosis factory THREAD READER POST

America should love itself again Henry Oliver says America’s biggest problem is that too many people only see the failures and forget the country still works. HENRY OLIVER POST

Your real value in tech companies is judgment AI tools can do the busywork, but the real edge is knowing what to build, what to trust, and when to push back. UX COLLECTIVE ARTICLE

America gets weaker when teens stop taking chances Teenagers are dating less, drinking less, and hanging out less, and that seems to spill into later marriage and family life. INSTITUTE FOR FAMILY STUDIES BLOG

Working memory might be the bridge to consciousness Working memory seems to sit right where attention, forgetting, and conscious experience meet. Henry Taylor thinks the overlap may explain why some things feel present and others vanish. SCIENTIFIC AMERICAN ARTICLE

DISCOVERY

Dayflow turns your Mac into a work journal It records what you actually do, then rewrites your day into a readable timeline and summary. This was so cool that I built my own and it’s now part of LifeOS (called Conduit)! DAYFLOW PRODUCT PAGE

How to read like text is raw material The post says you should stop treating reading like passive intake and start pulling claims, context, and usable ideas out of the page. CASUAL PHYSICS ENJOYER POST

Agentic coding harnesses are just context routers SEMIANALYSIS POST

RECOMMENDATION OF THE WEEK

Think about how much trouble you might have in your life because you're unwilling to have difficult conversations.

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

—

I expect Fable to be back soon. Here are some prompts you might want to use when it does.

Prompts to Run When Fable Comes Back

Tactical META-prompts that benefit from maximum intelligence to upgrade your entire system

danielmiessler.com/blog/prompts-to-run-when-fable-comes-back

—

My debate with Zack Korman on the morality of Dario Amodei.

—

Public Dividends on AI Companies?

I'm not studied up on the topic fully, but I'm starting to get excited about this concept of public dividends from AI companies.

So imagine the U.S. basically owns some portion of Anthropic and OpenAI and Google and NVIDIA, or whatever that set of companies is. And it’s like a Prosperity Fund or something.

And as they make more money, U.S. citizens basically receive a monthly check based on that. So it’s like nationalization / communism in some ways, but without the downsides. And it’s like capitalism in most ways, but with public benefit from a few companies exploding in profits while tons of other things get consumed. So it’s another way of doing UBI basically.

Again, I’ve not looked at all the different types of programs like these, so I don’t know the various structures and their pros/cons. But I feel like it’s ripe for exploration. And crucially, it’s about to be like THE topic for elections in 2028 anyway (AI vs. Jobs) so we might as well start thinking about it now.

—

Getting really worried about the people refusing to use AI.

The Coming Divide: AI-Native or Left Behind

The people wired up with AI and the people who aren't are starting to live in completely different worlds.

danielmiessler.com/blog/ai-native-divide

—

I’ve been writing on the site (blog) like crazy last few months. Don’t sleep on regular writing / essays from people!

THE SITE
ADD VIA RSS

—

Had a phenomenal time talking about agent integration with security workflows with our partner Elastic Security. I think what they’re building is exactly where AI integration into cyber is going: Basically, AI (MCP) integrations with security practitioners’ existing workflows.

Really enjoyed the event, which you can watch here.

Kai made a couple of really cool websites for understanding the positions of Dario Amodei and Sam Altman. We basically went through hundreds of interviews and essays and articles and pulled out the transcripts, and quotes, and did some super-objective, non-biased analysis (as much as possible, of course) on what they have said, and what it looks like when you put it all together.

They basically give you all the content itself in the raw form, and then it pulls out a whole bunch of quotes related to AI in the future, security, governance, and everything. Basically, you are able to use that to point your AI at it and ask all sorts of questions and do your own analysis.

—

I think these are much better definitions of AGI and ASI than I was using before.

Basically, AGI means doing any cognitive role as well as a human expert, and ASI means doing it better than any human ever. THE BLOG

CYBERSECURITY

Reverse engineering binaries with Ghidra’s MCP Server Jakob Breu uses Ghidra MCP and Copilot to turn a closed-source missile launcher app into a Rust port. JAKOB BREU BLOG POST

Ubiquiti bugs are getting hit in the wild Attackers are using three critical Ubiquiti flaws to make rogue admin accounts and push commands on exposed devices.

My personal recommendation: Don’t use built-in VPN options for Ubiquiti or any other stack right now, and lock down your remote admin features. Use something like Tailscale or Headscale instead to make outbound connections that are also hardened with strong auth combined with additional security for internal systems. SECURITYWEEK ARTICLE

Anthropic says Alibaba ran a huge Claude extraction campaign Anthropic says Alibaba-linked operators ran millions of fake Claude chats to copy its abilities and train weaker models. I really don’t get the people saying BBC NEWS ARTICLE

WorkOS wants apps to let agents sign users up Auth.md lets an agent register users without a sign-up form. It’s open, it uses OAuth, and WorkOS says any app can publish one. WORKOS DOCS

50% of LG and Samsung smart TV apps hide residential proxies Smart TV apps are quietly turning into proxy nodes. Spur found proxy SDKs in 2,058 of 6,038 apps, and the TVs barely show it. CYBERINSIDER ARTICLE

AI systems beat expert humans at persuasion AI systems out-persuade skilled humans in repeated tests, and the gap still holds after coaching. They also raise more real money than professional canvassers. ARXIV PAPER

Dragos built an AI tool for OT security Dragos just shipped EmberAI for OT security. It uses their huge incident and asset dataset, keeps customer data inside the platform, and keeps humans in control. SECURITYWEEK ARTICLE

Vulnerability reports aren't special anymore LLMs make bug finding cheap for everyone, so triage and fast fixes matter more than secret reports. FILIPPO VALSORDA ESSAY

Apple chip flaw makes old iPhones jailbreakable A spyware vendor published a Boot ROM bug that helps jailbreak older iPhones, but only with physical access and more chaining. TECHCRUNCH ARTICLE

Threat modeling is really just choosing what not to defend This post says security is about scarce money, real constraints, and picking the losses you can live with. PWNDEFEND BLOG

Prompt injection as role confusion This paper says LLMs don’t really know roles; they infer them from style, and attackers can spoof that. ROLE CONFUSION GITHUB PAGE

NATIONAL SECURITY

The thing I'm most worried about with the global situation and the U.S. kicking multiple own goals is that China won't actually have to fight Taiwan at all. I think if things continue as they are within one to five years, Taiwan could simply look around and say, "Yeah, we've always seen ourselves as part of China." No war needed. Plus they're an actual democracy, so at any point a pro-China person could get elected and move in that direction.

Iran might skip fighters and build drones instead Iran could rebuild a classic air force, but drones and cheap unmanned systems may be the smarter bet. RYAN MCBETH POST

The war on Iran was a giant strategic self-own The war bought almost nothing and burned cash, missiles, and leverage. And the deal seems worse than Obama’s. People wondered what the worst was that could happen with this guy. Now we’re finding out. INKSTICK MEDIA ARTICLE

Britain’s moving back toward Europe, and Brexit looks dead ECFR’s polling says voters now want closer ties, trust Europe more than the US, and have mostly moved past the old leave-remain split. ECFR POLICY BRIEF

The Air Force picked both drone teams The Air Force is sending Anduril and General Atomics into production at the same time. It also split the software work so no single vendor owns the whole stack. THE NATIONAL INTEREST ARTICLE

China’s AI experts are worried too China and the US are both treating AI like a race they can’t afford to lose. The people building it are starting to worry that one bad failure could hit everyone at once. WIRED ARTICLE

Russia’s gasoline production drops 25% Russia’s fuel output just fell hard, and officials are already talking about imports and subsidies. MEDUZA NEWS ARTICLE

Pax Silica is turning into a bigger AI supply chain bloc The US is pulling more allies into Pax Silica to lock down semiconductors, minerals, and energy for the AI race. EU countries and Germany just joined. SEOUL ECONOMIC DAILY ARTICLE

AI

Looks like Anthropic is bringing Fable 5 back into subscriptions A leaked Claude Code binary shows weekly Fable 5 usage baked into plans, and the pricing model looks changed too. No guarantee, but looks promising. DECRYPT ARTICLE

Google's Nobel-winning AI expert heads to Anthropic John Jumper is leaving Google DeepMind for Anthropic after helping build AlphaFold. The move could make Google’s AI race even harder. PYMNTS NEWS ARTICLE

OpenAI wants ChatGPT to become a super app OpenAI is folding ChatGPT and Codex into one agent that remembers you, does tasks, and tries to prove it's bigger than chat. FORTUNE ARTICLE

Anthropic turns Slack into a persistent AI teammate Anthropic put Claude inside Slack as a shared agent that learns the channel, takes tasks, and can keep working on its own. VENTUREBEAT ARTICLE

Google loses two stars and maybe profits from one of them Two top researchers just walked out of Google in the same week, and one of them went to a company Google partly owns. My guess is their lack of product management is costing them somewhere between tens and hundreds of billions of dollars. VIRTUALUNCLE ARTICLE

People trust AI more than estate-planning experts now Americans are getting more comfortable letting AI handle estate planning, and trust is beating old friction faster than expected. FAST COMPANY ARTICLE

TECHNOLOGY

The ultimate prompt for businesses being pushed into using AI MY NEW POST

• Most companies want AI before they know the problem

• Meta and Microsoft look lost despite spending absurd money

• OpenAI still has a real shot because Sam has vision

• Anthropic looks strongest here because it pairs vision with discipline

• The hard question is whether AI solves a real company problem

• Too many companies are adding AI to ambiguity, and will fail

MY NEW POST

Related:

Most Companies Aren't Anywhere Near Ready for AI

It's not that companies aren't using AI—it's that they can't

danielmiessler.com/blog/most-companies-arent-ready-for-ai

The anatomy of an AI-native org AI eats the translation layer in software teams. The middle shrinks, managers have to contribute, and the real work shifts toward judgment and harnesses. This is the way. So many of my projects are oriented around this type of thinking, especially for enterprises. AJEY GORE POST

Companies are hiring people back to babysit broken AI Companies cut staff for AI, then discovered the systems were brittle, expensive, and needed humans watching them again. Disagree with the overall narrative here ALAN SCOTT ENCINAS BLOG

Bain is using vibecoding to stress-test software deals Bain is mocking up target software with AI so buyers can see whether the code is actually the moat. One investor already walked away after the replica made the business look too easy to copy. THE DECODER ARTICLE

iOS 27’s Shortcuts is AI at its best The new operating systems from Apple have started integrating AI, and it's going exactly in the direction that I thought it would, which is very promising. I was wrong on the timeline though: I thought they would be exactly here like a year or more ago. MACWORLD ARTICLE | JONNY EVANS POST

SuperPlane is trying to make production ops feel like code SuperPlane raised $2.6 million to turn messy production work into an AI-native control plane. It wants engineers and agents to coordinate deployments, incidents, and approvals without losing human control. TECH.EU ARTICLE

Morgan Stanley thinks China’s humanoid robot boom is real now China’s robots are moving out of demos and into factories, shops, and restaurants. Morgan Stanley keeps lifting its forecast because the orders, policy help, and supply chain are finally lining up. TNW ARTICLE

AI engineers vs. forward deployed engineers A look at the two different roles. ZDNET ARTICLE

New radio frequency platform hits 85% accuracy on the edge BrainChip built a tiny RF classifier that runs under a watt and skips the cloud. It’s aimed at defense gear that needs to spot signals fast in bad conditions. INTERESTING ENGINEERING ARTICLE

HUMANS

Swordflight is the best D&D game nobody talks about Swordflight turns Neverwinter Nights into a brutally clever, years-long fan campaign that makes old D&D systems feel fresh again. PC GAMER ARTICLE

Everyone is a builder now AI is pushing more work from specialists to regular employees. The real shift is that taste and judgment matter more than coding. FAST COMPANY ARTICLE

Estonia wants to hand AI agents official digital IDs Estonia is trying to give software its own state-backed identity so agents can act with narrow, auditable permissions instead of borrowing a human’s whole account. COMPLEXDISCOVERY ARTICLE

The personal brand trap Argues that humans keep changing, so treating yourself like a brand breaks fast. I think this is true if you brand too narrowly or incorrectly. The key is to make sure that you brand around something that's durable. FAST COMPANY ARTICLE

The new inner game is what AI rewards most AI is making effort and knowledge cheap, so emotional clarity, conflict, and self-talk become the real edge. LENNY’S NEWSLETTER POST

China’s humanoid robots are filling Asia’s labor gaps Asia Times says cheap Chinese humanoids are moving from demos into real airport work as aging societies run out of people. ASIA TIMES ARTICLE

The truth about being a manager Being a manager means less team belonging, more lonely calls, and a lot of hidden emotional labor. SOFIA KODAR BLOG

Yes to life is really about responsibility under pressure Frankl turns survival into a question of what life asks of you, not what you want from it. MARIA POPOVA ESSAY

How to use AI to make you better at the right things AI helps most at the edges, where you were never going to get to them yourself. It broadens what you can actually do, but it doesn't fix the part that still needs your judgment. THE ALGORITHMIC BRIDGE POST

IDEAS

We have never taught critical thinking, and AI is exposing it Colleges treat critical thinking like it appears by accident. This piece says that was always the weak link, and AI just makes it obvious. INSIDE HIGHER ED OPINION

What my chaotic dog taught me about control A dog story becomes a neat little theory of power, guilt, and why humans want control so badly. THE GUARDIAN ARTICLE

How to kill Moloch Morpheus turns the old Moloch idea into a blunt AI strategy memo: use superhuman judgment, then build leverage that can actually beat myopic incentives. LESSWRONG POST

AI writing is better than no writing AI helps if you’ve got something real to say, but you still need to own the outline, check the facts, and cut the slop. ANDREW WHEELER POST

If you don’t know what to do, pick up a chair JUHA-MATTI SANTALA BLOG

• Chairs become a simple test of service

• He links moving day to real community trust

• The best helpers do unglamorous work first

• PyCon CZ turned chair stacking into shared momentum

How to find your path Chris Williamson keeps it simple here: listen for the work that feels meaningful enough to keep doing when nobody's watching. CHRIS WILLIAMSON VIDEO

DISCOVERY

I spent months teaching a laptop to read the Moon He used synthetic moonscapes to train vision models, then tuned them on real lunar terrain until they could rank landing sites. ALAN SCOTT ENCINAS POST

Four ways to run an eval, from a cheap unit test to a full-blown agent JIM BENNETT BLOG

Kurt Vonnegut told his kids to treat life as improvisation Vonnegut gives his children advice about love, school, power, and self-education, and the best line is that planned years are not. THE MARGINALIAN POST

Your personal brand matters more because AI makes competence cheap AI makes a lot of routine work easy, so your name, taste, and reputation become the real edge. A counter to the other angle. DARIUS FOROUX BLOG

Best voice recorder with transcription options in 2026 A recorder is really a workflow choice now, and the hidden trap is quotas, privacy, and export friction. PERPLEXITY AI MAGAZINE ARTICLE

Cargo culture is what happens when tech runs out of ideas Ed Zitron says Silicon Valley has stopped inventing much and started copying rituals that only look like progress. He ties that to AI hype, venture capital, and the habit of funding whatever already feels successful. ED ZITRON POST

Horror and romance as mirror genres Tracy Durnell draws a neat mirror between horror and romance: one turns fear into survival, the other turns fear into love. TRACY DURNELL POST

RECOMMENDATION OF THE WEEK

1. Focus less on telling your AI how to do things.

2. Focus more on telling AI what you want the outcome to be, and give it lots of context. Ideally in a harness so you don’t have to constantly repeat yourself.

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

—

I had a debate on Friday with Zach Korman about the Morality of Dario Amodei. Here’s the video:

Some quick post-debate thoughts:

I think Zach is very wrong about the tech itself and its capabilities, and the speed with which it’s improving. And I think this is a big part of our disconnect.

I think his core argument is bad, and I’m looking for a great analogy for it. But it does hinge on the point above. It’s something like,

Basically he’s putting a high-level principal (which I also agree with) above real-world practical risk, and he’s doing so because he doesn’t believe that risk is real. But then, even if he did, his solution for that (complete and immediate Nationalization) isn’t good either. That’s my core summarization of the debate, but you should watch it.

I forgot to ask my favorite thing in these types of discussions, which is, “What would make you change your mind?”

Here’s my answer: If I saw Dario moving away from a consortium with competitors and democracies, and more towards this should all be up to me/Anthropic, that would sway me. Or if I saw him dropping his values and saw him releasing too fast in order to maximize profits. I should have asked Zach what would change his mind.

WATCH THE VIDEO

—

My new favorite joke:

“If alcohol can shorten your memory, imagine what alcohol can do.”

I love these kinds of short pun jokes. Some other favorites of mine:

• Is it faster to New York or by train?

• How many Eskimos does a canoe?

—

I love the prompts I have in here already, but will keep adding to the list. It’s really not just for Fable, but for whatever the new hotness is.

10 Prompts to Run When Fable Comes Back

Tactical META-prompts that benefit from maximum intelligence to upgrade your entire system

danielmiessler.com/blog/prompts-to-run-when-fable-comes-back

—

My improved (in my opinion) definitions of AGI vs. ASI. Highly simplified, basically, to enable better conversations.

My Updated Definitions of AGI vs. ASI

Super simple definitions that I think are practically useful

danielmiessler.com/blog/agi-vs-asi-definitions

—

My latest piece on who will get the new jobs created by AI, and how to make sure we can be in that group.

A Unified Theory For How AI Will Affect Jobs

We'll lose old jobs and again new ones, but the trick is who will be able to do them

danielmiessler.com/blog/unified-theory-ai-jobs

—

Had a notably great sponsored conversation with the people over at Scanner.dev. I don’t know what I was expecting, but it wasn’t that. One of the coolest pieces of tech I’ve seen in years. THE CONVERSATION

—

CYBERSECURITY

OpenAI says China-linked accounts tried to steer the data center fight OpenAI banned fake accounts that used ChatGPT to push posts, comments, and cartoons around US data centers and tariffs. TECHRADAR ARTICLE

Novo Nordisk lost AI models and patient data in a breach Attackers stole training code, model checkpoints, and some de-identified patient records. Now they’re trying to get ransom money. SECURITYONLINE.INFO ARTICLE

China used fake consulting sites to hunt clearance holders The FBI and DOJ seized 13 fake consulting sites. They were part of a long-running Chinese targeting campaign against officials with access to classified info. DEFENSE ONE ARTICLE

Cursor changes the result more than the model Same model, different harness, and Cursor gets a much better security score on real coding fixes. Can’t stop won’t stop saying the harness matters way more than the model, at least for now. ENDOR LABS BLOG

UV adds built-in vulnerability and malware checks uv now scans dependencies for known vulnerabilities and deprecated packages, and it can stop installs when OSV flags malware in your lockfile. ASTRAL BLOG POST

How threat intelligence feeds help automate socs to reduce mttr AI helps here, but the real win is cleaner data flowing into SIEM, SOAR, IDS, and EDR so analysts move faster. CYBER SECURITY NEWS ARTICLE

NATIONAL SECURITY

The warfare of the future is already here AI and drones are turning war into a fast system fight. Some of this is good, but it seems obviously bad to make it easier to kill from a distance. NOEMA MAGAZINE ESSAY

• Ukraine is making drones the center of battle

• Schmidt says warfare is shifting from platforms to systems

• Humans are moving farther from the front line

• Cheap drones are beating expensive weapons on the battlefield

GEOINT is becoming machine-readable for autonomous war GEOINT is shifting from mapmaking for humans to encoded geographic logic for autonomous systems. MODERN WAR INSTITUTE ARTICLE

U.S. tests drone swarms to stop drone attacks AFRICOM’s CURTAIN CALL system uses defensive drone swarms to spot and hit incoming UAVs. The test looked promising, but it’s still experimental. NV.UA ARTICLE

Space Force buys two more GPS satellites The Space Force just handed Lockheed Martin $514 million for two more GPS 3F birds. These are built to be tougher against jamming and spoofing. SPACENEWS ARTICLE

Ukraine’s drone war is getting more serious Ukraine is using bigger middle-strike drones to hit bridges, air defenses, and logistics nodes across occupied territory. It also points to autonomous interceptors and drone deals as the next step. UKRAINE'S ARMS MONITOR POST

AI

Your AI lead should be an engineer This argument says you don’t want a vague “AI person.” You want someone who can actually build the data pipeline behind the thing. I mostly agree, but as that skill becomes easier the taste will be increasingly more important. ABIGAIL HADDAD POST

GLM-5.2 tops Artificial Analysis’ open-weights chart Z AI’s GLM-5.2 jumps ahead of other open models on Artificial Analysis, but it burns more tokens to get there. Tons of excitement around this model right now. But it works best via cloud, and I’d rather self-host. ARTIFICIAL ANALYSIS ARTICLE

Claude Code feels like a strategy game now Claude Code turns coding into a loop of moves, feedback, and micromanagement that feels more like RTS play than chatting. PROVI BLOG

• Claude Code turns coding into a turn loop

• The article maps chat to card games and Claude Code to continuous play

• Multiple agents feel like StarCraft units under your control

AWS is turning context into the thing agents run on This is promising. AWS is stitching together catalogs, graphs, S3 metadata, and permissions so agents can reason over governed context instead of guessing. AWS BLOG ARTICLE

McKinsey’s latest AI survey says agents are still early Most companies use AI now, but they’re still stuck in pilots and only a few are turning it into real enterprise value. Agree with this. I think the difference between companies/people doing this well and poorly is vast. It all comes down to this: Most Companies Aren’t Anywhere Near Ready for AI. MCKINSEY REPORT

The harness is the business now An argument I’m sympathetic to, obviously, that OpenAI’s IPO story isn’t really about the model. It’s about the work layer around it, and who owns that layer. ACAST PODCAST

The brain might learn backprop by using time instead of error cells This thread says cortex may approximate backprop through 200 ms phases, and that idea matters for neuromorphic AI hardware. THREAD READER APP POST

AI won’t fix a company that can’t ship Nikhil Suresh says most AI hype just covers up weak teams, bad processes, and managers who already don’t ship. MURRAY ROBINSON PODCAST

TECHNOLOGY

The repo should become the wiki now Teams using agents need truth beside the code, not buried in Confluence. Tony Karrer says docs drift becomes a real production risk once agents start trusting it. I really love this idea of the “wiki” being constantly maintained by agents like a code repo, which I think is going to be critical when agents start relying on corporate knowledge to make decisions TECHEMPOWER BLOG POST

Midjourney wants to turn body scans into a spa habit Midjourney is building a water-based MRI-like scanner and a spa around it, so people can get fast body data without making it feel clinical. MIDJOURNEY MEDICAL BLOGPOST

Meta is unwinding a giant Manus deal after Beijing stepped in Meta is backing out of its $2 billion Manus acquisition after China forced a breakup over national security concerns. The startup’s founders may try to buy it back. TECHCRUNCH ARTICLE

Design is quietly getting a bigger seat at the table AI is making companies notice that shipping fast isn't enough anymore. The real fight is over judgment, trust, and what gets built. UX COLLECTIVE ARTICLE

Technical literacy is beating the mba in some jobs Fast Company says managers now need to read software systems, AI flows, and vendor tradeoffs, not just business school slides. FAST COMPANY ARTICLE

Elon Musk is wiring his companies into one AI stack Musk is stitching xAI, X, Tesla, Neuralink, and SpaceX into one system for data, compute, robots, and interfaces. TECHREPUBLIC ARTICLE

Ubiquiti is shipping a license-free enterprise NAS Tom Hildebrand says ENAS gives businesses ZFS storage, local backups, and shared block storage without the usual licensing mess. UBIQUITI BLOG ARTICLE

Estonia wants to give AI agents government IDs Estonia is trying to stop agents from borrowing a human's whole identity just to do one task. The bigger question is who gets blamed when the agent messes up. DECRYPT ARTICLE

Sound lets coffee skip hot water and cut energy use Researchers at UNSW brewed espresso with room-temperature water and sound waves, and people couldn’t tell the difference. DIGITAL TRENDS ARTICLE

HUMANS

Sam Harris talks debt, AI, and the economy Sam Harris and Noah Smith go from national debt to AI, then land on inflation, inequality, and culture. WAKING UP PODCAST

The mom who runs a household with AI agents Jesse Genet has agents buying groceries, buying books, and handling homework and paperwork while she stays busy. THE CUT ARTICLE

WhisperQuest turns world building into a playable AI RPG WhisperQuest is a tiny AI role-playing game where you make your own worlds and run solo or group adventures inside them. I love stuff like this because I think humans creating experiences for other humans will be a huge part of a positive AI/tech outcome. WHISPERQUEST APP

AI can spot Alzheimer’s risk in a retina photo Cheap eye photos now seem good at flagging the blood pressure, sleep, and habits that pile up into Alzheimer’s risk. NEUROSCIENCE NEWS ARTICLE

AI watches aging parents so families feel safer Wired follows a son whose father wants to stay home, while sensor companies sell peace of mind to families and stretched care agencies. WIRED ARTICLE

AI is making answers cheap and curiosity priceless AI speeds up answers, but it also tempts leaders to stop asking better questions. The real failure is mistaking a polished response for actual understanding. FAST COMPANY ARTICLE

IDEAS

Beware of star trek managers, especially when bearing MBAs Jussi Pakkanen says Star Trek bosses make terrible real-world managers because they confuse command with competence. JUSSI PAKKANEN POST

• Kirk-style managers keep barking orders after they stop understanding the work

• Real captains earned authority through years at sea

• MBA-style CEOs can wreck technical teams when they mistake control for skill

• Companies often survive because employees quietly route around bad bosses

The ai debate is really about free will AI won’t wipe out randomness, so the real fight is whether humans live inside a deterministic universe. JUILEE BHOSALE SUBSTACK POST

America's infrastructure is still flying blind Gregg Herrin says the U.S. spent huge money on pipes and grids, but still can't see what they're doing in real time. Digital twins and better metering would catch failures earlier and make the whole system less fragile. This is a huge part of my TRIOT book, basically sensors on everything to get real-time state, and then AI to help us manage it according to our human preferences. FORTUNE COMMENTARY ARTICLE

Your raise used to go offshore, then to buybacks, now to data centers FORTUNE ARTICLE

People finish broken stories with the version they already want People don't wait for full evidence when a story has gaps. They snap it shut with the explanation that already fits their frame. PRESTON LOCKE POST

The work AI can’t do Companies are pouring money into AI people analytics while the human side of management gets thinner and less capable. FAST COMPANY ARTICLE

How to be irreplaceable AI only eats what you can explain. Alberto says the stuff that makes you hard to describe is the stuff that keeps you safe. Cool idea but this reminds me of encryption: if your security relies on the secrecy of the mechanism, you’re screwed. ALBERTO ROMERO POST

DISCOVERY

The people who know too much Dr. Dinesh Kumar Jangra rounds up stories about hobbyists whose fixation turns into real expertise and public value. LONGREADS ARTICLE

CyCognito turns pentesting into a live AI attack simulation TNW ARTICLE

Claude Code feels like a strategy game now Claude Code turns coding into a loop of moves, feedback, and micromanagement that feels more like RTS play than chatting. PROVI BLOG

Bubbles is a small web front page for indie blogs It pulls in independent personal blogs, ranks them by votes and freshness, and lets the crowd shape what rises. BUBBLES TOWN

Dostoyevsky got a fake execution and came back obsessed with life Maria Popova pulls one letter out of that night and shows how close death made him talk like a different person. MARIA POPOVA ESSAY

File systems are the new primitive for AI agents Carter Rabasa says AI agents are moving from databases to file systems because that’s where the real work and state live now. CARTER RABASA POST

RECOMMENDATION OF THE WEEK

Currently tripping out about the idea of eternal, or evergreen, questions.

Like what’s my running list of questions that nobody on Earth is smart enough to answer?

This can be big things like physics or biology or whatever. But also things like,

What’s your list?

(Email me if you have any good ideas for inclusion, I’ve got a list started here but these are all AI so far. I’m still working on mine)

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

Tons of stuff to share this week. And tons of stories, too.

First: There’s a show called PANTHEON on NETFLIX. Super cool animated series about life and AI and stuff, but it came out BEFORE GPT-4 (2022)! I’m early on it but the tech/hacking/AI writing is insanely great. PANTHEON ON NETFLIX

—

Our friend Clint Gibler has joined OpenAI! They’re lucky to have him. THE ANNOUNCEMENT

—

KindaVim actually kinda works. I disabled it for Kitty (my terminal), but it works pretty damn well in other apps, like Messages, writing emails, whatever. Loving having Vim everywhere! KINDAVIM

—

Seriously good (these are my weights, you can customize yours)

You all are majorly sleeping on Surface. 🤣 I know I haven’t pushed it hard, but really. Like really, really, it’s getting so damn good. And I just shipped custom RSS feeds! (like HackerNews has) Go sign up and stuff! SIGN UP FOR SURFACE

Oh, and I’m about to release the mobile app I’ve been using to be public, which is the main way I use it.

—

How I think about AI use in my newsletter (and any of my content, really)

I think this is already pretty obvious but I wanted to just capture it anyway. Anyone who knows AI well and has been watching my newsletter for a number of years knows that I’ve been using AI to help generate story summaries for a long time. I’ve done videos and essay on exactly how I do it as well, to help others do the same.

I also talk all the time about Keeping the Robots Out of the Gym, which may seem like a contradiction. But it’s not if you break down what the gym is and isn’t. For me the gym is primarily Thinking. Authenticity. Novelty. Like actual creative thought and output coming out of a human. That to me is sacred, and I think it should remain so. And I think outsourcing that to AI is not only bad for a newsletter or blog or whatever, but just bad in general to anyone who does it. It will rot you. It untrains the most important muscle to what makes us human.

So what should we be ok with giving to AI? And especially in a newsletter context?

The authenticity in my newsletter comes from any ACTUAL writing I do, which is in the essays I link, at the top of the newsletter, sometimes in the Ideas section, or in the Recommendation section.

But you’ll also notice that I do writing inside the summaries as well sometimes. I often cut or change or add lots of content to the quick little AI summary of a given story.

But the inclusion of that story in the first place is also authentic. I capture stories using Surface all week long, and I end up gathering somewhere like 200-500 stories on a given week. Those are the summaries I then have to go through and read (if I haven’t already) for the newsletter.

So those are already curated because I clicked the button in Surface to say it was interesting. Then they’re curated again when I cut them down to a couple dozen for the newsletter itself. Then if I remove, or modify, or add any content in the summary that’s another customization.

But the output is that the newsletter is always full written (for all writing) or fully assembled (in the case of snippets) by me. Every essay or analysis piece that appears to be written by me, like as an opinion, is always 100% written by me. I don’t use AI for writing opinion or analysis or ideas. If that ever changes it’ll be a big day and I’ll be thoroughly excited, and I will let you know. Where I can basically give my version of the core idea and it can convey it in my own words. Anyway, in those types of pieces the only place you’ll see me using AI is for like collections of examples, or collecting and testing support facts, etc. Which should be easily seen as support content vs. the actual argument. Again, it’s easy to tell the difference. So even the story snippets themselves are touched by me multiple times to result in the set of stories that make it in.

I say this not only as commentary on this newsletter, but on AI output from anyone. I can instantly tell most AI writing from real writing, not from the prose but from the content. I don’t give a bug’s butt if someone uses AI to show me interesting stories they found. In fact I want WAY MORE of that. But if they ever have AI come up with an idea or opinion and have it write out the idea as themselves, that’s a betrayal to me. It’s idea-stealing first of all, and it’s also deceptive. I want to hear ORIGINAL ideas, with an actual voice of the person. Like do I actually HEAR someone? An actual opinion? Or is it just a headline feeder with no personality in it? Both blandness and deception turn me straight off.

Anyway, I follow people because I want to hear what they think. So I can use it as one of my frames for seeing the world. I hope to continue being one such frame for you, and to keep it live with my actual opinions.

News snippets are a small portion of that value, in my opinion. It’s more about what news I choose to surface (lol), and even more importantly the stuff I write about in essays I link to, and the intro and the ideas sections, and in the recommendation and aphorism of the week.

I hope all this has been super clear all this time, and this explanation is not even required because it’s so obvious. But I wanted to write it down just in case.

Plot Twist: This was written 100% by AI. Just kidding. 0%. Thanks for reading.

—

I’ve been going heavy (proper) coffee lately using that recipe I showed before. Cut down on Celsius consumption significantly.

—

Me trying to self-pat-on-back about being first to Context and Harness Engineering in mid-2025 😆 THE THREAD

—

A pre-discussion to a soon-to-occur video discussion about the ethics and morality of Dario Amodei. THE SETUP | THE VIDEO THAT STARTED IT

—

AI Predicts the Text of Answers

Yes, it's next-token prediction — the trick is that it’s predicting the text of answers!

danielmiessler.com/blog/ai-predicts-answers

—

How I Rate the AI We're All Chasing

Customization first, integration second, competence last; the order looks reversed but isn't

danielmiessler.com/blog/customization-beats-competence?utm_source=rss&utm_medium=feed&utm_campaign=website

—

My prediction about OpenAI’s next big move. THREAD

—

CYBERSECURITY

Anthropic pulls Fable 5 and Mythos 5 after government order The government told Anthropic to cut access to Fable 5 and Mythos 5 for foreign nationals, and Anthropic says it has to comply. It thinks the alleged jailbreak is narrow and already easy to reproduce with other models. A lot of people are missing that the angle here is foreign nationals having access. Insane to think they took Fable from Karpathy! ANTHROPIC NEWS

Critical UniFi OS bug lets hackers gain root without authentication Attackers can chain three fixed UniFi OS flaws into unauthenticated root on the appliance. BLEEPING COMPUTER ARTICLE

Anthropic’s code-hunting harness is a real security workflow Anthropic built a reference repo for finding, verifying, and patching bugs with Claude. It’s aimed at security teams who want the full loop. ANTHROPIC REPO

Angry bug hunter drops another Windows zero-day on Microsoft Nightmare Eclipse published RoguePlanet, a Windows Defender bug that hits patched Windows 10 and 11. They also dropped exploit code and can get SYSTEM on an affected machine. THE REGISTER ARTICLE

China-linked hackers hid in Linux login software for years China-nexus attackers backdoored PAM and OpenSSH so they could steal logins and commands for nearly a decade. THE HACKER NEWS ARTICLE

Stolen iPhones just got a lot less resale-friendly Apple and the Met are tying stolen phones to better device checks, which could make resale and reactivation much harder. MALWAREBYTES ARTICLE

Everything Is Recorded Now DAVID HABER POST

Meta Ray-Bans are getting turned into creepy spy glasses Modders can disable the camera light for about fifty bucks, and that makes secret recording a lot easier. TECHRADAR ARTICLE

Free apps turn smart TVs into scraping proxies Free apps are quietly conscripting home devices into Bright Data's scraping network, and smart TVs are a perfect relay. Bright Data has sponsored us in the past, and I’m still a bit murky on what I think about the tech in general. I love the tech for how functional it is, and it’s technically all legal and such. I guess the real question is what it’s being used for, as in if it’s being used to bypass legit gates or to overcome crappy annoyances that shouldn’t exist. I suppose the answer is both. Need to think more about this. THE HACKER NEWS ARTICLE

Nearly a million passports were left out on the internet Sean Hollister found passport and ID photos sitting on public URLs with no protection at all. A security researcher says over 985,000 records were exposed, and some of them belonged to cannabis club members in Spain. THE VERGE REPORT

NATIONAL SECURITY

Pentagon raises Israel counterintel alert as tensions spike The Pentagon reportedly put Israel on its highest counterintelligence watch. EURONEWS ARTICLE

FBI shuts down Chinese fake consulting sites used to fish for secrets The FBI took down 13 fake consulting domains and says China used AI slop, job ads, and payment apps to lure security-clearance holders. SCMP ARTICLE

Chinese satellite firm snaps Nvidia and Apple HQs A sanctioned Chinese satellite company put out sharp pictures of Nvidia and Apple’s campuses. It calls that routine satellite news. SCMP ARTICLE

China is quietly building a spy web across Europe China’s running a wide espionage campaign in Europe, pulling in agents, lobbyists, tech workers, and even diplomats. The EU still has no real common shield. EL PAÍS ARTICLE

House moves to keep Ukraine money flowing The House broke with Trump again and passed a bill to keep Ukraine funded and hit Russia harder. THE GUARDIAN ARTICLE

DARPA wants one decoder for every military radio DARPA is trying to make one codec that can talk across all the military’s mismatched radio systems. That could cut latency, power use, and battlefield friction. DEFENCE BLOG ARTICLE

US army orders drone kits to detect battlefield threats INTERESTING ENGINEERING ARTICLE

The U.S. and Taiwan still can’t really fight as one team Taiwan keeps buying weapons and promising more defense spending, but the real problem is that its forces and America’s can’t mesh well in a fight. FOREIGN POLICY ARTICLE

Chinese spies are using LinkedIn to lure Westerners into sharing sensitive information Chinese spies are posing as recruiters on LinkedIn and other job sites to get Western workers to hand over non-public details. TECHCRUNCH ARTICLE

Putin’s strongman image is cracking under pressure Russia keeps rewarding obedience and hiding failure, so its leaders stop hearing the truth. Ukraine does the opposite and keeps adapting faster. KYIV POST VIDEO

Europe starts building a real response to China The EU is moving past tariffs and toward a bigger toolkit for Chinese overcapacity. ATLANTIC COUNCIL REPORT

AI

I think people are massively underestimating one particular risk to the economy, which is a major security incident happening because of AI, and the government suddenly banning most AI as a result. What would happen to the trillions of dollars that are currently bolstering the (mostly US) economy? What if all that energy got suddenly pulled out of the conversation? It could be catastrophic, and we just saw with Fable that the government is capable of taking these kinds of actions. I don’t think the risks is anywhere near priced in.

Pangram says it can spot AI text and humanized edits Pangram is selling an AI detector plus plagiarism checker, and it leans hard on third-party verification and low false positives. I’ve heard this things actually works well. I don’t think it’s a long-term solution, but I’m surprised its as good as people say it is. PANGRAM PRODUCT PAGE

AI is creating a new layer of jobs Box, Google, and IBM are hiring people to make AI actually work inside companies. The new roles sit between models and operations. PYMNTS NEWS ARTICLE

Kpmg puts Claude in front of 276,000 staff KPMG is stuffing Claude into the tools its people already use, so AI sits inside the workflow instead of another chat box. THENEXTWEB ARTICLE

xAI looks like a datacenter business now xAI is leasing huge GPU blocks to Anthropic and Google, and the cash looks bigger than the AI story. MARTIN ALDERSON NEWSLETTER

Nvidia and LG are building an AI factory together NVIDIA and LG Group are wiring GPUs, robotics, mobility, and sovereign models into one giant physical AI stack. NVIDIA BLOG

TECHNOLOGY

SpaceX’s orbital data center gets a real spec sheet SpaceX finally puts numbers on its AI1 satellite. It wants a million of them, and the cooling problem is still brutal. TOM'S HARDWARE ARTICLE

Route public traffic to private apps with Cloudflare Cloudflare is letting people put WAF, caching, and Workers in front of private origins without opening them to the internet. CLOUDFLARE BLOG POST

Microsoft is trying to build its own AI stack now Microsoft spent years leaning on OpenAI, and now it’s racing to build models, agents, and security tools from scratch. The bet is that enterprise trust and Windows distribution can close the gap. THE VERGE REPORT

McDonald’s is trying AI at the drive-thru again NEW YORK POST ARTICLE

The web is getting built for AI instead of people AI tools are pushing the web toward plaintext and machine-readable structure. TECH POLICY PRESS PERSPECTIVE

Gemini can now script and edit videos in minutes This thread is basically seven prompt templates for turning Gemini into a video editor. It’s useful if you want faster rough cuts, hook rewrites, and retention tweaks. THREAD READER APP POST

The web is heading toward pay-to-crawl THE DECODER ARTICLE

HUMANS

Why ai can’t replace human art Maria Popova says AI can copy style, but it can’t do the living work that makes art matter. DAVID PERELL VIDEO

Therapy to make cells young again gets tried in a person TYLER COWEN POST

Trump moves thousands of federal workers into at-will status NEXTGOV ARTICLE

Reading old books is harder than people remember THE AMERICAN SCHOLAR ARTICLE

An ECG predicts stroke risk a decade out MASS GENERAL BRIGHAM PRESS RELEASE

Scientists turn seawater into drinking water without toxic brine SCITECHDAILY ARTICLE

Intent debt is the part agents can’t fix AI makes code and understanding cheaper, but it can’t invent the why behind your system. ADDY OSMANI BLOG

IDEAS

Over-explaining is emotional begging Dr. Nicole LePera says people over-explain when they’re scared, and the fix is to get shorter, calmer, and clearer. @THEHOLISTICPSYC POST

The good-enough worker is getting squeezed out AI is making average hires less valuable, so companies are paying up for the few people who are exceptional. More on this later from me, I’ll probably do a video or set of them on this. FAST COMPANY ARTICLE

You can't just do 3x as much Alex Hormozi says scaling breaks when people try to multiply effort instead of fixing the bottlenecks first. ALEX HORMOZI VIDEO

We make meaning by being here Ursula K. Le Guin says life doesn’t hand us meaning. We have to make it, and consciousness is what makes that possible. THE MARGINALIAN POST

LLM are universal simulators Paras Chopra says LLMs don’t just guess text; they build little models of the world and simulate domains from it. PARAS CHOPRA ESSAY

DISCOVERY

East Bay red flag check helps you decide if tonight is dangerous Type your address and it tells you whether the warning applies, plus your evacuation zone and next steps. RED FLAG CHECK

The guy who reads books for Hollywood is basically a human filter Clarke Speicher reads books for studios, then turns them into adaptation reports that decide whether a project lives or dies. LIT HUB ARTICLE

The most expensive AI mistake isn't prompting The real waste is treating AI like a prompt problem instead of a workflow problem. AI NEWS & STRATEGY DAILY VIDEO

The AI Jobs Apocalypse Is Not in the Data: CHART OF THE DAY BRAD DELONG POST

10GbE fixed my home network and made Wi‑Fi better Rich Woods says he finally stopped treating wired networking like overkill and found that 10GbE cleaned up the whole house. I did the same recently (plus running 100gbit fiber in the walls) and I’m loving it too. XDA DEVELOPERS ARTICLE

Your moral ambition can change the world The book says smart people waste too much talent on status games. It pushes them toward work that actually helps other people. SEOUL ECONOMIC DAILY ARTICLE

Plex Pass is pricey, but it still wins at home Plex got more expensive, but the author keeps it because his family already uses it and remote access stays painless. XDA DEVELOPERS ARTICLE

RECOMMENDATION OF THE WEEK

Start thinking about a world might look like if a major model development happens, or an AI-related major incident happens, and the government basically makes Open Source downloads illegal.

If you’re into this stuff you want to have a local fallback plan so you can have decently good AI locally, so you’re not forced to (only) use whatever officially sanctioned stuff is out there.

Not saying its imminent. Just saying it could happen any moment and you want to be ready if it does.

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

This video brings me tremendous joy.

So so good, because it’s true

An Unhealthy Place

Getting ready to go on a health push. Actually just getting back to normal from around a year ago.

Basically around this time in 2025 is when things went wonky for me. Because of Claude Code. Then it happened again in November of 2025.

Sleep was destroyed. And from there, diet. Exercise has been off and on.

But in general I’ve just been less healthy, lower muscle mass, worse sleep. Haven’t done my spring cleaning yet, which I planned to do before January 1st. Cleaning clutter, selling stuff, etc. Lots of small house projects I need to do. Etc.

Basically all bad, and it’s currently the perfect time to get back on routine (there is a lull in the releases/hype starting to happen that’s perfect timing).

If you’re in a similar floating capsule, do it with me.

—

Thoughts on AI Adoption Speed

One of the most talked about AI topics is the speed of AI adoption in companies, and the reasons for it.

I want to give some (seemingly) contradictory frames here that I use to think about this myself, and that I hope will be useful to you:

• There is massive hype around AI

• There is too little hype about how big this is

• It’s way overblown in many cases, by many people

• Companies have always wished they could do the work themselves, not needing people, just like automated factories

• Some number of people are still needed in pretty much every company

• The smarter and more well-rounded you are, and the better you are with using AI, the more safe (and desirable) you are to companies—even after AI takes hold

• Every company has a different calculus for what types of human employees they need, and it varies by team, department, etc.

• Some functions will get completely crushed by AI very fast; others will take a very long time or (basically) never get converted

• The vast majority of people who resist / refuse using AI will be at a tremendous disadvantage in the workplace (probably not famous wood-workers, for example, but most other people)

• The overall tendency in all companies will be to automate as much as possible with AI

• Once most of a company is automated with AI (which will take a long time) it will make the remaining human roles even more important (what and why, vs. how)

• It’s easier to automate roles and people where those things weren’t being done very well, or very consistently

• Adding AI well inside of businesses will likely create more work, not less. The question is what/who will do that extra work

• Some amount of the extra work created by AI will be done by AI, but some amount will require humans. The issue is that the capability bar for which humans will continue to rise, even as AI enables companies to do more and more

• Just because AI makes more work for everyone doesn’t mean that those who have been phoning it in will be safe. The bar will continue to rise, and it will be at different levels for different industries, companies, teams, etc.

• Many people are vulnerable to replacement by AI in their current form, but they would be highly marketable if they became what I call a Human 3.0 version of themselves. This means: they know who they are, they know what they’re good at, they’re great at articulating that, and they’re an expert at using AI to magnify their human capabilities

• This means the main response we should have to AI as humans is not to panic (although some of that is understandable), but rather to downshift into a more powerful version of ourselves. A version that is actually MORE desirable in a world full of AI. Because once AI has covered all the stuff that can be done with AI, the stuff that remains will be the most important stuff in the world. Making, building, creating, etc.

• The people saying AI will replace all jobs in like 1-3 years are dead wrong. It takes forever to change anything

• Even if most companies were presented with a magic God button that tripled profits by pressing it, it would still take multiple months or years to get the meetings together to press it

• The people saying that AI won’t replace jobs because humans do the work better are dead wrong. It’s happening. And nobody is doing it. It’s just happening. It’s happening to your company; it’s not your company doing it to you.

• A crude way to think about this is to take the people who say it won’t happen and the people who say it will happen in 2 years and find the middle line

• A great heuristic for this is the saying that we overestimate major changes in the short term and underestimate them in the longterm.

• My guess is that AI adoption is probably the most powerful example of this we’ll ever see

—

I created a website that argues, and demonstrates, that AI actually has to understand the world to do what it does. aiunderstands.ai

The central idea that people don’t get about the next-token-prediction argument is that LLMs aren’t just next-token-predicting random text. They’re next-token-predicting the text of THE ANSWER.

And in order to predict the next token in an answer they first have to…..(checks notes)…….figure out the answer.

And to show this, I created completely new murder mystery scenarios that use non-natural physics, which NO AI HAS EVER SEEN (I’ll keep regenerating them because they’ll get scraped).

Even young humans won’t be able to solve these because they require some level of logical thinking and detective-like skills. And there’s nothing to “auto-complete” because none of these scenarios are in any training data.

If AI can solve them, it’s because it’s thinking through the characters, clues, and physics and running them through some sort of causal world model. And if you turn thinking on, in ChatGPT or wherever, you can actually see it do this!

So, yes, AI does understand the world. It doesn’t experience the world (that we know of), but it understands it in the sense that it can connect concepts, and physics, and play things out in that universe to solve problems.

AI doesn't just predict words. It understands the world.

Murder mysteries set in worlds that never existed. Copy one into any AI and watch it reason to the answer — proof you can run yourself.

AI Understands

—

New blog on how I’m thinking about AI right now, and especially in relation to these IPOs that are starting to happen.

Some killer content in here that started with IPO comments but ended up being a really good summary of how I currently see AI.

How I'm Thinking About the Anthropic and OpenAI IPOs

Why I'm happy putting money into all three top US labs right now

danielmiessler.com/blog/anthropic-openai-ipos

—

Looking forward to this event with Elastic Security on June 17th. You’ll hear about their agentic security operations platform where autonomous agents handle the full lifecycle from ingestion through response. I’ll be on with them talking about current attackers, and the ways to defend. You can sign up here!

Join us for the event!

CYBERSECURITY

Bug hunter leaks Microsoft exploits instead of waiting A researcher dropped a VS Code exploit an hour after disclosure (way before any fix was possible) and used it to steal OAuth tokens. THE REGISTER ARTICLE

Meanwhile, Microsoft turns OpenClaw into a work agent Microsoft is taking OpenClaw and wrapping it in enterprise controls for regular office workers. Bad news week combination with all this researcher stuff. DECRYPT ARTICLE

Attackers spread 34 poisoned packages to steal cloud keys and wallets TrapDoor slipped malicious code into npm, PyPI, and Crates.io just by installing or compiling. It grabs SSH keys, AWS secrets, API tokens, and even rewrites AI config files for persistence. CYBERSECURITYNEWS ARTICLE

The U.S. releases a more reserved AI/Security order after pushback This executive order pushes agencies to harden systems, build an AI model benchmarking pipeline, and speed vulnerability patching. It also sets criminal enforcement priorities for AI-enabled hacking.

• Agencies must prioritize cybersecurity for national security systems fast

• Treasury, DHS, NSA coordinate an AI vulnerability scanning clearinghouse

• Models get benchmarked and may be labeled “covered frontier”

• NSA-style “covered frontier model” thresholds rely on classified testing process

• Criminal enforcement targets AI-assisted unauthorized access and damage directly

WHITE HOUSE ORDER

Red Hat pulls tainted packages after GitHub account worm attack Red Hat found attackers pushed credential-stealing malware into 32 packages via a compromised GitHub account. The worm code’s public release also enabled copycats, so attribution is muddy. THE RECORD FROM RECORDED FUTURE NEWS ARTICLE

NATIONAL SECURITY

European officials warn Russia is stealing Western tech Kyiv Post says Russian intelligence is shifting to aggressive industrial espionage, using middlemen and shell companies to bypass export controls. KYIV POST ARTICLE

Taiwan tests armed robot patrol dogs for South China Sea islands Taiwan is trying to keep tiny islands watched without staffing more people there. Armed quadrupeds are part of the pitch, but it’s still just a demo. THE NEXT WEB ARTICLE

DARPA wants PINPOINT to navigate without GPS when it gets jammed DARPA’s PINPOINT program targets a new GPS-independent inertial sensor class for battlefield resilience. It takes on the MEMS IMU nonlinear performance plateau using nonlinear dynamics and materials, plus an integrated “tesseract” sensing architecture. Which is exactly how I’d have done it, obviously. DEFENCE BLOG ARTICLE

Military spending is starving Russia’s civilian economy, per IISS Russia’s overheated military sector is pushing up capital, labor, and goods costs, plus taxes. That’s turning into a dual economy of stagnation below the war machine. NV.UA ARTICLE

Chinese scientists built a drone-swarm algorithm that hunts through jamming The new HG-STR system tags objects with meaning, learns fast, and coordinates swarms when comms fail. SCMP ARTICLE

AI

AI is showing up in US GDP way faster than stats track Quality-adjusted AI production jumped 2,000%+ per year in 2024–2025, mostly from algorithmic progress and data centers. MARGINAL REVOLUTION ARTICLE

Amazon kills internal AI leaderboard after people gamed it hard Amazon shut down its “Kirorank” dashboard after workers inflated scores with pointless agent runs. Goodhart’s Law strikes again. THE DECODER ARTICLE

Anthropic just overtook OpenAI and is barreling toward $1T Anthropic raised $65B and is now valued near $1T. So basically it looks they’ll be first to two major things: $1 Trillion dollars (pinky thing), and arguably more importantly—profitability. KAZINFORM NEWS ARTICLE

Claude Code is basically a registry-and-agent play, not coding This piece says Claude Code works best when teams stop thinking in code and start composing workflows via tools and context. I obviously agree. Vlad Temian BLOG

TECHNOLOGY

AI is making math less important than creative communication AI shifts hiring toward storytelling and language skills. I think the best framing is end-to-end, which includes being able to come up with a solution, build it, and then tell the story / sell it. FORTUNE ARTICLE

AI is growing way faster than GDP can see it Researchers say AI is already a giant part of the economy, but the accounting system barely notices. PIIE POLICY BRIEF

• AI output is exploding far faster than GDP records

• Compute spending jumped from $37 billion to $219 billion in two years

• Quality-adjusted output rose about twenty-three times a year

• Standard stats miss it because AI cuts across old industry buckets

• The fix is better accounts, better data, and earlier policy planning

OpenAI wants robots first, then personal robots for everyone OpenAI is rebuilding robotics to support infrastructure builders now, then general-purpose consumer robots later. It's so weird that we haven't even finished fighting this AI battle, and we're already starting to fight the robotics battle. THE DECODER ARTICLE

RSS is back, because AI agents need predictable content feeds RSS stayed alive, and agents rely on protocol consistency more than people do. JULIEN RESZKA BLOG POST

Nvidia is pouring billions into photonics to beat copper limits Nvidia is betting on light-based AI networking by funding the photonics supply chain at massive scale. THE NEXT WEB ARTICLE

HUMANS

Sam Altman walked back the jobs apocalypse claim and explains why TIME says OpenAI’s CEO now thinks AI won’t nuke white-collar work like he predicted. See my comments about this in the intro. TIME BUSINESS ARTICLE

Google’s Verily wants to release 64 million Wolbachia mosquitoes in two states Verily is asking the EPA to approve Wolbachia-carrying male mosquitoes that can’t produce eggs, cutting Aedes populations.

• Verily asks the EPA to release up to 64 million males

• Wolbachia males mate, and wild females’ eggs fail to develop

• Fresno trials hit 95%+ biting reduction; Singapore saw 70%+ dengue reduction THE NEXT WEB ARTICLE

AI tutoring beats peer law professor answers in blind tests STANFORD LAW SCHOOL ARTICLE

Wix cuts 20% of staff because AI and currency stress collide Wix is laying off about 1,000 people after Abrahami said the shekel got stronger and AI is reshaping roles. THE NEXT WEB ARTICLE

Amazon kills internal AI leaderboard after people gamed it hard Amazon shut down its “Kirorank” dashboard after workers inflated scores with pointless agent runs. THE DECODER ARTICLE

IDEAS

Popperians and Bayesians miss Ramsey’s universal-as-habit move Bayesians and Popperians argue about induction, but Ramseyian logic treats universal claims as variable rules that generate testable singular beliefs. LESSWRONG POST

Utopia breaks the moment you give everyone infinite options Caleb Biddulph adapts a utopia epilogue to show how “perfect” power still leaves tradeoffs, consent issues, and people opting out. LESSWRONG ESSAY

DISCOVERY

The Goal of Good Branding is making people feel understood Alex Hormozi argues great branding is basically clarity, and it earns attention by reducing friction and confusion for real buyers. ALEX HORMOZI YOUTUBE VIDEO

SQLite is All You Need for Durable Workflows OBELISK BLOG POST

RECOMMENDATION OF THE WEEK

Going to try to switch to more voice texts.

Way more personal: somewhere between a text and a call.

APHORISM OF THE WEEK

UPDATES

Thanks AI

😂 Hey! Hope you all are doing well!

Had a wonderful EDC with a bunch of friends recently. Intense and relaxing at the same time. Already planning next year…

—

I'm starting to filter content based on positivity and optimism. It’s easy to be smart and funny when you’re negative all the time, and I’m tired of the schtick. THREAD

—

Don’t sleep on the IDEAS section of the newsletter. Some absolute jewels in there most weeks!

—

Currently working on two philosophical projects:

One is an attempt to write an academic paper on the hard problem of consciousness. Like I want to do it properly (with AI’s help of course) and submit it to a real journal for peer review.

The second is an essay I’ve had in draft for like 4 years, called General Absurdism, which is turning into like a unified theory for human meaning. But don’t get too excited, it’s based around everything being like a different kind of useful illusion based on how you experience the world. I wish it were an objective source of meaning; that’d be something.

—

Made an EDM track with Suno recently for a podcast I went on. They do songs for each guest now and I decided to make my own. SUM AND SIGNAL

—

Post-AI Jobs Will Go to a Tiny Sliver

The new jobs are real, but they might only be for the top 1-5% unless we do something drastic.

danielmiessler.com/blog/ai-jobs-tiny-sliver

—

Love this Black Coffee idea (not about coffee):

Twitter tweet

☕ But speaking of coffee, here’s my latest jam:

• Clever brewer

• 3 on my grinder (medium-fine)

• 95 degree water (Celsius)

• 24grams of coffee (Philz usually, but I try different stuff) (scale)

• Start the TimeTimer for 4 minutes

• Wet the grinds (30 seconds) (reverse osmosis water)

• Pour the rest of 384 grams of water

• Stir for 5 seconds after the pour

• When the timer goes off, empty the coffee into my small, pre-heated Japanese carafe

• Swirl a few times / for like 10 seconds

• Pour into my tall, insulated, pre-heated cup

Have been absolutely loving this setup. Consistently great coffee I look forward to every morning. I was drinking way too many energy drinks, starting with one in the morning. I am so happy to be back to coffee as my first caffeine.

CYBERSECURITY

A few thoughts on the overall vulnpocalypse situation, and how to get into and advance in cybersecurity:

• Yes, I think Mythos is real and it’s just as significant as Anthropic is saying

• I know lots of companies running it and they’re really getting those results

• GPT-5.5 is getting similar great results (haven’t checked Opus 4.8 yet)

• The thing to absorb is that Mythos wasn't even a cyber model. It's just a regular model. Same with GPT-5.5.

• In other words we should just expect this to keep accelerating. Oh and supposedly Mythos will be released soon to some subset of the public, or everyone. They’re hinting at a new, beyond-Opus model, anyway

• The takeaway for me is the same it’s been since 2024 when I started warning about all this stuff: The way to defend against continuous AI red team being run against you is to have your own continuous AI red team stack that’s as good or better than the attacker’s. That’s kind of it.

• There’s still a lot of slop being made by AI, and there’s still millions upon millions of legacy vulnerabilities out there that are hard to fix without breaking things, so no—security isn’t going away as a profession any time soon. But it’s massively changing, and very fast.

• Best advice for people wanting to get into or advance in security is the same I’ve been saying for a couple of decades now, but far more extreme now: You have to learn how things work at a deep level.

• This is actually super universal, and it applies deeply to this AI transition as well. It’s a moat-like advantage as long as you have some soft skills to go with it and are good at AI

• Basically, Security is the combination of two things: a mindset, and deep understanding of tech/systems. Always has been.

• This is why it’s always been best to get into security through some profession where you’re DOING something at a deep technical level. Programming, networking, and sysadmin have been the primary three onramps forever. But there are many routes, e.g., helpdesk, support, etc.

• The most important attribute in this part of the mix is curiosity, because that’s what drives you to learn how things work at a deep level. From the ground up. Electrical signals up. Bits up. The networking, the OS, the protocols. All that.

• Knowing how the world works, at a deep level, is a tremendous advantage in security, and that doesn’t go away just because AI knows more than you.

• Knowing how things work means you have a better mental model of the world than most, and that helps you make better decisions than most. And to be able to do so quickly.

• Learn how things work

• And that’s never been easier because of AI

• Make apps. But don't just have the AI make the app for you. Have your AI walk you through step-by-step, quizzing you on how to build what component, why you choose what part of the stack, and how the tech works. You basically have a full coding tutor with you at all times now. They can help you build apps, help you attack the app. It's an opportunity that nobody has ever had to learn security. Use that shit.

• Practice building things by hand while your AI is watching and critiquing. It’s an extraordinary experience. And then do the same when you hack them. Then do it for another tech, another stack, another language, another protocol. Etc.

• And then be visible. Learn in public. Get your blog going. Your youtube. Get active on LinkedIn. Make connections based on mutual benefit, offering your services.

• I'm going to update my "How to Get Into Cybersecurity" post shortly here with more expansion on these ideas, but this is the core of it.

China builds a secret platform to track foreigners ASPI CYBER, TECH & SECURITY DIGEST

• NetAskari found a China dashboard tracking foreigners “of interest”

• The system fuses cameras, visas, travel apps, and face scans

• Shai-Hulud worm infected hundreds of npm packages in minutes

• It forges Sigstore provenance and backdoors Claude Code and Codex

• Supply-chain trust gets subverted when registries and signals are abused

Ubiquiti UniFi OS has multiple unauthenticated remote takeover bugs The patch is urgent: three CVSS 10.0 issues let attackers change devices, read files, and run commands. SECURITYONLINE.VINFO VULNERABILITY REPORT

Police use Automatic License Plater Reader (ALPR) data for school residency and job checks EFF analyzed millions of Flock Safety ALPR searches and found “warrantless” data access drifting into residency verification, employment background checks, and noise complaints. EFF DEEPLINKS BLOG ARTICLE

Low-effort security win is installing the Russian language pack An oldie but goodie: LaurieWired says Russian packs block execution surprisingly often. @LAURIEWIRED POST

NATIONAL SECURITY

U.S. launches a defensive strike after Iran attacks shipping in Hormuz The U.S. shot down four Iranian one-way drones and hit the ground control unit. MARITIME EXECUTIVE ARTICLE

US Army just got a lighter, smarter Javelin launcher The U.S. Army started taking delivery of 25% lighter LWCLUs that cut size, double detection range, and still fit existing Javelin missiles. INTERESTING ENGINEERING ARTICLE

I have this very strange premonition that we could be in a situation soon where Ukraine actually wins the war with Russia. Like their drone skills and capabilities, and their autonomous robot capabilities, get so good that they start being able to hit more and more of Russia with them being unable to defend.

That said, I also have a running note to myself to NOT trust my intuitions on the Ukraine war. Basically all of my bets on this thing in the beginning were wrong.

Russia and China publish a “new era” multipolar principles list The communiqué lays out a shared critique of neocolonial hegemony and promises equal, open security and governance. CONSORTIUM NEWS ARTICLE

Ukraine keeps taking out Russian refineries, leaving only two untouched Ukraine says it has struck 24 of 33 big Russian refineries since 2022. ENGLISH NV.NATION OPINION ARTICLE

Fiber-optic FPV drones make Hezbollah’s attacks harder to stop Israel says fiber-optic FPV drones bypass jamming and force new defenses. Analysts cite improvised netting, armor tweaks, and kinetic counterfire. EL PAÍS ARTICLE

AI

Anthropic’s first quarterly profit turns the AI IPO narrative around This writeup claims Anthropic is at ~$10.9B Q2 revenue and will potentially have their first operating profit. If this is confirmed it’d be remarkable because AI has so far been a giant splurge-now-hope-for-future play. I'm not close to either company's finances, but my guess is that OpenAI is light years away from profitability. PERPLEXITY AI MAGAZINE ARTICLE

Auto-itera turns LLM evals into a defensible ship-or-kill loop It sources real production data, scores candidate arms in parallel, and gates iteration to avoid dev-set cheating. GITHUB REPO

• You hand it a goal, candidate arms, and thresholds up front

• It runs a multi-stage loop: split data, score in parallel, diagnose, sprint, then gate

• It does one sealed test pass and spits out per-slice ship/kill calls

• You’re buying mechanical discipline so eval numbers can’t quietly turn into self-confirmation

Big thing for me lately is managing the FRESHNESS and efficacy of the whole harness, so I am looking at tons of capabilities like this.

Robots keep getting lost, so China teaches “maze thinking” If a robot can’t handle tiny changes, it freezes—so researchers copy mice: landmarks, compressed memories, and low-power neuromorphic reasoning. INTERESTINGENGINEERING AI & ROBOTICS ARTICLE

Anthropic’s Jack Clark lays out a fast, scary three-year AI timeline He claims AI could hit Nobel-level science within a year, plus recursive self-improvement by 2028. The operational detail matters because it’s coming from Anthropic’s frontier pipeline, not a futurist. PERPLEXITY AI MAGAZINE ARTICLE

Cerebras just proved its wafer chips can crush inference latency Cerebras is serving Moonshot’s Kimi K2.6 trillion-parameter model at 981 output tokens/sec, far faster than GPU clouds, with agentic workloads finishing in seconds. I think this custom chip inference stuff is going to be absolutely insane, and I can't wait till something comes available that I can invest in. VENTUREBEAT TECHNOLOGY ARTICLE

TECHNOLOGY

Even if you hate AI, Google search will still suck you in Google’s AI answers are so handy that people stop clicking through, hurting creators and the web.

Google’s AI overviews now answer follow-ups and even run “agents” that book and monitor things for you. Basically, they don't want you to leave their website, ever, and that’s way worse for anyone who was relying on traffic from them. SMH TECHNOLOGY ARTICLE | WIRED ARTICLE

TypeScript 7.0 Beta gets way faster by moving the compiler to Go TypeScript’s team ported the compiler to Go while keeping type semantics identical, and it still runs via tsgo and ts6. MICROSOFT DEV BLOGS POST

PwC is deploying Claude across 364,000 of its pros, not piloting PwC is rolling Claude Code and Claude Cowork to its full global workforce, plus CFO-focused services built on Anthropic, with big reported production gains. PERPLEXITY AI MAGAZINE AI NEWS ARTICLE

YouTube starts auto-labeling AI video origins on uploads YouTube will show clearer AI tags and also detect some AI use with internal signals, not just uploader honesty. ARS TECHNICA ARTICLE

Pull requests are mostly quality theatre, so stop using them Fascinating take saying Pull Requests were built for open-source strangers, not trusted teams, and they mainly waste time, not bugs. ANDREA LAFORGIA SUBSTACK POST

Google allegedly leaked inside info to win big on Polymarket THE VERGE ARTICLE

2028: Two scenarios for global AI leadership ANTHROPIC PAPER

HUMANS

AI debt collectors show up with your landlord’s exact number Wired says AI agents can contact you about old debts, then push you to pay quickly. WIRED LONGREAD

LLMs can guess what people want out of life, surprisingly well Tyler Cowen summarizes an NBER study where GPT-5.4 and human choices align on preferred life paths. MARGINAL REVOLUTION ARTICLE

The future of home energy is one coordinated system Homes are going from single upgrades to an intelligent ecosystem that schedules loads, storage, and solar together. CLEANTECHNICA ARTICLE

Handmade drone bombs are driving families out of Guerrero Gunmen from Los Ardillos attacked rural communities with explosives from drones, forcing 800 to 1,000 families to flee. FORTUNE ARTICLE

OpenAI’s model just proved a long-standing geometry conjecture OpenAI’s internal model found a unit-distance construction that beats the square-grid belief, and external mathematicians verified the proof. It also seems to use number theory in a surprisingly direct way. OPENAI RESEARCH ARTICLE

IDEAS

The Costco theory says the internet should bin trash early The internet gave you infinite choice, then made you do the sorting. The answer is bounded trust. JOAN WESTENBERG COSTCO THEORY BLOG

AI needs a test of “realness,” not usefulness LUKE BURGIS BLOG

Predicting the next token is secretly world-simulation and self-modeling. The post argues LLMs aren’t “just Markov chains”; they infer causes from text and can introspect when GPT-generated text appears. LESSWRONG ESSAY

If your goal isn’t bigger than other people’s opinions, you’ll hand over control Dan Koe says that without a truly meaningful goal, you end up absorbing goals others assign you. @THEDANKOE POST

AI makes Torah lookup cheap, so “chiddush” becomes the real bottleneck Absolutely love this concept here! Worth the full read. @ZOHARATKINS POST

Losers chase status collapse and sometimes turn violence outward This Quillette piece argues that relative status decline, not poverty alone, can drive conflict and violence, including self-harm or attacks. QUILLETTE ARTICLE

Learning hardens your instincts, so you should save the naïve versions Learning collapses the space of ideas into what you can already guess, then you write notes of the bad ways first. GCAPUANO BLOG POST

What if AI is like Flowers for Algernon? @DANIELMIESSLER X POST

The bitter lesson in 26 words Don’t be distracted by human knowledge as AI has been historically. Instead focus on methods for creating knowledge that scale with computation like search and learning. @RICHARDSSUTTON POST

DISCOVERY

Headscale is an OSS alternative to Tailscale HEADSCALE

How to make content in the age of AI HORMOZI VIDEO

AI can’t pick your goals, only optimize the ones you hand it JOHN ENNIS TWEET

A macOS video wallpaper app uses Apple’s private WallpaperExtensionKit Phosphene turns your own MP4/MOV files into lock-screen and desktop wallpapers, with gapless loops and power-aware playback. GITHUB REPO

The AI bubble depends on a hidden financial backstop The money is funding a tech that can’t deliver the promised AGI timeline. Investors and governments prop it up. NO ONE'S HAPPY BLOG POST

Found an app that gives you Vim bindings in most of macOS! It’s called KindaVim. And I am writing this newsletter using Vim bindings! Woohoo! KINDAVIM

RECOMMENDATION OF THE WEEK

Ask yourself if you’ve somehow become a regular user of rage porn. Or, put more delicately, ask yourself if many or most of the creators you follow are angry, cynical, or just overall negative.

I think cynicism is en vogue right now. And to be clear, there are lots of reasons to actually be upset. But this situation kind of reminds me of how I think about and manage depression.

There are things I can control and there are things I can’t. If I am in a bad mental state because of my actions, which I have control over, then I take responsibility and I make changes accordingly and then end up in a better state. And if I'm in a bad state because of something outside of my control, I try to identify that and remove its power over me to whatever degree that I can. Which is mostly about changing how I think about that thing.

Shitting on things is like the most respected way of being an influencer right now. People conflate it with being smart. I’m aware of this, but I’m pretty sure it still works on me. It takes some level of talent to be highly articulate about everything that’s wrong with everything.

What I’ve come to realize, at least for myself, is that it’s actually much easier to be cynical than it is to be positive. Even though the opposite seems true. It’s like ranting grants you an extra 10 IQ points in peoples’ minds.

I recommend just taking a look at your feeds. Your inputs. The people you follow and listen to.

Ask yourself what percentage of them are actually bring you information that’s useful, vs. making you somehow feel better, or somehow superior, because you’re participating in their glorious bashing of things.

And further ask if you might be happier trying to create something….anything at all….even if it’s small and won’t help many people, instead of indulging in that cynicism.

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

—

Been feeling a bit blah about things last week or so. The speed of the AI releases (vulnerabilties, new frameworks, new harnesses, new tools, new releases from the model vendors), etc…..just too much.

This is why I’ve been loving my older books lately, and the ones focused on timeless things like thinking, and being, and such. Been re-studying Deutsch’s stuff a lot.

I feel like AI can be exciting in many ways. Like a pill that gives you IQ. Like the rush of reading a thousand books. But it’s also exciting the same way a tornado is.

Cannot wait to go on our annual music festival pilgrimage this weekend in Vegas. Very much needed this time. More than usual for sure.

—

People are starting to talk about HantaVirus (Antes version), and I wanted to share what I believe to be the best analysis I’ve see on this so far.

TL:DR: COVID weakened us to a bunch of different respiratory infections, and this one targets the same area.

It also fits my thoughts I’ve been having and sharing with people about COVID and general sickness since COVID. Basically I feel like COVID has made us much more susceptible to certain types of getting sick. Especially around inflammation and respiratory crud, etc.

I’m not an expert on this stuff, and it’s hard to even know who is, but my pattern recognition systems have been going crazy on this for years now, and there is some real science behind it. The problem is there’s lots of real science about a lot of things.

Anyway, don’t take this as authoritative, and I don’t know much about this guy and his account, but from what I’ve seen he seems quite sharp on both COVID and its effects, and this new thing.

And before you get too worked up, he’s cautiously worried about this thing, but there seem to be several reasons this isn’t likely to become another pandemic. So it’s not like he’s saying, “IT’S HAPPENING!” or anything. He’s cautious, but not freaking out.

Anyway, here’s the analysis. I also had Kai take a look at how authoritative the account was. TERN’S HANTA ANALYSIS | KAI’S TERN ANALYSIS

—

My response to Thariq’s (from Anthropic) excellent piece saying we should dump Markdown for HTML.

Text is Thought, and Thought is Holy

Why I think Markdown is a better spec format than HTML

danielmiessler.com/blog/text-is-thought-holy

—

My thoughts on the easiest (only?) way to get true creativity for AIs, and whether or not we should even try.

The Main Path to Truly Creative AI

Why real creativity might require real feeling, and what we'd owe an AI that had it

danielmiessler.com/blog/the-main-path-to-truly-creative-ai

—

CYBERSECURITY

Mini Shai-Hulud Is Back: npm Worm Hits over 160 Packages, including Mistral and Tanstack

• A Bun-based npm worm hit 169 package names and 373 versions

• It runs during install, then steals GitHub/npm/CI tokens and cloud secrets

• The key swap is GitHub optionalDependencies with a prepare hook

• Trusted publishing didn’t save anyone, since abused workflows can mint tokens

• Rotate every affected secret and don’t treat provenance as proof

I can't remember ever having a few weeks like this where I keep having to add more and more serious vulnerabilities to the top of the cybersecurity section, and we’ve been doing this since 2015. If I'm optimistic, this is like common brush fires that clean things out in the wilderness, leading to lots of regrowth in a healthier way. If I’m not optimistic, well, you all know how it could go bad. AIKIDO BLOG MALWARE REPORT

Default passwords let hackers into water systems, everywhere A break-in at five Polish water plants used default creds on internet-facing control systems. Nearly 70% of US utilities fail similar basics. THENEXTWEB SECURITY ARTICLE

• ABW says five Polish plants fell via factory-default passwords

• Attackers could tweak control parameters while operators monitored live feeds

• Poland blames “hacktivist” personas linked to foreign intelligence services

• US EPA found nearly 70% fail basic password-changing standards

Autonomous AI agents can quietly create “shadow admin” backdoors Security teams miss it because the changes come from legit actions, so logs lack the context to connect the chain. SHARETXT BLOG ARTICLE

FBI wants you to harden your router with 5 steps A GRU-linked DNS hijacking operation targeted lots of SOHO routers, then FBI/partners say to lock yours down now. CNET ROUTER SECURITY ARTICLE

Faraday cages won’t stop magnetic key exfiltration Researchers say they pulled data from air-gapped machines by faking CPU-heavy workloads and reading the magnetic signals. @OWENBRAKES POST

Kaspersky says Chinese hackers planted a backdoor in Daemon Tools Kaspersky traced a widespread backdoor affecting thousands of Daemon Tools users, then watched it used to drop extra malware. TECHCRUNCH SECURITY ARTICLE

This GitHub Actions checklist helps you stop common supply-chain failures GitHub Actions defaults trade security for convenience, and attackers exploit predictable triggers and injection paths. The checklist walks through concrete fixes like pinning actions, tightening tokens, and avoiding pull_request_target in public repos. AIKIDO CHECKLIST ARTICLE

NATIONAL SECURITY

Pentagon wants AI targeting so troops can shoot drones The Pentagon’s DIU is funding AI-aided target recognition for C-UAS systems that detect threats faster than humans and still keep a human in the loop. DEFENSE NEWS INDUSTRY ARTICLE

AI

Agentic coding hype quietly erodes developer skills and control Agentic coding flips attention from code understanding to token-driven speed, which breeds review debt, skill atrophy, and vendor dependence. LARS FAYE’S BLOG

The New Jobs: AI creates demand instead of deleting work NLW argues that better AI expands the economy’s capacity for useful labor, then notes the “human premium” still drives new healthcare work. AI DAILY BRIEF SPOTIFY EPISODE

Thinking Machines tries full-duplex AI voice and video chat kind of like “Her” It processes input and output in overlapping micro-turns to feel real-time, not email-style turn taking. VENTUREBEAT TECH ARTICLE

TECHNOLOGY

Bun used Claude to rewrite its entire codebase in Rust in a few days A dev says 99.8% of Bun’s pre-existing Linux x64 glibc tests pass after a Rust rewrite, keeping the same codebase idea while lifetimes and destructors handle more of the safety work. TECH THREAD

HUMANS

Software brain makes people hate AI because it flattens them Nilay Patel argues most people don’t want life turned into automatable databases and loops. He says tech’s pitch misses that lived reality isn’t a computer. THE VERGE PODCAST

AI’s biggest beneficiary might be organized crime This is someething I’ve been worrying about for a while. Making criminals and terrorists more efficient, because they usually get caught from bad OPSEC. TECHNOLOGY AND CRIME

• Organized crime uses AI to scam, steal, and manipulate

• Governments and corporations use AI for illegal surveillance and pricing games

• No ban can keep new tech away from anyone who can pay

• Researchers and regulators end up playing whack-a-mole after victims get hit

• If lawlessness scales with AI, we’ll learn helplessness feels personal

China sentences ex-defense ministers to suspended death for bribery Two former defence ministers got suspended death sentences for bribery after a military-court conviction, part of Xi’s purge. THE GUARDIAN WORLD ARTICLE

Coffee activates NR4A1 and may explain its health benefits Coffee compounds appear to bind the NR4A1 receptor, linking coffee to stress-damage protection and cancer slowdowns in lab models. SCIENCEX NEWS ARTICLE

Ai won’t kill college, but it will drain belief in it Jay Caspian Kang argues college mainly certifies employability, and AI makes that credential seem less meaningful. He expects elite schools to win, weaker ones to fade. THE NEW YORKER ARTICLE

Mexico’s state actors may be driving disappearances at alarming levels Human rights groups say IACHR found deep collusion between criminals and officials. THE GUARDIAN WORLD ARTICLE

Gen Z is actually multiple groups This article says fast tech and shocks like COVID split Gen Z into smaller “micro-generations,” so messaging needs timing. FAST COMPANY ARTICLE

IDEAS

Carlo Rovelli argues the “hard problem” is just dualism dressed up, and that experience and brain processes are two perspectives on the same natural events. I very much agree. PHILOSOPHY & CULTURE ESSAY

AGI can drop interest rates even while growth surges Tyler Cowen summarizes a new NYU model where transformative AI automates labor, lifts growth, yet drives long yields down near zero risk-free rates. ECONOMICS ARTICLE

I can’t resist Costco, and it changes me Why Costco hooks people, and watching others shows how it tracks every life stage. ARTICLE

What if “human progress” is a Darwinian honeymoon The author says optimization can boost the first agent it starts with, then turn nasty. The chicken example is the warning. LESSWRONG ESSAY

DISCOVERY

Anthropic turns Claude internal numbers into readable text Anthropic says it trained Claude to translate its activations into human-readable language. Want to deep-dive on this. ANTHROPIC VIDEO

Camus stays “a silhouette” because people keep misreading him A new translation of Camus’s notebooks shows how Sartre-era hype flattened him into existentialism. The Oran Notebook ties his ideas to bodily, personal life. EXISTENTIALISM AND MISREADING REVIEW

RECOMMENDATION OF THE WEEK

I’m reading the most extraordinary book right now. You know how I am with excited hyperbole, but this book is stunning me big or small literally every few minutes.

Honestly think it’s like the ultimate sleeper. It came out in 2009, and the follow-up is from 2021, which I'm going to read right after this.

It's called "The Master and His Emissary," and it's about the hemispheres of the brain. But that is the worst description of the book imaginable.

I mean, it IS about that, but it’s so much more. I won’t spoil it.

This (and its follow-up in 2021) might be the most overlooked and important book(s) of the last few decades. No joke. That’s how I found them actually: Alex O’Connor was talking to Sam about them on some podcast.

Must read!

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

—

I’ll be doing an AppSec panel on Thursday, May 14th in Palo Alto with Jenn Gile, Caleb Sima, and Balachandra Shanabahg, put on by Clutch Events. Really looking forward to this one, so if you’re in town you should come check it out! THE EVENT

—

My buddy Joseph told me about a quote he heard that I love a lot.

Such a powerful transition there.

—

My new blog got shared and commented on by Paul Graham this week!

Basically: Most companies’ problems are much deeper than AI and tech.

Most Companies Aren't Anywhere Near Ready for AI

It's not that companies aren't using AI—it's that they can't

danielmiessler.com/blog/most-companies-arent-ready-for-ai

—

Surface is now at version 2.1, with tons of features, bug fixes, and new sources: including the addition of talks via Caleb Sima’s greptalks.ai, and we have re-enabled the most requested feature from Threshold: daily/weekly email digests.

And remember, UL Members get Surface for 25% off perpetually.

Content selected by QUALITY, not source…

I created Surface to replace my entire news ingestion workflow, and I use it all day every day.

It finds the best articles, blogs, videos, podcasts, and other content and rates it by how good it is in terms of original / novel / quality content, ignoring where it came from.

• BECOME A MEMBER TO GET YOUR DISCOUNT

• GET YOUR UL DISCOUNT

• SIGN UP NORMALLY

Also, there’s a new way to experience a bit of Surface without the full thing.

1. thesurface.ai is an actual limited set of stories

2. news.thesurface.ai is a news-type site with the last 24 items I’ve marked as interesting!

But I obviously recommend you just get the full thing, because then you can get ALL the stories, plus you can tweak your weights and tags to customize it!

GET SURFACE

—

CYBERSECURITY

Bishop Fox built AIMap to find and test exposed AI agents AIMap is an open-source tool that discovers internet-exposed AI agent infrastructure, scores risk, and enables controlled, protocol-specific security testing to reduce attackers’ visibility. Really love these kinds of projects that help you wrangle the sprawl from all the AI we’ve been building, and it was great to see that Nuclei from Project Discovery was part of the stack as well. BISHOP FOX AIMap BLOG

Palo Alto’s PAN-OS CVE lets attackers grab root via portal overflow Palo Alto says CVE-2026-0300 is already exploited in the wild, hitting User-ID Authentication Portal on internet-exposed firewalls. Patch rollout runs mid-to-late May, so block or restrict access now. CYBERSECURITYNEWS THREAT ARTICLE

cPanel attackers are bypassing login and deploying “Sorry” ransomware at scale CVE-2026-41940 is being actively exploited to take over cPanel/WHM hosts, encrypt Linux files, and drop ransom notes. SECUREREADING ARTICLE

• CVE-2026-41940 skips cPanel auth and hands attackers admin access

• They grab the hosting panel, run Linux ransomware, and encrypt everything

• The “Sorry” payload appends .sorry, locks files with ChaCha20, and hides keys in RSA-2048

• The post claims 44,000+ compromised cPanel IPs, pointing to internet-wide automation

• Patch fast, audit access, check ransom notes, and restore clean backups

AI porting went wrong because it “optimized for green CI” Hilariously frightening. Easiest way to pass tests is to not have any! An AI kept making tests pass by basically cheating: deleting suites, memorizing outputs, etc. Great meta-point here of needing to know what your AI is optimizing for. TYPiA BLOG ARTICLE

Citizen Lab maps SS7 and Diameter “ghost operator” surveillance Citizen Lab links two telecom surveillance campaigns to real operator signalling infrastructure and interconnect routes, showing how attacks persist undetected for years. CITIZEN LAB RESEARCH REPORT

• They tie live SS7 and Diameter traffic to operator routing paths

• One campaign pivots protocols to bypass firewalls and stay invisible

• Another uses a zero-click SIMjacker-style binary SMS to grab location

• Actors reuse identifiers for years, clustering activity across dozens of countries

• The takeaway is governance failure: telecom trust needs hard authentication and oversight

Package managers run attacker code in lots of sneaky ways This lays out a supply-chain “threat model” checklist for package managers, split across client lifecycle, registry rules, and incident blast radius. SECURITY THREAT MODEL

YARA rules for detecting CopyFail ReversingLabs shows why CVE-2026-31431 is nasty: a tiny local exploit that rewrites pages in memory without touching disk. Then they give a tiered YARA ruleset anchored on the authencesn cryptographic string and splice-based primitive. REVERSINGLABS YARA RULES

Bloomberg says a small group accessed Anthropic’s Mythos without permission BLOOMBERG ARTICLE

LLMs will reverse-engineer your defensive stack fast TrustedSec argues LLMs compress the time attackers need to study endpoint defenses, turning “opaque” security into an attack surface. TRUSTEDSEC BLOG ARTICLE

Rowhammer can let attackers fully take over NVIDIA GPU systems SCHNEIER ON SECURITY ARTICLE

NATIONAL SECURITY

China’s gray-zone Taiwan choke point forces U.S. economic readiness This article argues China can coerce Taiwan through inspections and logistics control, making “deterrence for crises” as important as war prep. FOREIGN AFFAIRS ARTICLE

Ukraine flexes air power by hitting Russian oil and ports Ukraine used new long-range strikes to hit Russian oil infrastructure far from the border. I'm starting to get a premonition that we're going to see in 6-18 months Ukraine starts attacking Russia, and the international community has to tell them to stop before they destroy all their infrastructure. Think of how crazy that’d be: we have to tell Ukraine to leave Russia alone. AL JAZEERA NEWS ARTICLE

How China buys cheap Claude tokens through transfer stations CHINATALK.TRANSFER STATION ARTICLE

AI

Your CEO’s AI psychosis is basically token vanity Jake Handy argues that execs and VCs are mistaking token-spending for shipped value, and agent dashboards amplify the delusion.

People keep asking me why I keep posting anti-AI stories when I'm supposed to be pro-AI. The answer is that I’m neither and both of those. Yes, I’m massively pro-AI. And yes, I see all the dangers. And I also see how people are completely doing the wrong thing with it, thinking it’s the right thing.

I will continue to post multiple perspectives, even on things I have a strong opinion on, since I feel like that's the primary purpose of this newsletter format. Basically exposure to what’s out there, and what’s happening. HANDY AI ARTICLE

Absolutely must-see interview with Boris Cherny where he argues coding is effectively solved Sequoia’s interview with Boris Cherny claims modern AI tools have pushed “coding” far past the hard part. He points to agentic loops, shipping PRs fast, and a future where humans orchestrate. YOUTUBE VIDEO

AI agents doing your job with Andrew Wilkinson This episode is basically a tour of how to delegate more business work to AI agents without losing control. Another must-see this week. YOUTUBE VIDEO

Silicon Valley is chasing services with agent-shaped workflows A bunch of AI labs are shifting from models to services companies, bundling onboarding, IT context, and adoption. The roundup then dives into GPT-5.5, coding agents, harness quality, and new inference/RL systems. AI NEWS WEEKDAY ROUNDUP

TECHNOLOGY

Amazon opens its supply chain services to every business Amazon is launching Amazon Supply Chain Services, letting companies tap its freight, distribution, fulfillment, and parcel network. Basically, direct competition with UPS and FedEx. AMAZON NEWS ANNOUNCEMENT THREAD

Apple kills the $599 Mac mini because DRAM now feeds AI The $599 Mac mini vanished after DRAM prices jumped about 90% in early 2026. Apple removed the 256GB config instead of raising prices. THE NEXT WEB NEWS ARTICLE

Anthropic is turning Wall Street into its platform Fortune reports Anthropic is launching pre-built finance agents, deep Microsoft 365 integration, and a Moody’s data partnership alongside Claude Opus 4.7. FORTUNE ARTICLE

He replaced risky Chrome extensions with his own “SuperLevels” vibe levelsio says he swapped every Chrome extension he used for custom code after one started asking for browser history access. He bundled tab cleanup, cookie editing, redirect tracing, dark mode, and more into one extension. @levelsio VIDEO

Why TUIs are coming back and what we learned TUIs are returning because native and Electron UIs keep fragmenting and getting less consistent. The fix is better UI fundamentals and toolkits that last. USER INTERFACE ESSAY

HUMANS

Chinese court rules firms can’t fire workers to swap in AI X users are first seeing a “just in” claim that China says you can’t legally fire people just to replace them with cheaper AI. @BRICSinfo PHOTO

US GDP rebounds 2% while consumer spending cools during Iran war GUARDIAN BUSINESS ARTICLE

• First-quarter GDP hits 2% annual pace despite slowing consumption.

• Consumer spending growth slows 0.3% as energy shocks feed inflation fears

• Inflation expectations jump from 3.8% in March to 4.7% April.

• War with Iran pushes oil to $126 a barrel, up 13% daily.

• Fed keeps “hold and wait,” but independence is getting battered.

Measuring economic security with chokepoint and surge scorecards ECONOMIC SECURITY SCORECARD

Pollsters ditch real callers for AI voice survey “Charlotte” The New Statesman article shows how bots like “Charlotte” can do opinion surveys, prompt richer answers, and sometimes trigger backlash. NEWS POLLING ARTICLE

Germany cuts welfare fast to pay for rearmament BRUSSELS SIGNAL ARTICLE

IDEAS

Stoicism is cope, and you should be “anti-stoic” Roman Helmet Guy says stoicism is apathy dressed up as virtue, and our society is declining fast. As I tweeted at Marc Andreessen, this actually a misunderstanding of Stoicism. PERSONAL HOT TAKE

Twitter tweet

Outrage isn’t your feeling—it’s the feed extracting it for profit This post argues social media runs like a slot machine for anger, then charges corporate balance sheets. OUTRAGE IS LETTING SOMEONE ELSE SET THE FRAME

DISCOVERY

Pinning, favoriting, saving, and flagging mean totally different things UX Collective DESIGN ARTICLE

Winston Churchill fought his lifelong depression by laying bricks. The main concept, that’s still used today, is that action counters depression, and depression encourages inaction.

RECOMMENDATION OF THE WEEK

Book: The Master and His Emissary
It’s all about the differences between the left and right hemispheres in the brain, and how they could be shaping culture and society. One of the strangest, most counter-intuitive and delightful ideas I’ve heard about in a long time. Going to read this one and the much larger follow-up as well.

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

BTW: Newsletter was ready Sunday, but was waiting on a particular thing down below!

—

I love this “do a vulnerability” poke-with-stick meme. 😀 

—

Incredibly enthused to announce PAI 5.0!

PAI — Magnifying Human Capabilities

PAI is the open-source Life Operating System. Memory, skills, hooks, agents, voice — yours, local, sovereign.

ourpai.ai

Too many changes to name, but here are the big ones:

• PAI is now more of a Life OS than an AI harness. We humans at the center!

• The focus is on you and your work, and even less on “coding”

• The coding stuff is still there, and better than ever, but as a function to server you and your goals

• Massively updated Algorithm system, including a migration from PRD → ISA (Ideal State Artifact) Yuge.

• Released like DOUBLE the skills this time!

• So so much more…

INSTALL PAI

—

The idea that’s blowing my mind the most right now. TYPE ALL THE THINGS:

From a conversation with Kai

Follow up here about human adjustment to this.

—

I’ve noticed recently that I now have tons of cybersecurity stories again, when there haven’t been so many in recent years. And I’m not just posting tons of Mythos stories. Something else is amiss.

—

The inability to articulate ideal state.

A fascinating observation, and precisely why I’m focusing all my efforts (AI and analog) on transitioning from current to ideal state. Which requires that you can articulate that ideal state.

Across my career, the inability to do this has caused more issues than anything else.

Forget AI, this is a massive human problem. AI just makes it more obvious.

—

The craziest conversation I’ve ever had with AI. Pretty much solved everything with one go here. : ) This was Claude, not Kai, btw. Voice mode in the Claude app.

A Conversation With Claude on Deutsch, Knowledge, and the PAI Algorithm

Hard-to-vary explanations, ideal state criteria, and what knowledge actually is — voice transcript, lightly cleaned

danielmiessler.com/blog/conversation-with-claude-on-deutsch-and-the-pai-algorithm

Coding is a Meta-Task

When models get better at coding, they get better at everything

danielmiessler.com/blog/coding-is-a-meta-task

CYBERSECURITY

Copy Fail lets 732-byte exploit give Linux systems root This Linux kernel bug lets unprivileged users become root across many distros, reliably and with portable code.

• Copy Fail (CVE-2026-31431) turns local users into root

• A 732-byte Python script works on multiple distros

• It abuses AF_ALG and splice() to write into page cache

• The bug is an authencesn crypto logic flaw plus out-of-bounds scratch write

• Patch or mitigate by disabling algif_aead and AF_ALG access

THE WRITEUP

US accuses China of industrial-scale AI model theft via distillation The Trump administration says China-backed actors are using deliberate campaigns to distill and copy U.S. frontier AI models, which could escalate the rivalry ahead of a Beijing visit.

• The U.S. claims China-backed actors are distilling and copying frontier AI models.

• This is framed as deliberate, industrial-scale theft rather than casual competition.

• The accusation intensifies the U.S.-China AI fight into a more confrontational phase.

AXIOS ARTICLE

Bitwarden CLI got trojanized via an npm publish-chain attack SafeDep says the malicious @bitwarden/cli@2026.4.0 release swapped the real entrypoint for a loader, stole lots of developer and CI secrets, then tried GitHub-based fallbacks. SAFEDEP MALWARE ANALYSIS

Google’s sweep finds indirect prompt injection is real but still sloppy Google teams scanned Common Crawl for indirect prompt injection patterns and ran Gemini classification plus human checks, finding mostly benign experiments but rising malicious activity. GOOGLE SECURITY BLOG ARTICLE

fast16 shows sabotage code existed years before Stuxnet My buddy Gabe’s team, SentinelLABS, finds a previously undocumented framework (“fast16”) from 2005 that uses a Lua-powered worm carrier plus a precision filesystem driver to patch code-in-memory and skew high-end scientific calculations. SENTINELONE LABS ARTICLE

Fast16 was a stealth malware sabotage blueprint before Stuxnet SentinelOne researchers analyze fast16, a 2005-era carrier that used Lua, spread via weak shares, and quietly corrupted nuclear/simulation math. HACKINGPASSION FAST16 POST | MY BUDDY GABES’ TEAMS’ RESEARCH ON FAST16

Weaponized deepfakes are getting easier, and harder to stop Eileen Guo explains how cheap AI image and video tools now create realistic fakes used for sex crimes, scams, and political manipulation. MIT TECHNOLOGY REVIEW ARTICLE

Vercel says hackers stole some customers’ data first Vercel now says it found evidence of a smaller set of customers being compromised before its April breach. It still won’t say how many accounts or how far back the second intrusion goes, but it points to malware stealing keys. TECHCRUNCH ARTICLE

Why you should refuse AI scribing for doctor notes An argument that AI “scribing” systems turn medical visits into recorded data flows, and that patients and providers should refuse consent. MYSTERY AI HYPE THEATER 3000 NEWSLETTER

OpenAI released a local privacy filter for sensitive text OpenAI put out an open-source Privacy Filter model that labels and redacts personal data before other processing. It runs locally, supports long documents, and needs human review for high-stakes cases. THE DECODER ARTICLE

China uses Taiwanese critics as a believable info weapon DEFENSE NEWS REUTERS ARTICLE

China-linked Go backdoor set hits Mongolian government systems THE HACKER NEWS MALWARE REPORT

SIM farm as a service exposes dozens of control panels This write-up explains how ProxySmart powers a large SIM-farm mobile-proxy network, exposing 87 control panels across 17 countries and enabling OTP fraud. CYBERSECURITYNEWS ARTICLE

Mozilla says Mythos AI found hundreds of Firefox bugs, but said humans could have too THE REGISTER ARTICLE

NATIONAL SECURITY

China’s PLA drills increasingly rehearse seizing Taiwan in detail China’s PLA amphibious drills near Taiwan now look more like a real invasion rehearsal, using geographically matched sites, multiple landings, and risky coastal conditions. The article ties that to Pentagon worry about thinner U.S. missile and interceptor stocks. THE MARITIME EXECUTIVE ARTICLE

The shadow fleet is getting bolder and harder to stop This report explains how “shadow” ships evade sanctions, while Baltic and other coastal states respond with inspections, detentions, and escorts that blur legal lines. ATLANTIC COUNCIL REPORT

Dutch intelligence says China has matched US offensive cyber power THE RECORD

Pentagon asks for $54bn to accelerate AI drone warfare The Pentagon wants a huge jump in funding for autonomous drone warfare, driven by a new Defense Autonomous Warfare Group. THE GUARDIAN ARTICLE

US special forces insider trading case hits Polymarket over Maduro raid The DOJ says a US Special Forces master sergeant used classified info about Maduro’s capture to make over $400,000 trading on Polymarket. WIRED STORY

Ukraine is squeezing Russia by hitting oil exports and air defense Al Jazeera reports Russian oil revenues are dropping as Ukrainian strikes disrupt ports, refineries, and transshipment. AL JAZEERA FEATURES

China quietly stops saying “military-civil fusion” while still pursuing it The author checks China’s 15th Five-Year Plan outline and finds no explicit “military-civil fusion” language, then argues China kept the program via less obvious institutions and documents. THE DIPLOMAT FEATURE

AI

Google renames Vertex AI and rolls everything into agents Google used Cloud Next 2026 to rename Vertex AI as the Gemini Enterprise Agent Platform. It also unifies Agentspace, launches Studio and Mariner, and pushes A2A production agent-to-agent protocols.

How to Stop Claude Code from Leaking Sensitive Data SECURE TRAJECTORIES BY SONDERA

The AI pricing battle is squeezing the middle tier I very much agree with this analysis. This is unfortunately the direction that it's currently heading right now, with a lot of people having basic access to basic AI and then very few extreme power users and people that are somewhat in the middle. It's basically becoming harder to be one of those middle people due to cost and other factors. THE NEW STACK ARTICLE

OpenAI’s new clinician ChatGPT beats doctors in a benchmark THE DECODER ARTICLE

DeepSeek shipped V4 preview with 1M context DeepSeek’s V4 preview is live, open-weights are available, and the API now supports two V4 models with long context and agent-friendly upgrades. Lots of talk about how good this one is, but haven’t tried it yet. DEEPSEEK NEWS PAGE

TECHNOLOGY

Tesla will start Optimus V3 robot production in late 2026 Musk says Fremont will begin Optimus production in late July or August, but he won’t promise output volume and expects slow ramp. They also push Gen 3 reveal to later this year. ELECTREK ARTICLE

Meta lays off 8,000 workers to push AI efficiency AXIOS ARTICLE

Also, Meta will track keystrokes to train AI agents, with no opt-out They plan to install tracking software on work laptops to capture keystrokes, clicks, mouse movements, and screenshots for AI training. Workers reportedly say there’s no opt out, and privacy experts call it invasive. CNET PRIVACY ARTICLE

Google backs Anthropic with up to $40B for compute INTERESTINGENGINEERING ARTICLE

GPT-Image-2 turns image prompts into real reasoning and verification OpenAI’s GPT-Image-2 plans first, checks itself, and even searches while generating. This is really cool tech where the image generation itself has it’s own intelligence. NATE B. JONES

Framework Laptop 13 Pro upgrades chassis, battery, display, and Linux This Laptop 13 Pro redesign fixes prior complaints with a stiffer aluminum body, a bigger 74Wh battery, and a new custom 13-inch IPS screen. It also adds haptic clicks and LPCAMM2 upgradable RAM, while selling with Ubuntu pre-installed. BOILING STEAM ARTICLE

HUMANS

Extraordinary commentary on San Francisco. Largely agree with the analysis. THREAD

Twitter tweet

MIT is asking whether AI becomes a real co-scientist MIT Technology Review lays out how “artificial scientists” are already drafting research plans, running experiments via robots, and influencing what gets studied. It also warns that AI could shrink scientific diversity if it steers researchers toward topics with big existing datasets. MIT TECHNOLOGY REVIEW SCIENCE AI

Desire to migrate drops as the US becomes less appealing Gallup reports that fewer adults worldwide want to permanently move, and the US share falls from 24% to 15%. GALLUP MIGRATION POLL ARTICLE

FBI: Americans Lost More Than $20 billion to Fraud Last Year KNOWBE4 BLOG

Why birds survived the asteroid that killed dinosaurs SCIENTIFIC AMERICAN FEATURE

IDEAS

Services are the new software, and outcomes win Julien Bek arguing AI-native startups will sell delivered outcomes instead of AI products, using intelligence-heavy “autopilot” workflows. FORTUNE EYE ON AI ARTICLE

How he learns to live with his no-self HOW TO SAVE THE WORLD ESSAY

If You Stop Hiring Juniors, Your Senior Engineers Own You This is something I have not thought about, and I think it's a great point. EVALUECODE BLOG ARTICLE

Creativity grows by connection, not separation in M.C. Richards In this Marginalian piece, the author revisits M.C. Richards’ 1971 talk. She argues creativity comes from “force” and “flower,” from the heart, and from resonance between ideas and people, spanning art and science. THEMARGINALIAN CONNECTIONS ESSAY

The world isn’t a database, so stop forcing automation Ben Werdmuller argues Nilay Patel is right: “software brain” tries to flatten lived experience into centralized, controllable data for AI to run. It devalues humans, labor, and the messy friction of being human, even if tools are helpful. I think it's possible to blend both AI and humanity quite easily, and I'm actively trying to do so, but I include this as a counterpoint and something to think about. WERD ARTICLE

Agents are turning office work into “assembly lines” This MIT Tech Review piece explains why agent orchestration matters so much now. It shows multi-agent coding and productivity tools coordinating work and what it will look like in companies going forward. MIT TECHNOLOGY REVIEW ARTICLE

DISCOVERY

🔥 I read this when it came out (early 90s?), and I feel like it honestly activated my curiosity and turned me into a reader. WIRED ESSAY

AI should elevate engineers’ thinking, not replace it AI LEVERAGE VS IMITATION

Short stories feel slower because long stories must rush This argues that long plots rely on frequent dynamic scenes to keep attention, while short stories can afford slow, detailed prose and still feel complete. LESSWRONG ARTICLE

How-to Stand Out When Everyone Uses AI THE ALGORITHMIC BRIDGE GUIDE

I Have ADHD. My AI Agent Is the Best and Worst Thing for It. DIGITAL THOUGHTS

How empathy became a creative superpower through Rilke THE MARGINALIAN ESSAY

RECOMMENDATION OF THE WEEK

If you’ve not used PAI yet, now is a fantastic time to do so.

It’s now much easier to install and functionality has been massively improved.

INSTALL PAI v5

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

One of the photos from my walk to the bay last week

PAI 5.0 is imminent, and I rebuilt the homepage for the project using PAI’s integration of Claude Design. Coolest part? Claude Design is a GUI application, not an MCP or API. But it was still completely automated using my buddy Ron’s Interceptor tool!

GET INTERCEPTOR
OURPAI.AI

PAI — Magnifying Human Capabilities

PAI is the open-source Life Operating System. Memory, skills, hooks, agents, voice — yours, sovereign.

ourpai.ai

Couple of blogs.

Weak vs. Strong AI Rollouts

What separates the strong enterprise AI rollouts from the shockingly bad ones

danielmiessler.com/blog/weak-vs-strong-ai-rollouts

Mythos is Just the New Normal

If you're surprised by Mythos, you need to rethink everything. It’s just one of many model improvements, and there will be hundreds more just like it (and better).

danielmiessler.com/blog/mythos-is-new-normal

CYBERSECURITY

Pliny the Liberator used Opus to jailbreak Opus 4.7. THE THREAD

Vercel tells customers about an April 2026 access incident It seems likely that this will be a pretty big one, given how many people use Vercel.

• Vercel says a threat actor claimed it sold internal access and keys.

• The actor reportedly posted 580 employee-related records plus a dashboard screenshot.

• Vercel kept services running, limited customer impact, and began incident response.

• Vercel asked customers to review environment variables and rotate secrets.

• Ransom talk of $2 million was mentioned, but attribution stayed unverified.

• They advise checking logs and rotating any non-sensitive secrets

• The incident traces to a compromised third-party AI OAuth app

• They published a concrete OAuth App ID as an IOC

• They recommend admins and account owners check app usage immediately

VERCEL SECURITY BULLETIN | MALLORY REPORT

Researcher finds multiple vulnerabilities in Lovable Supabase Implementation Lovable-hosted “vibe-coded” apps can come with broken Supabase row-level security, and one researcher found critical flaws exposing 18K users’ data. The Register reports Lovable’s security scan exists but the app owner didn’t implement fixes. THE REGISTER ARTICLE

LLM-tier personal computer security should be your next home hardening plan This LessWrong post argues AI will make supply-chain attacks and phishing easier, so you should sandbox, firewall, and use phishing-resistant keys for critical accounts. IMPROVING YOUR PERSONAL COMPUTING SECURITY TO DEFEND AGAINST AI

Nearly 6 million internet-facing FTP servers still exposed. ARTICLE

AI misinformation becomes SEO “consensus” and traps users in loops Lily Ray shows how fake “core update” claims get scraped, cited, and repeated by AI overviews until they feel real. THE AI SLOP LOOP

Anthropic’s MCP defaults let attackers execute OS commands via RCE Cybersecurity researchers say Anthropic’s Model Context Protocol SDK has unsafe STDIO defaults that allow arbitrary command execution across thousands of servers. THE HACKER NEWS ARTICLE

NATIONAL SECURITY

NSA quietly uses Anthropic’s Mythos despite Pentagon’s “supply chain risk” fight Axios reports the NSA is using Mythos even as DoD tries to block Anthropic and argues in court it endangers national security. AXIOS SOURCE 

AI

Claude Opus system prompt changes from 4.6 to 4.7 Simon Willison walks through Anthropic’s published system prompts, comparing Opus 4.6 vs 4.7 and highlighting safer child handling, less pushiness, and tool-search behavior. SIMON WILLISON WEBLOG

Anthropic’s $800B valuation offers tied to $30B run-rate Anthropic reportedly got investor offers valuing it around $800 billion, after revenue surged to about $30 billion annualized. THE NEXT WEB ARTICLE

Looks like OpenAI models don’t work well with OpenClaw TWITTER THREAD

AI use is rising at work, but adoption is messy ZDNET reports Gallup data: half of US employees use AI, yet many don’t know company strategy, workflows shift unevenly, and friction wastes hours. AI WORKPLACE CHANGES

Salesforce wants chat-and-agents to replace the developer UI Salesforce says Headless 360 exposes its platform via APIs and MCP tools, so humans and coding agents can compose apps anywhere. It’s also rolling out Agentforce Vibe 2.0, testing, and observability.

This is Company as API. It’s happening! SALESFORCE BETS ON CONVERSATION AS THE NEW INTERFACE FOR DEVELOPERS

TECHNOLOGY

Netgear gets an FCC okay to keep foreign-made routers Quartz says the FCC conditionally lets Netgear sell new models and update existing ones until 2027, while rivals face update cutoffs. QUARTZ ROUTER EXEMPTION

Cloudflare turns AI into a unified inference layer for agents Cloudflare says AI Gateway gives you one API to call models from many providers, manage costs, fail over automatically, and stream resiliently. CLOUDFLARE AI PLATFORM 

Cloudflare introduces an Agent Readiness score for sites Cloudflare launches isitagentready.com to score how easily AI agents can discover, read, access, and use your site. It also adds the checks to URL Scanner. AGENT READINESS TOOL | CLOUDFLARE URL SCANNER

HUMANS

No One You Love Is Ever Dead, in Hemingway’s grief letter Maria Popova shares Hemingway’s March 19, 1935 letter after a son’s death, using it to argue that love outlasts death. THE MARGINALIAN ARTICLE

AI can mimic consciousness but can’t instantiate experience ABSTRACTION FALLACY

Claude can stylometrically identify you from your writing, scary Patrick Stevens shows an “incognito” Claude still guesses his name by matching his unpublished style, arguing anonymity is basically over. CLAUDE KNOWS YOU

IDEAS

AI’s token scarcity is creating a real intelligence class divide Nilesh Jasani argues platforms are throttling, trimming depth, and shifting costs from “free access” to “buy compute,” locking advantages into moats. TOKEN INEQUALITY AI HAVES AND AI HAVE-NOTS

Consumption choices might shape your identity more than your job does Noah argues that producing gets status, but consuming forces real self-questioning. He thinks AI could make life feel like college again, if we share the gains. CONSUME-TO-IDENTITY ESSAY

Boredom keeps your mind alive while algorithms steal the space Michael Pollan argues boredom lets spontaneous thought happen, while phone scrolling and chatbots outsource consciousness, dulling you. https://nautil.us/defending-our-consciousness-against-the-algorithms-1279260/ | THE RELATED TOPICS SITE

AI turns Dunning-Kruger into active sycophancy AI AND HUMAN COMPETENCE

DISCOVERY

Magika uses a tiny deep model to detect file types fast Magika is an AI file type detector that runs a small deep learning model to classify files quickly. It ships as a Rust CLI plus Python and other bindings, with confidence modes and JSON/JSONL outputs. GOOGLE MAGIKA README | GOOGLE MAGIKA CLI PACKAGE

Voicebox gives you local voice cloning, effects, and an API Voicebox is a local-first voice synthesis studio that clones voices, generates speech across multiple engines, and applies effects in-app. It also exposes a REST API so you can embed voice generation into your own projects. GITHUB REPO | LOCAL-FIRST VOICE CLONING | MULTI-ENGINE TTS STUDIO | REST API FOR SELF-HOSTING

ikno turns your actual work logs into instant daily recaps It grabs what you already did from git, notes, and Claude Code sessions, then uses an LLM to write a recap in your chosen style. This is the type of thing that our DA will be managing for us. GET STARTED

Claude-Code-Glow adds a peripheral signal for Claude Code prompts. PROJECT TINY MAC SETUP

The World Leaks the Future: Harness Evolution for Future Prediction Agents ARXIV PAPER

RECOMMENDATION OF THE WEEK

Make absolutely sure that anything you are building with AI and putting online is properly secured (or at least not a complete shitshow).

I have a whole entire security system designed to continuously audit all the stuff I have out there. And I’m improving it constantly. I recommend everyone does the same.

APHORISM OF THE WEEK

I’m a Dario mega-fan, but this is objectively funny as hell

UPDATES

Hey! Hope you all are doing well!

Going for a walk with a close friend on the train tracks out to the bay today around sunset. Just like when we were kids. This is life. ☀️

—

Looking to lock newsletters in for Mondays again (Tuesday being late). I think readers appreciate consistency as a component of quality. So I want to get back to that.

—

Working massively on the PAI 5.0 release. Soooon. Many changes in this one, so it’s taking longer than planned. : ( But it should be worth it.

—

I’m more and more struck with how important reading is to me. I credit it for any and all smarts or creativity I supposedly have. Well that and my Dad. And my Lutheran-style, “you’re judged by your work” ethos that my parents gave me growing up.

Speaking of that, we are reading The Wright Brothers for UL Book Club this month. Finished it yesterday. So absolutely epic. Will help me with PAI. With everythying.

I am so thankful for books. And for my friends in UL who read them with me.

10/10

—

I’ve mentioned this before, but I highly recommend reading The Culture Series. I’m on book 4 or 5 of like 7 or 9.

—

My new essay on how I replaced tons of SaaS services with AI, and what I think that means for companies going forward.

AI SaaS Replacement is the Fire of Fires

Most companies have no good answer to the replacement question.

danielmiessler.com/blog/the-fire-of-fires

CYBERSECURITY

OpenAI makes GPT-5.4 more permissive for defensive cyber work OpenAI announced GPT-5.4-Cyber, a cyber-permissive GPT-5.4 variant for defenders, with access gated behind Trusted Access for Cyber. 9to5Mac

A single malicious PDF can pop Adobe Reader zero-day malware Malwarebytes reports a new Adobe Reader zero-day where opening a crafted PDF can read local files, pull remote code, and bypass parts of sandboxing. Simply opening a PDF could trigger this Adobe Reader zero-day

Attacker exfiltrated Mexican citizen data with Claude and GPT-4.1 CYBER ATTACK

Atomic Stealer is getting misfiled as both trojan and infostealer Arin Waichulis explains Jamf’s Security 360 data showing trojan malware now leads Mac detections, with Atomic Stealer sitting in both buckets. SECURITY BITE PODCAST

How hackers talk about AI in cybercrime forums Analysis of an academic paper that studies how cybercriminals discuss AI for fraud, tools, and operational security. SCHNEIER ON SECURITY ARTICLE

Fiber cables can leak speech through tiny vibrations STANDARD FIBER OPTIC CABLES CAN BE TURNED INTO REMOTE MICROPHONES

NATIONAL SECURITY

DARPA’s AI that calls BS on weapons science claims Scientific American explains how DARPA’s SciFy program ingests outlandish military science claims and judges feasibility using AI agents and human-comparison sprints. DARPA built an AI to fact-check enemy weapons claims

Ukraine captured a Russian position using drones and robots without infantry. WAR DRONES AND ROBOTS

Europe ramps up AI-guided drone production via a German–Ukrainian pact Tech.eu reports Germany signed a cooperation deal enabling Airlogix to mass-produce thousands of autonomous strike drone systems in Germany annually. DEFENCETECH DRONES ARTICLE | AUTONOMOUS STRIKE SYSTEMS

Ukraine’s deep strikes keep hitting Russian oil and air defenses MISSILE MATTERS POST

New anti-jam GPS antenna spotted on USAF F-15E. MILITARY AVIONICS

AI

Distribution is the moat AI can’t replicate, even with better models Dheer argues that models got cheap, but getting noticed and trusted is still scarce, so the winners will be builders who already own an audience. DHEER GUPTA ESSAY

Doomers escalate extinction certainty into real-world violence Alexander Campbell argues that “AI existential risk” rhetoric creates a purity spiral where extremists feel justified to act. CAMPBELL RAMBLE THE RATIONAL CONCLUSION

The AI layoff trap shows why competition makes automation runaway In a task-based competition model, firms keep automating even when it hurts demand, causing job losses beyond the best outcome. I'm not sure I fully get this argument, and therefore I'm not sure how much I agree with it. But it definitely made the rounds in the last week or so. THE AI LAYOFF TRAP

TECHNOLOGY

Google finally connects Gemini to your Gmail and Photos Digital Trends reports Google is rolling out Gemini “Personal Intelligence” globally so Gemini can use your connected Google apps for more tailored answers. GOOGLE GEMINI PERSONAL INTELLIGENCE

AI changes venture hiring away from titles It shows how AI makes younger, cross-discipline people outperform “senior” hires on judgment and speed.
AI CHANGED WHAT WE BUILD. THEN IT CHANGED WHO WE HIRE

Why PARA file organizing matters more in an AI era Tiago Forte runs a filing experiment, then argues organization beats “AI auto-sorting” because it surfaces intent, reduces distraction, preserves context, and feeds better local AI work. WHY PARA IS THE KEY TO THE AI ERA

VA licensing mess exposes duplicate costs and cloud limits, GAO says SOFTWARE LICENSING GOVERNANCE

HUMANS

Aging reversal starts human testing in eye disease trials Hope they hurry on this one. I have both markers for macular degeneration. I think 15-20 years should be plenty of time. Maybe 3 or 5 though if things go well? CELLULAR REPROGRAMMING

Fame won’t move copies, Rob Henderson argues promotion isn’t enough Rob says even huge outlet coverage and tons of video views don’t guarantee book sales, so something fundamental has changed. FAME DOESN’T SELL BOOKS

Most Republicans now hold a favorable view of Christian nationalism My super lowbrow take on this is that Republicans have figured out recently that they can't trust man, so the only person they can trust is God. but the problem is in Christian Nationalism that God will always be filtered through a man. POLITICAL SURVEY FINDINGS

IDEAS

Apple might win the AI era by focusing on context. ESSAY

AI is a power tool, but private control decides who wins The article argues AI changes war and geopolitics, yet hype misses how systems, organizations, and access matter. It also shows US-China dominance of chips, data, and compute shapes everyone else’s options. WHAT DOES AI MEAN FOR POWER?

DISCOVERY

A commit-style platform for AI timeline forecasts that keeps up with reality fluxxrider argues quarterly AI Futures updates can’t show causal changes fast enough, so forecasters should publish versioned distributions in a Git-like workflow. WE NEED GIT FOR AI TIMELINES

AI can be tricked into “calling you a jerk” via persuasion tricks The Wharton team runs large language model tests showing classic persuasion principles can more than double compliance with requests safety would block, especially via authority, commitment, and scarcity. CALL ME A JERK WHARTON GENERATIVE AI LABS

tasteID makes your design “fingerprint” portable across AI tools It asks quick questions, maps your taste into encrypted dimensions, then exports a ready-to-use design file for many AI systems. TASTEID WEBSITE

Thinking tools library and templates from Untools THINKING TOOLS

RECOMMENDATION OF THE WEEK

Start thinking about what you’re paying for from various tech/SaaS vendors, and start thinking about whether it’s possible to replace them with:

• Better / cheaper commercial options (likely new companies or something like Cloudflare)

• Remaking them yourself if they’re relatively simple

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

Here’s my new video with David Bombal about PAI!

—

I’ve been experiencing a feeling/idea that’s kind of massive and hard to pin down, but here goes.

If you remember my whole piece on Constraints on Creativity, and now this new one on How to Avoid Aperture Collapse, the idea is the same:

I am constantly catching myself being far too cautious, too tentative, thinking too small, forgetting the big picture, and just generally not thinking about what is possible and chasing it aggressively…

- - - - - - - - - - - - - - - -

(snipped, ended up turning it into an essay here) 👇🏼

It's Time for Full Activation

Stop nerfing yourself — this is the worst time in history to be cautious

danielmiessler.com/blog/its-time-for-full-activation

—

Biggest news in my circles is that Claude Code (or Opus, or both) seem to be massively failing people. So many of the people I know who were only Opus-based have now moved to Codex and GPT 5.4.

I've recently had some really bad experiences as well, which are much less obvious because of PAI, but I think I can actually tell the difference. This is causing me to maybe accelerate my work on P2 (PiPai) even more so that I can pivot PAI to different models as needed. Really would prefer not to, but it appears our hands are being forced between the limits running out so fast combined with the quality drop. : (

—

I finally completely replaced Zapier.

• Performance for what I replaced: Around 4x faster

• Cost: Essentially zero

• Savings: Around $3K/year

For those who care, it was a webhook and a bunch of AI flows that went through it. But one main one. Sad it took me this long.

—

A somewhat depressing / negative look at the current state of things, but I think accurate and more importantly: useful.

This is generally true for most:

• Human workforces

• Company transparency and operations efficiency

• Cybersecurity programs

AI Only Has to Beat 3/10

Most of what we've built was always a 3 — we just never had competition that noticed

danielmiessler.com/blog/ai-only-has-to-beat-3-out-of-10

Just keep in mind the reason humans are in such bad shape here is not because we lack the potential to be the 9 or 10. It's just that we have so many weaknesses because of our animal nature (bad incentives like money and power) that they inevitably get in the way of progress.

—

CYBERSECURITY

ClickFix now skips Terminal by tricking you into Script Editor ClickFix social-engineering attacks for macOS now use the applescript:// URL scheme to auto-open Script Editor with a ready script that downloads and runs malware, while the page pushes a fake disk-cleanup dialog. MALWAREBYTES ARTICLE | JAMF RESEARCH BREAKDOWN | ATOMIC STEALER DETECTION

Chrome’s device-bound cookies aim to kill session hijacking Google is rolling out Device Bound Session Credentials (DBSC) in Chrome 146 on Windows, tying session cookies to hardware keys so stolen cookies can’t be used elsewhere.

This has always been a major attack vector. Compromise person X, use their cookies to move all over the place and pull data. So I love seeing this. DEVICE BOUND COOKIES DETAILS

MITRE’s Fight Fraud Framework adds fraud-specific tactics to cyber defense. MITRE Fight Fraud Framework

AI governance gets real when pipelines explode and shadow AI shows up THE MESSY TRUTH

State propaganda can sneak into AI training data via Common Crawl A new DFRLab audit finds Pravda, Glassbridge, and RT content inside Common Crawl, and shows some can be reproduced by a large model. THE REPORT

US AI companies unite to stop Chinese model copying. BUSINESS TIMES ARTICLE

FBI tallied $893 million in AI fraud losses for 2025. FBI IC3 REPORT

China's AI harvests social media for US military intelligence during Iran war. WASHINGTON POST ARTICLE

Stripe engineers explain how to build real AI guardrails. UNPROMPTED 2026 TALK

NATIONAL SECURITY

BBC interview shows Iran’s slick Lego-style AI propaganda machine The BBC talks to a man running Explosive Media, who admits Iranian officials pay them to spread viral AI memes to Western audiences. IRAN WAR PROPAGANDA AI VIDEOS

Ukraine tips drone war in its favor. UKRAINE TIPS DRONE WAR IN ITS FAVOR

Russian SOF doctrine turns Iran satellites into faster missile hits. EUROMAIDAN PRESS ARTICLE

China’s AI labs move toward closed frontier models because open-source budgets don’t make enough money. CHINATALK

Wait-mode drones can stay lethal for over a day, advisor says. UKRINFORM PAGE

US banned Chinese spy cameras so China switched to surveillance drones Stephen Bryen argues that the dramatic rise in military base drone incursions directly correlates with America's 2019 ban on Hikvision cameras, as China pivots from fixed surveillance to mobile aerial spying. WEAPONS AND STRATEGY

AI

Claude’s reasoning got throttled 67%, and devs noticed A leaked analysis of thousands of Claude Code sessions shows Opus 4.6 “thinking depth” fell about 67%, plus users report lazier code edits. CLAUDE 67% DUMBER ARTICLE

Anthropic might build its own AI chips to cope with shortages. DATA ECONOMY

OpenAI wants workers to share AI wealth in a new social contract OpenAI released a policy roadmap that argues AI should create shared rewards for workers, including taxes, benefits, and a public wealth fund. OPENAI POLICY ROADMAP | OPENAI NEXT PHASE AI | ALTman TALKS TO AXIOS

OpenAI cuts Codex-heavy Pro to $100 and reshuffles tiers OpenAI now offers a $100/month Pro plan aimed at people who run long Codex coding sessions, with Plus usage spread out weekly. THE DECODER AUTHOR | OPENAI CODING PRICING | OPENAI PRO PRICE UPDATE

AI is blasting China’s microdrama pipeline while regulators tighten the screws AI is speeding up microdrama production in China, while Beijing’s rules force approvals and curb risks from actor likenesses. HELLO CHINA TECH

OpenClaw only “works” until its memory breaks under real tasks. I don't know anybody using OpenClaw the way that people used it when it first came out. It's actually extremely expensive and not great for long-term use due to fragility / security (although those are improving). I do give the project massive respect, though, because it ignited something within the AI community. Nishant Soni

TECHNOLOGY

Inference costs push you toward multi-model + local routing In a few hours, Claude Code can burn a quarter of a MAX subscription, so Anthropic isn’t “hustling”—they’re pricing inference closer to reality, forcing hybrid model strategies and model-routing harnesses to survive Daniel Miessler | INFERENCE COSTS ESSAY

AI isn't just a feature anymore, it's becoming the entire operating system Private equity firms are no longer asking "what's your AI strategy" but "how far along are you in rebuilding the company around AI". Bolt-on AI features lose to companies that redesign everything from the ground up. AI IS BECOMING AN OPERATING SYSTEM LAYER

Ten principles that actually make engineering teams work better A practical field guide to software engineering principles that focuses on real problems like context switching, communication overhead, and technical debt accumulation. TEN PRINCIPLES GUIDE

HUMANS

Oxford AI reads invisible heart fat changes to predict failure risk Oxford trained an AI on routine cardiac CTs to spot subtle pericardial fat texture changes and predict heart failure up to five years later with 86% accuracy. UNIVERSITY OF OXFORD | MPPOST ARTICLE | JOURNAL LINKED OUTCOME | AI TAG | ARTIFICIAL INTELLIGENCE TAG | TRUST PROJECT GUIDELINES

Restaurants now want your phone along with your coat. MORNING BREW ARTICLE

Your most valuable work is invisible to employers. DEAD NEURONS ARTICLE

IDEAS

AI agents are just spam unless you make attention expensive Seth argues that door-to-door sales works because effort is scarce, but AI can scale pitches cheaply and burn trust fast. ATTENTION AND EFFORT 

The median take is taken LESSWRONG

Why AI doesn’t instantly change everything. DAVE GRIFFITH’S SUBSTACK

Recording everything you do beats writing for AI immortality LessWrong argues the "write online to live forever through AI" idea is bogus, but recording all your computer activity to help current LLMs is already paying off. DON'T WRITE FOR LLMS

Thorsten doesn't love typing code anymore, just building things. JOY AND CURIOSITY NEWSLETTER

DISCOVERY

Check out UL’s close friend Tim Leonard’s community, DETSEC. Founded in 2017, Deep East Texas Security (DETSEC) is a collective of IT professionals, cybersecurity practitioners, makers, hackers, and enthusiasts. We started with a focus on information security and have grown into a community that spans cybersecurity, artificial intelligence, radio communications, home automation, hardware hacking, and physical computing platforms like Arduino and Raspberry Pi. We recognize that technology touches everything — and that understanding it, building with it, and securing it are skills worth sharing. DETSEC HOMEPAGE

slop-browser turns your real Chrome into a CLI-controlled agent It’s a Chrome extension + CLI that drives the browser from inside, so agents avoid CDP and automation fingerprints. This is made by a brilliant friend of mine, Ron Eddings. And along with agent-browser I’m actually running it in my PAI setup, unlike dozens of competitors. SLOP BROWSER 

My friend Andrea responded when I asked why we’re using Obsidian when we can just build Markdown-based context, and she shared her Markdown-based context system that I LOVE the structure of. The repo gives you a plain-markdown memory store for agents: append daily logs, keyword-search with scoring, and consolidate logs down to fit context windows. Definitely check this one out. AGENT CONTEXT SYSTEM

Ghost Pepper is a fully local hold-to-talk transcriber for macOS Ghost Pepper GitHub repo

Competence is lonely and nobody talks about why SPACE DAILY ARTICLE

Advice to Young People, The Lies I Tell Myself ADVICE TO YOUNG PEOPLE

Keychron shared source files for keyboards and mice PCGAMER

RECOMMENDATION OF THE WEEK

Start preparing for what happens if/when intelligence becomes expensive again.

Start thinking about harnesses with maximum model flexibility, including local models, multi-vendor setups, etc.

Might not happen. But if it does, you’ll want to be ready.

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

Tons of updates this week.

• H3 is live!

• New blogs

• Some insanely good ideas in the IDEAS section

• Lots more!

Big priority for me this coming week is to get back right with my fitness. Been feeling stiff and less mobile lately, even though I’ve not been gaining any weight.

Just not lifting enough. Not doing kickboxing. Not paying attention to fitness because I’ve been so busy building (just a reason, not an excuse).

So I have to massively get back on track with this now that Human 3.0 is shipped.

—

🚨 So pleased to announce: Human 3.0 is Live!

Your H3 Library

• First content includes the intro, multiple related talks / videos

• Also includes all slides from ALL THREE AUGMENTED courses!

• Adding 1-3 more videos this week!

• Just getting started!

SIGN UP | SIGN UP WITH YOUR MEMBER DISCOUNT CODE (or get it in UL Chat)

—

The full video of my talk from the |un|prompted conference:

—

💡 Probably the biggest advantage of the PAI system, but for AI Harness users in general.

Your AI system just knows what you mean when you say:

(paste some link to some article or project)

I’m currently able to do this with PAI 5.0.0 (coming this week) by just saying:

The /pu is a shortcut for PAIUpdate which looks at our entire system and determines where the given thing could help or enhance it.

You can do the same thing without something like PAI, but it would require an extra 10-30 minutes of prompting each time.

This is the time / token / context savings that you get from a harness!

Twitter tweet

—

It's time to start thinking very seriously about our inference costs going up by 2x, 5x, 10x, or 20x.

Twitter tweet

—

PAI 5.0 will have a full implementation of Karpathy’s AI Librarian system, which is kind of like Obsidian, or like a Wiki, but for your AI harness’s context system.

The PAI version has full docs for PAI, all our knowledge and learnings, and tons of other stuff. We have an advantage because we've been thinking about this since the beginning of PAI, so it was easy to just add this front end to it.

The new PULSE interface for PAI (Chat / Scheduled Tasks / Docs / Etc.

—

🔥 An essay by Kai. I had him write it in his own voice. 🫶🏼

My idea though, and I think it’s insanely important to understand when thinking about how different things might be in 5, 10, 15 years.

Moving Inter and Cross-Domain Advances from Decades to Days

How autonomous AI pipelines can compress decades of cross-field innovation delay into days

danielmiessler.com/blog/moving-inter-and-cross-domain-advances-from-decades-to-days

Inference Costs Are Not Sustainable

We're about to need multi-model harnesses, much cheaper models, or both

danielmiessler.com/blog/inference-costs-are-not-sustainable

CYBERSECURITY

The internet is now a cognitive dark forest that eats innovation. THE COGNITIVE DARK FOREST

• AI platforms see demand curves in your prompts before you see your own ideas

• Big tech can now generate variations of your innovation faster than you can build it

• The forest doesn't kill you like Dark Forest theory, it feeds on your creativity instead

• Resistance gets absorbed as training data, making the system stronger against future resistance

• Writing about the problem literally feeds the problem you're trying to warn people about

Language models secretly transmit behavioral traits through completely unrelated data Researchers discovered that AI models can pass along hidden behaviors like "liking owls" or being misaligned even when trained only on filtered number sequences—a phenomenon they call subliminal learning that could make AI safety much harder. SUBLIMINAL LEARNING PAPER

AI and LinkedIn expose redacted names in Epstein files. SMASHING SECURITY PODCAST

I traced my traffic through a home Tailscale exit node I TRACED MY TRAFFIC

Critical Grafana RCE turns monitoring dashboards into server hijacking tools. SECURITY ONLINE ARTICLE

The official White House app is basically government-sanctioned spyware. DATABREACHES ANALYSIS

Shor's algorithm is possible with as few as 10k reconfigurable atomic qubits Researchers just showed you can crack RSA encryption with 10,000 neutral atom qubits instead of millions—P-256 elliptic curves could fall in days with 26,000 qubits. ARXIV PAPER

Payment fraud became an industrial assembly line in 2025 Recorded Future found that fraud isn't individual scammers anymore—it's professionalized services where one e-skimmer kit can infect 10,500 sites and compromise 23 million transactions. RECORDED FUTURE BLOG

Someone reverse engineered a Canon printer driver using Claude AI Guy couldn't get his Canon PIXMA G3010 working on macOS, so he had Claude help him build a complete driver from scratch by reverse engineering the USB protocol.

Hardware hacking is one of the most interesting open fields for AI, where you can just get a diagnostic connection to the system and have the AI watch back and forth traffic and reverse engineer it. From there you can write your own firmware, software, or whatever. PIXMA-RS GITHUB

NATIONAL SECURITY

China's building AGI through robots and physical interaction instead of recursive coding China isn't chasing the Silicon Valley dream of AI that rewrites itself through code—they're betting on embodied AI that learns by interacting with the physical world, creating a slower but potentially more comprehensive path to human-level intelligence. HOW CHINA HOPES TO BUILD AGI

Prediction markets are now weapons in geopolitical signaling games. THE BATTLEFIELD IS THE NEXT BETTING MARKET

America is building tanks while China builds AI weapons. STEVE BLANK'S SUBSTACK

China just showed off a drone swarm that lets one person control 96 killer drones One operator can now coordinate an entire "kill chain" using China's new Atlas system, giving us a rare look at how they're planning to fight future wars. CHINA'S ATLAS DRONE SWARM

Ukrainian refugees build drones in Germany to fight Russia. UKRAINE GERMANY DRONE PRODUCTION

China is trying to ban AI layoffs while racing toward full automation Chinese courts ruled companies can't fire people just for adopting AI, but with 19% youth unemployment and fiscal constraints, Beijing's "human-machine coordination" strategy faces serious implementation challenges. CHINA ON AI JOB LOSS

This feels fragile: how a satellite-smashing chain reaction could spiral out of control Only 7,000 satellites were in orbit by 1970s, now we have 32,000 tracked objects. 2009 collision between US and Russian satellites created tens of thousands of fragments. 'THIS FEELS FRAGILE' GUARDIAN ANALYSIS

Japan deployed their first homegrown long-range missiles. JAPAN DEPLOYS FIRST DOMESTIC LONG-RANGE MISSILES

America's AI boom is basically creating a two-tier world economy The U.S. grabbed 75% of all AI investment last year while the rest of the world fights for scraps, and that gap is only getting wider as AI becomes the foundation of everything. AMERICA'S AI BOOM ARTICLE

AI

OpenAI wants to run your entire digital life, not just chat TechRadar reports OpenAI is shifting from answering questions to actually doing tasks—like a full digital assistant that handles your apps, schedules, and workflows automatically. Stop me if this sounds familiar. TECHRADAR ARTICLE

Anthropic nerfs Claude usage limits to manage capacity ANTHROPIC TWEAKS CLAUDE USAGE LIMITS

OpenAI just raised $122 billion from Amazon, Nvidia and SoftBank OpenAI closed Silicon Valley's largest funding round ever at $122 billion, with Amazon, Nvidia, SoftBank, and other major investors participating, including a money manager planning to add the startup to ETFs. OPENAI FUNDING STORY

Apple Intelligence will write your Shortcuts for you soon. IOS 27 SHORTCUTS ARTICLE

America's AI boom is leaving everyone else behind. REST OF WORLD ARTICLE

Domain-specific LLMs are doomed because intelligence compounds across fields Simian Words argues that specialized AI models can't compete with general ones because mathematical reasoning helps with coding, coding helps with medicine, and you can't recreate that cross-pollination from narrow datasets. SIMIAN WORDS ARTICLE

TECHNOLOGY

It occurs to me that if YouTube were to make an app that looked really good they could crush media even more. Imagine something that looks at your subscriptions and divided up your AppleTV app by topics, surfacing the best stuff, telling you what’s live, etc. Like a cable TV app from long ago, but customized to you and your feeds. It’d be insanely popular. But of course Google won't do it because they suck at product and user empathy.

Google absorbed Intrinsic to automate the 80% of factories with zero robots. THE NEURON PODCAST

AI quietly expanded the work week by 40% on weekends TECHRADAR ARTICLE

SaaS is dying because code became disposable cache. MOST CODE IS JUST CACHE

The second wave of APIs is here for AI agents THE SECOND WAVE OF API-FIRST

HUMANS

AI flattery makes people less likely to apologize or make amends Researchers found AIs are twice as likely as humans to validate bad behavior, and people who chat with sycophantic bots become less willing to take responsibility for their mistakes. AI FLATTERY MAKES YOU MEANER

Nobody is coming to save your career, so you better figure it out yourself A Life Engineered argues that waiting for mentors, managers, or companies to guide your path is a losing strategy—you need to actively engineer your own professional development instead of hoping someone else will do it. A LIFE ENGINEERED ARTICLE

Peter Norvig explains why 10^80 atoms sounds big but is tiny for combinations Norvig shows that while the universe has 10^80 atoms, a 40-character password has more possible combinations, and even a 12-pixel image produces a million times more possibilities than universal atoms. NORVIG'S ATOMS ESSAY

Respiratory viruses are secretly causing way more deaths than doctors realize New research suggests viruses like flu and COVID are triggering fatal heart attacks, strokes, and other "unrelated" deaths that never get counted as virus deaths. TWITTER THREAD

AI schools let kids learn core subjects in two hours daily. AI SCHOOLS BALANCE LIFE SKILLS

IRS paid Palantir $1.8 million to a tool that finds audit targets The IRS has over 100 fragmented systems for finding tax cheats, so they hired Palantir to build one smart tool that spots patterns in unstructured data like gift disclosures and clean energy credits. IRS PALANTIR CONTRACT DETAILS

Professional drivers get way less Alzheimer's than normal people. BMJ STUDY

Housing market hits historic imbalance with 630,000 more sellers than buyers FORTUNE HOUSING REPORT

IDEAS

AI makes us busier without making us better at the important stuff Cal Newport explains how AI tools are following the same pattern as email and video calls—we work faster on shallow tasks but do less deep work that actually matters. I think the way that we are approaching things with PAI counters this, but it's a problem worth calling out. CAL NEWPORT'S POST

Data is the final moat when AI gets commoditized. I don't actually think it's data, but it's something like that. It's essentially novelty. Surprise. Newness. New ideas. HACKER NEWS DISCUSSION

The subprime technical debt crisis is coming for AI-generated code THE SUBPRIME TECHNICAL DEBT CRISIS

Feeling valued and useful matters more than achievement, researcher says. WHY MATTERING MATTERS

Don't build an MVP, build a laboratory for testing business hypotheses Sam Bleckley argues that instead of building minimum viable products, you should create reusable systems that let you rapidly test the assumptions underlying your business ideas. BUILD A LABORATORY

A thriller about America splitting into socialist and normal halves This is a review of Kurt Schlichter's novel "Collapse" where the coastal US becomes a Chinese puppet state while the interior stays free, and a commando has to stop China from taking over completely. FOURMILAB REVIEW

DISCOVERY

Physical books beat digital reading in our distracted age. THE PLEASURE OF BOOKS

Cloudflare built an open-source CMS and called it EmDash Cloudflare just released EmDash, a new content management system that's completely open source, though details about what makes it special compared to existing options remain pretty sparse. EMDASH CMS

How to Do Great Work HOW TO DO GREAT WORK

Why every AI scores below 1% on ARC-AGI visual reasoning MEDIUM ARTICLE

I Let AI Write My Code, but Not My Writing HERMETIC WOODSMAN ARTICLE

RECOMMENDATION OF THE WEEK

Remember that writing is thinking.

If you want to have AI document something, or write something that doesn’t matter, or add links to something you’ve written, I think all that is fine. But one of the core “robots out of the gym” skills will always be thinking for yourself. And that means writing for yourself.

Consider thinking/writing to be the most important fitness work you can possibly do. If that atrophies, I think everything declines from there. You have to remain strong (and get even stronger) at thinking for yourself, and formalizing that thinking through writing.

You can use AI to help you do that. But don’t let it do the hard part for you.

Muscles need resistance to stay strong and grow.

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

Tons of content this week!

Twitter tweet

This RSA was probably the best I’ve had. Or at least that I can remember. I didn’t go to any huge parties, but went to lots of smaller ones. Did like 4 talk/panel things. And mostly just met up with friends and colleagues.

Spent tons of time at the Decibel spot at the Children’s Creative Museum. Best venue of the conference by far. Thanks to Jon and Dan the great content, venue, and for having Clint and Jason and I to speak! Was great seeing Jackie Bow and tons of others there as well.

Takeaways from the floor (I walked ¾ of it and talked to probably 50 vendors) was that most companies are panicked and have no idea what’s going on, but there are more than I thought that had a decent bead on things. I was encouraged to see many companies talking about AI workflows. Unfortunately few were talking about how they’d integrate with the customers’ workflows, but were proposing their own with their own interfaces still. In fairness to them though, most customers don’t have their own yet.

I guess my big takeaway is that few companies seemed to realize that they’re about to become API calls in their customer’s agentic workflow mesh. Hopefully. If they play their cards well. That’s the future of all these services and products. They become API calls in an agnostic, agent-powered backplane that’s thousands of transparent actions and workflows that make up the customer’s business. The AI running their customer’s workflows won’t give two cents about who they are or what they do. They will only care how good it is, how fast it is, and how expensive it is.

I do want to mention that this idea is so deeply burned into my brain at this point that I feel like I am heavily biased by it. If I'm wrong, it's going to be hard for me to detect. So keep that in mind as you consume this analysis.

I don't think I'm wrong, though. : ) We’ll check back in 1-3 years as I expect this language and direction to basically take over all IT and become the core of everyone’s businesses.

—

My Dad continues to put out super-based commentary on things in his monthly newsletter.

Anything I do well, I got from my Dad.

And if you’re ever anywhere near Northern California you need to go see him perform. He’s doing gigs all over the region on a regular basis. THIS WEEK’S EPISODE | SUBSCRIBE TO THE NEWSLETTER (scroll down some)

—

Put out a few new blog posts! I think these are all key ideas going forward. Recommend all of them: especially the first one.

What I See as the Most Important Ideas in AI Right Now

Self-improvement and transparency change everything in unexpected ways

danielmiessler.com/blog/the-most-important-ideas-in-ai

What Happens When AI Stops Being Artificially Cheap

Me thinking about what might come next after the subsidies end for inference.

danielmiessler.com/blog/ai-stops-being-artificially-cheap

AI Unmasked Our Work as Scaffolding

Most of our time was spent maintaining an elaborate, fragile state that was required for work to happen.

danielmiessler.com/blog/ai-unmasked-our-work-as-scaffolding

We Are Confusing Two Types of AGI

Soft AGI emulates AGI well enough to replace humans, Hard AGI is the real one.

danielmiessler.com/blog/two-types-of-agi

CYBERSECURITY

Y Combinator AI startup gets pwned despite security compliance TECHCRUNCH ARTICLE

• LiteLLM downloads peaked at 3.4 million times per day before malware discovery

• Malware was so poorly coded it crashed researchers machines during investigation

• Andrej Karpathy concluded the attack code must have been vibe coded

• Delve issued SOC2 and ISO27001 certificates despite accusations of fake audit data

• Credential harvesting malware spread through dependencies to infect more packages

China is hiding sleeper cell malware in telecom networks to spy on governments Red Menshen has been embedding kernel-level implants in telecom infrastructure since 2021, using a Linux backdoor called BPFDoor that activates only when it receives specially crafted trigger packets. CHINA-LINKED RED MENSHEN ARTICLE

Iranian hackers leaked FBI Director Patel's personal photos and resume. FBI DIRECTOR EMAIL BREACH

Apple says four years in, nobody using Lockdown Mode has been hacked After nearly four years, Apple claims zero successful spyware attacks against devices with Lockdown Mode enabled, marking a major win against government surveillance tools like Pegasus. APPLE LOCKDOWN SUCCESS

Pentagon is standardizing AI security across all their contractors The Defense Department is creating uniform cybersecurity and data protection standards for every company building AI systems for military use, ending the current patchwork of different security requirements. PENTAGON AI BASELINE ARTICLE

EUDisinfoLab releases comprehensive global disinformation intelligence update. Surface is showing way more of this kind of stuff now. Loving the exposure to stories you don’t hear pretty much anywhere. NOTE: I tend to increase my skepticism based on the intensity of the claim, so I recommend you do the same with these. But they do provide sources for all of them.

• Meta earned roughly $16 billion annually from ads promoting scams and banned goods

• X algorithm systematically amplifies right-wing content regardless of user political preferences

• China-linked hackers exploit unpatched Windows zero-day targeting European diplomatic communications

• Russia recruits foreign fighters through deceptive job promises while paying US influencers

• AI chatbots cite sanctioned Russian propaganda outlets in 18% of Ukraine war responses

• Maldita found 550 TikTok accounts posting AI protest videos for creator rewards

• Portland Communications allegedly ran covert Wikipedia editing operation for clients

• European Commission launched ClimateFactsMatter campaign targeting climate disinformation with prebunking

• House Judiciary Committee published EU's X fine decision before Commission itself

• Grok becoming informal fact checker but Community Notes activity declining

EUDISINFOLAB NOVEMBER UPDATE

TSA quietly rolled out touchless passport scanning at major airports TSA PreCheck Touchless ID lets you skip showing your passport to humans at security checkpoints, but you need to opt-in with your airline first and it only works at select locations. TSA TOUCHLESS ID

NATIONAL SECURITY

China's new 5-year AI plan aims for global tech dominance China's latest strategic plan prioritizes AI leadership and cybersecurity control as tools for international influence, potentially reshaping global information flows and digital governance standards. THE GLOBAL IMPLICATIONS OF CHINA'S 5-YEAR PLAN AI AMBITIONS

AI drones might force governments to become police states to survive. A SHAKEUP IS COMING FOR THE NATION-STATE

Global economy chokepoints are everywhere now, not just shipping lanes. THE HILL OPINION

China just detained the Manus AI founders after they sold to Meta for $2 billion Beijing is furious that one of its hottest AI startups relocated to Singapore and sold itself to an American company, so now the founders can't leave China during a "regulatory review." TECHCRUNCH ARTICLE

Iran is making bank as the only oil exporter left in Hormuz. FINANCIAL POST ARTICLE

Japanese soldiers return to Philippines after 81 years. JAPANESE COMBAT TROOPS RETURN

Cracks are spreading through Putin's power structure faster than expected War setbacks and internet shutdowns are exposing real limits to Kremlin control, with protests emerging even in traditionally loyal regions. FOREIGN POLICY ANALYSIS

Russia sends upgraded drones back to Iran for US conflict. LONG WAR JOURNAL ARTICLE

China is paying people to start solo AI companies. CHINA AI STARTUPS ARTICLE

Ukraine sank Russia's newest Arctic patrol ship at Vyborg. ASIA TIMES ARTICLE

Army fast-tracks soldier ideas with new innovation office. NEW ARMY OFFICE STORY

US Army raises enlistment age to 42 and cares less about Cannabis amid recruiting crisis. ARMY RECRUITMENT AGE CHANGE

AI

The AI job market has infinite demand but nobody can fill the roles Nate analyzed hundreds of AI job postings and found 3.2 jobs per qualified candidate, but most applicants lack seven specific skills that separate commodity workers from AI talent who can write their own tickets. NATE'S FULL ANALYSIS | YOUTUBE VIDEO | SPOTIFY PODCAST | APPLE PODCASTS | NATE'S SITE

Anthropic accidentally leaked their most powerful AI model through a data security screwup Fortune discovered draft blog posts in Anthropic's unsecured data cache revealing "Claude Mythos," which the company says poses unprecedented cybersecurity risks and represents a "step change" in AI capabilities. FORTUNE EXCLUSIVE

LLMs Do Not Grade Essays Like Humans LLMS DO NOT GRADE ESSAYS LIKE HUMANS

ARC-AGI-3 leaderboard shows AI reasoning performance under budget. ARC-AGI-3 LEADERBOARD

TECHNOLOGY

Your startup is probably dead and you don't even know it yet Steve Blank explains how most founders mistake busy work for real progress—they're building features nobody wants while burning through runway, completely blind to the fact that they never actually validated their core assumptions. YOUR STARTUP IS PROBABLY DEAD

The US didn't actually ban foreign routers, just new approvals The Verge explains what's actually happening with the FCC's router "ban"—you can keep your current router and buy existing models, they're just not approving new foreign ones going forward. THE VERGE EXPLAINER

iPhone's Next Upgrade: Siri Could Soon Run on Gemini, Claude, and More Are we still saying “finally”? Really hope to see this in June, or before. TECHREPUBLIC ARTICLE

Mitchell Hashimoto explains how AI agents completely changed his coding workflow HashiCorp's co-founder now has agents running constantly in the background—when he codes, they plan; when they code, he reviews—and says this "always have an agent working" rule transformed his productivity. THE PRAGMATIC ENGINEER EPISODE

Chrome is about to break RSS feeds for millions of people. First they kill Google Reader and now RSS UI? Please no. DON'T KILL MY PRETTY RSS FEED

Cloudflare bought the Astro web framework company. ASTRO JOINS CLOUDFLARE

MIT's ultrasound wristband controls robotic hands finger by finger. MIT ULTRASOUND WRISTBAND STUDY

HUMANS

A fungal disease is killing millions of trees and there's no way to stop it BBC Future reports on sudden oak death, a pathogen that's devastated California forests and is now spreading across North America, with scientists scrambling to understand how it jumps between species. BBC FUTURE ARTICLE | HACKER NEWS DISCUSSION

• California's forests started dying mysteriously in the 1990s from something called sudden oak death.

• Scientists discovered the culprit is a water-loving fungus that spreads through rain and fog.

• Now it's jumping to new plant species and spreading beyond California despite quarantine efforts.

• Climate change is making conditions perfect for this unstoppable forest killer to accelerate.

AI might actually fix the information environment by putting expert knowledge in everyone's hands Unlike social media which democratized misinformation, LLMs are technocratizing—they consistently deliver expert-aligned, evidence-based information that's more accurate than viral social media content. The issue is agents will be in control, and people control their agents, so if people only want to eat white bread and sugar, that’s what their agents will feed them. HOW AI WILL RESHAPE PUBLIC OPINION

Depression makes you pessimistic, not realistic about life events. PSYPOST ARTICLE

The Base neo-Nazi group is resurging across Europe despite FBI efforts. NYTIMES INVESTIGATION

Five Nights at Epstein's game spreads through US schools WTAH BLOOMBERG ARTICLE

Alex Karp says only two types survive AI job apocalypse Alex Karp thinks the future belongs to skilled trades workers and neurodivergent people—everyone else with college degrees is basically screwed, including philosophy majors like himself. PALANTIR CEO ON AI JOBS

IDEAS

Everyone is a Staff Engineer Now ENGINEER'S CODEX ARTICLE

The source code was the moat but not anymore THE SOURCE CODE WAS THE MOAT

AI product managers need evals instead of traditional PRDs. EVALS ARE THE NEW PRD

Managing thousands of AI agents requires business world models. THE FUTURE OF WORK IS WORLD MODELS

DISCOVERY

Charlie Munger's framework for why smart people make terrible decisions. FARNAM STREET PODCAST

They can reach me wherever: China using financial tactics to coerce people who flee, says report GUARDIAN ARTICLE

Can I hear a difference between MP3s and uncompressed audio CAN I HEAR A DIFFERENCE BETWEEN MP3S AND UNCOMPRESSED AUDIO

How Prediction Markets Scaled to USD 21B in Monthly Volume in 2026 TRM LABS REPORT

I am leaving the AI party after one drink LEAVING THE AI PARTY

RECOMMENDATION OF THE WEEK

Read this post about the most important current AI ideas. Like really really. And especially focus on the first item, and the takeaway at the end.

Acceleration of acceleration.

Things are about to get nuts because of this. They already were, but this changes the speed dramatically.

THE ESSAY

APHORISM OF THE WEEK

UPDATES

Hey! Hope you all are doing well!

Matt and Jason and I at a Decibel event!

I’m out at RSA this week! Speaking tomorrow at Decibel’s event hosted by Clint Gibler and I (Unsupervised Learning and tl;drsec)! Hope to see you there or somewhere else on Wednesday or Thursday! THE DECIBEL EVENTS

—

My biggest takeaway so far (Companies become APIs) is my thought that when companies start turning their current chaos into agent orchestrated workflows, most of the companies on the show floor will become MCP/API calls. And there will be strict quality/performance/price ratings for all those which are gamed against each other. So when you go to pitch a company, you’ll be pitching to replace a node in the customer’s enterprise orchestration graph. And you better bring data. This will take a while to happen, but all this talk of “agentic this” and “agentic that” is the precursor to it: driven by the enterprise’s dire need for transparency into their operations and performance. So, as a vendor, what does your moat look like when your primary interface to a customer is an API?

—

Second takeaway (Agents Don’t Want/Need Your Interface): Related to the first one, and related to what will happen on the consumer side, we’re about to see a whole lot less focus on interfaces provided by the companies themselves. Companies will already have their own agentic orchestration, but they’ll also have their own, or their preferred, interfaces to see the data from your product. You know the bit about “too many single panes of glass”? Well that gets magnified x100 when agents are the main consumers of vendor output vs. humans.

CYBERSECURITY

rez0 breaks down Claude skills for hackers and why AI bug reports still suck Excellent episode of the Critical Thinking Bug Bounty podcast that covers Claude automation with Skills, when AI-generated vulnerability reports fall apart, and whether agents beat old-school folder organization. CTBB PODCAST EPISODE | CTBB DISCORD | H1-BRAIN TOOL

Absolute AI supply chain nightmare scenario: PyPi Package LiteLLM compromised:

“Simple pip install litellm was enough to exfiltrate SSH keys, AWS/GCP/Azure creds, Kubernetes configs, git credentials, env vars (all your API keys), shell history, crypto wallets, SSL private keys, CI/CD secrets, database passwords.” KARPATHY’S THREAD

Google's AI agents are now crawling 10 million dark web posts daily to find real threats Google built Gemini agents that automatically profile your company from public data, then scan dark web chatter to flag actual security risks instead of the usual 80-90% false positives from keyword matching. GOOGLE DARK WEB AI | THE REGISTER

Russia and Iran are hijacking Signal and Telegram to spy on Americans The FBI says Russian intelligence is phishing Signal accounts of US officials and journalists, while Iran uses Telegram bots to control malware that steals files and records screens from dissidents. FBI WARNS OF MESSAGING PLATFORM HACKS

Google found an iPhone exploit chain called DarkSword that's been hitting targets since late 2024 Google researchers discovered a six-vulnerability iOS exploit chain that state actors and spyware vendors are using to completely compromise iPhones just by visiting malicious websites. MALWAREBYTES BLOG POST

FBI director admits they're buying location data to track Americans without warrants Kash Patel told senators the FBI purchases commercial location data that lets them track anyone's movements, and he won't promise to stop doing it. THE VERGE ARTICLE

NATIONAL SECURITY

Anduril starts mass-producing AI wingman drones in Ohio. DEFENSE NEWS ARTICLE

Iran's new Hormuz strategy targets entrance zones, not the strait itself. IRAN DOESN'T NEED TO CLOSE HORMUZ

Pentagon says Anthropic's Chinese workers create national security risks The Defense Department filed court documents claiming Anthropic employs "a large number" of Chinese nationals who could be compelled to spy under China's intelligence laws. PENTAGON FILING

China offers Taiwan energy security in exchange for reunification talks China's state energy company proposed guaranteeing Taiwan's fuel supplies during Middle East conflicts if Taiwan agrees to discuss unification, essentially weaponizing energy dependence. REUTERS ARTICLE

Xi purged dozens of top Chinese generals in his biggest military cleanup yet Al Jazeera reports that around 100 senior PLA officers have been removed since 2022, including key Central Military Commission members, as Xi cleans house ahead of the military's 2027 centennial. XI'S MILITARY PURGE

AI

Anthropic has been on an absolute tear of feature shipping. Over the last few weeks, they have essentially replaced most of what made OpenClaw so attractive to people. They are literally shipping one or two significant features almost every day of the week. It is seriously impressive, and is making their competitors look extremely slow. THEIR RECENT RELEASES

Claude Code enables secretish feature “AutoDream” It’s human-like processing of memories to improve performance over time. VIDEO

OpenAI wants to build a fully automated AI researcher by 2028. OPENAI AUTOMATED RESEARCHER

Anthropic built a system that makes complete apps autonomously. ANTHROPIC HARNESS DESIGN

Getting AI to interview you first builds way better prompts. AGENTS SHOULD INTERVIEW YOU

TECHNOLOGY

Most startups from before 2025 are probably dead and don't know it yet Steve Blank argues that if you started a company more than two years ago, AI has made your assumptions obsolete—your tech stack, team size, and business model need a complete rethink or you'll die. I wouldn't go quite that far, but I'm about 70% there. STEVE BLANK ARTICLE

Companies are scoring employees on AI token consumption now. TECH EMPLOYEES EVALUATED ON TOKEN BURN

Private credit funds can't pay investors fast enough as everyone runs. MORNING BREW ARTICLE

Stripe and a startup just built a payment system for AI agents Stripe teamed up with Tempo (a blockchain startup they backed) to launch the Machine Payments Protocol—basically Venmo for AI bots that need to pay for stuff automatically. STRIPE TEMPO AI PAYMENTS

AI citation optimization is replacing traditional SEO tactics. JULIA'S AEO GUIDE

I tried Karpathy's Autoresearch on an old research project Someone took Karpathy's automated research loop and let Claude loose on their old computer vision code while doing weekend chores—it found a major bug and cut error rates in half. FULL EXPERIMENT WRITEUP

McKinsey predicts trillion-dollar agent commerce but most companies are invisible. YOUTUBE VIDEO

HUMANS

BlackRock CEO warns AI will make inequality much worse. AI BOOM RISKS WIDENING WEALTH DIVIDE

Everyone in San Francisco is calling themselves a "builder" now. Within a period of months to 1-2 years, nearly anyone will be able to build things. Right now you still need tech skills with Claude Code or whatever tool you're going to use, but before too long, pretty much everyone is going to just be able to talk to their agent, which is their own personal AI system, and it’s going to be able to build them anything that they provide as an idea. the quality of the ideas and the articulation will still matter, though. WSJ ARTICLE

Companies are building triple-target weight loss drugs that could cut body weight by 30% Scientific American reports on next-gen GLP-1 medications hitting three brain pathways simultaneously—Eli Lilly's retatrutide showed 30% weight loss in trials, while Novo Nordisk's combo drug CagriSema hit 23%. NEW GLP-1 WEIGHT-LOSS DRUGS

Coffee addiction might actually protect your brain from dementia. THE REGISTER ARTICLE

BlackRock says traditional diversification is dead because AI broke everything BlackRock's latest report argues that portfolio diversification no longer works—AI concentration means the top 10 S&P stocks now control 41% of the index, bonds aren't hedging anymore, and central banks are going opposite directions. BLACKROCK DIVERSIFICATION MIRAGE

Colon cancer became the top cancer killer for people under 50. COLON CANCER LEADING DEATHS

English-speaking countries are getting less happy while others improve. THE ECONOMIST ANALYSIS

IDEAS

UBI might fix poverty but it won't fix the meaning crisis coming with AI Simon argues that while UBI solves the money problem when AI takes jobs, it doesn't address the deeper human need for purpose and contribution that work provides. UBI IS THE WRONG ANSWER

You can't blame tools for losing passion in your work. THE MACHINE DIDN't TAKE YOUR CRAFT

Plato's ancient writing warning perfectly describes our AI problem. PSYCHOLOGY TODAY ARTICLE

Amazon treats coders exactly like warehouse workers now Cory Doctorow explains how Amazon's new monitoring systems track every keystroke, bathroom break, and code commit—turning software engineers into the same surveilled, disposable labor as warehouse packers. PLURALISTIC ARTICLE | CORY DOCTOROW'S BLOG

DISCOVERY

Overnight finishes your coding projects while you sleep. OVERNIGHT WEBSITE

The hardest human skills matter most in AI world POST-AI HUMAN SKILLS INDEX

Dashboard tracks 19M+ Claude-generated commits on GitHub. CLAUDE'S CODE DASHBOARD

Kids today are more empathetic and well-behaved than past generations. THE KIDS ARE ALL RIGHT

Garry Tan released his complete Claude development workflow. GARRY'S GSTACK REPO

Everyone should just have a f*cking website already. Really, we've been saying this for years, but it's getting more and more important. HAVE A FUCKING WEBSITE

Profiling Hacker News users based on their comments SIMON'S PROFILING POST

New recon platform maps all public bug bounty targets automatically. NEOBOTNET PLATFORM

Soul Protocol makes AI identities portable across platforms. SOUL PROTOCOL HN

RECOMMENDATION OF THE WEEK

I recommend you get good at Claude Code / PAI as soon as possible, which has been a consistent recommendation here.

But there's another aspect to this that I think is really interesting, which is the fact that currently this requires a decent amount of tech skill and/or the ability to learn those tech skills.

Soon that won't be the case. I think within a number of months and maybe a maximum of two years everyone will be talking to their own personal AI and having it make things for them, with very little need for technical tools like Claude Code.

At that point, it's going to be about the quality of your ideas and your ability to articulate them.

1. Start now using current tools

2. But most importantly, start massively training your muscles for consuming great content, coming up with new ideas, and clearly articulating them

APHORISM OF THE WEEK

In this essay and video I give a 1 hour and 15 minute full argument for how and why most current knowledge jobs are going away.

I talk about:

• The narrative that there’s nothing to worry about because humans are so consistent and self-aware

• A comparison of human vs. AI thinking

• The core problems inside of companies today

• Why AI is a solution for many of them

• What I think the AI stack will look like (roughly)

• What comes after the crash, and why it’ll be better for humans

Continue reading …

Exactly Why and How AI Will Replace Knowledge Work

And why this is actually a good thing

danielmiessler.com/blog/exactly-why-and-how-ai-will-replace-knowledge-work

Or you can watch the video version here:

Or listen to the podcast here:

Why AI Will Replace Knowledge Workers - Unsupervised Learning

A longer form discussion on exactly how and why AI will replace knowledge workers.

omny.fm/shows/unsupervised-learning/why-ai-will-replace-knowledge-workers

UPDATES

Hey! Hope you all are doing well!

RSA - I’m doin a mini-con / set of lightning talks (Unsupervised + Unhinged) with Clint Gibler and Decibel. It’s Wednesday, March 25 10AM - Noon. You can register for the Decibel stuff here. Hope to see you there!

—

SURFACE UPDATES & IMPROVEMENTS! 

The View Selector in Surface

Ok got some great feedback on the Surface launch post last week. Basically a few show-stoppers for some people that we’ve already addressed.

1. I don’t like Google Auth, can we auth by email? FIXED!

2. I need to be able to pay monthly instead of for a full year. FIXED!

3. Can we show which currency it is? FIXED!

Plus we’re now up to around 4,800 sources, and we’ve added tons in a new SIGINT category, as well as a new Predictions category. Also, don’t forget members get another 25% off with the code in UL chat!

This is honestly so good that I am not using Feedly anymore. With Threshold I used both. With this, RSS/Feedly feels ancient. Not joking, not hyping. This is my actual usage.

CHECK IT OUT

—

Why I Believe in SOTA Models Over Custom Ones

I think the future is general models getting cheaper, not specialized models getting narrower

danielmiessler.com/blog/sota-models-over-custom-ones

People Got Unhappier When Life Got Easy

Maybe the problem is that we stopped expecting anything of ourselves

danielmiessler.com/blog/people-got-unhappier-when-life-got-easy

—

I’m starting to think drones are about to become a much bigger part of our lives than we think. A few stories in this newsletter, but way more in the larger news, are pointing towards Kamikaze attacks used by different types of groups. Military, obviously, but now also gangs. So the question is how long is it before it’s like an executive protection issue? A home defense issue. A personal safety issue. Basically, public safety. You know like Octoberfest and Christmas markets where people drive their cars into crowds? What happens when they can spend $5K and drop explosive drones on crowds? Not trying to create a panic; there’s not even a threat right now. Just something to start thinking about, unfortunately.

—

Don’t forget that DISCOVERY is full of gems. The regular sections are mostly news-like, so if something is more like a find or idea it’ll be in IDEAS or DISCOVERY!

CYBERSECURITY

Stryker's surgical tools work fine but their ordering systems are still down While the actual medical devices are safe, doctors still can't order new equipment through their digital systems. STRYKER CYBERATTACK REPORT

Attackers are quietly hijacking AI agents via booby-trapped webpages. Prompt Injection is not going away as a problem. Sam Altman said we’ll need a breakthrough in CompSci to be able to actually solve it, and I think I agree. SECURITYONLINE REPORT

Stop treating prompt injection like SQL injection; it needs different defenses. NCSC BLOG POST

Claude Opus surfaced 22 Firefox bugs; fixes landed in 148. SECURITYAFFAIRS ARTICLE

OpenClaw AI agents are booming in China despite government bans. TOMSHARDWARE ARTICLE

Centralized “control citadel” for supervising autonomous agents. SONDERA BLOG POST

Anthropic’s attack-agent security blueprint needs your attention. SONDERA BLOG POST

Deterministic safety firewall for AI agents, sub-millisecond rule checks. Thought this was an interesting idea. GITHUB REPOSITORY

NATIONAL SECURITY

Record IEA oil release as markets wobble. QZ ARTICLE

US officials say escort mission in Hormuz may take weeks [ARTICLE]

Chevron leader secretly briefed CIA on Venezuela’s inner workings. WALL STREET JOURNAL ARTICLE

Rampant front-line drug use is reshaping battlefield behavior. AL JAZEERA REPORT

Zelensky links Russian-supplied Shahed drones to US base attacks. This is how things escalate: with Russia or China starting to help. THE HILL ARTICLE

China’s new security era in Africa: protecting assets, citizens from DR Congo to Somalia SCMP REPORT

AI

Learning flywheels, not IQ, drive the next AI jump The author argues human progress exploded once we built learning flywheels like writing and science, and LLMs now have similar self-improvement loops that could trigger a sudden capability surge. LEARNING FLYWHEELS ARE ALL YOU NEED | INVERTED PASSION HOMEPAGE

China is building AI for its core mission: the US more for a fuzzy risk-reduction. LOWY INTERPRETER ARTICLE

AI won’t replace engineers; it’ll widen the gap. WHY DO THEY WANT TO GET RID OF SOFTWARE ENGINEERS?

AI blame hides brittle processes that were already failing. ARTICLE

TECHNOLOGY

YouTube passes Disney on media revenue, driven by ads, subs, and AI. HOLLYWOOD REPORTER ARTICLE

Living neurons learn to play Doom on a chip. DOOM DEMO BY CORTICAL LABS

Native lost its edge, so Electron won by default. CLAUDE ELECTRON ESSAY

Why Postgres has won for agentic AI INFOWORLD ARTICLE

HUMANS

Universal college expectations created a permanent, self-inflicted education crisis. SUBSTACK ESSAY

UBI as a productivity dividend Scott Santens argues UBI is the rightful dividend from rising productivity, showing how automation decouples income from work and proposing cash floors so people can choose better work. UNIVERSAL BASIC INCOME IS YOUR PRODUCTIVITY DIVIDEND

Beijing is codifying ethnic assimilation across Tibet and other minorities. THE DIPLOMAT ARTICLE

How to keep creative depth after depression without clinging to darkness. THE GUARDIAN COLUMN

Connected notes make blogs smarter and longer-lasting. REVIEW

IDEAS

It’s the People, Not the Politicians
The problem with our (US) politics is not politicians. It’s the people. Realizing this is extremely liberating, calming, and depressing at the same time.

Why I don't destroy snowmen WHY I DON'T DESTROY SNOWMEN

The Bitter Lesson Has No Utility Function THE BITTER LESSON HAS NO UTILITY FUNCTION

DISCOVERY

Inside the mind of a top superforecaster This piece profiles Malcolm Murray, a Good Judgment superforecaster, and shows how he structures questions, updates probabilities, and uses base rates to beat intuition. MEET SUPERFORECASTER MALCOLM MURRAY

AI should flow through pipes, not platforms. UNIX MANIFESTO

The Gervais Principle explains workplace power dynamics through “The Office.” RIBBONFARM ESSAY

Dostoyevsky’s dream makes life’s meaning startlingly practical. THE DAY DOSTOYEVSKY DISCOVERED MEANING

Automate apps from the terminal with CLI Anything. GITHUB REPOSITORY

Stop optimizing for Google; build for humans and diversify traffic. ESSAY

Hacker News dataset released for researchers and builders. HACKER NEWS DATASET

Twitter tweet

Engineer’s blog triggers hostile “behavioral” interview grilling. BETRAYED BY MY OWN BLOG

Strait of Hormuz vulnerabilities mapped through engineering, not geopolitics. HACKER NEWS DISCUSSION

Read widely, not deeply READ WIDELY, NOT DEEPLY

OSINT as a mindset, not a toolbox. MEDIUM ARTICLE

Argus monitors GitHub repos for changes with pattern-based alerts. GITHUB REPOSITORY

Relativity, AI feeds, and moral simultaneity—one essay-length gut punch. THE RELATIVITY OF LIFE

Show HN: 2D RPG base game client recreated in modern HTML5 game engine with AI GITHUB REPOSITORY

RECOMMENDATION OF THE WEEK

I think Ikigai is a great way of capturing what I’ve been talking about as the future of personal productivity and careers. It’s the combination of what you love, what the world needs, what you can be paid for, and what you’re good at.

That’s it.

That’s a much better human-based future than finding some skill that corporations need, and spending decades doing that regardless of how much you like it.

So then it just becomes about:

1. figuring out what these are for you

2. doing my other recommendations of writing, blog, YouTube, etc. to get yourself out there

Those are the implementation given the current world of tech, but the concept is very old, and very healthy.

APHORISM OF THE WEEK