Welcome to another edition of Horizon AI,

Upwork is now integrated directly into ChatGPT. In today's issue, we'll share how growing businesses can use the new app integration to describe their work needs, discover expert talent, and draft job posts without leaving ChatGPT.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• OpenAI Launches GPT-5.4-Cyber for Defensive Security Work

• Anthropic Prepares Opus 4.7 and New AI Design Tool

• AI Tutorial: How to use the Upwork app in ChatGPT

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

OPENAI

OpenAI Launches GPT-5.4-Cyber for Defensive Security Work

OpenAI has introduced GPT-5.4-Cyber, a new model variant fine-tuned specifically for defensive cybersecurity work, following Anthropic's announcement of the frontier AI model Mythos.

Details:

• GPT-5.4-Cyber is a specialized version of GPT-5.4 with fewer restrictions on ​sensitive cybersecurity tasks such as vulnerability research and analysis.

• The model is designed to reduce "unnecessary friction" for legitimate security work, after partners flagged earlier models over-refusing dual-use cyber queries.

• The company is also expanding its Trusted Access for Cyber program to thousands of verified individual defenders and hundreds of teams protecting critical software, with new tiers to be added.

• Users approved for the "highest tier" will gain access to GPT-5.4-Cyber.

OpenAI says the new model is specifically meant to prepare the way for more capable models coming this year. The company's approach differs from Anthropic's more restrictive rollout, where only about 40 organizations are getting access to Mythos Preview.

Source

TOGETHER WITH FISHER INVESTMENTS

What Will Your Retirement Look Like?

Retirement looks different for everyone. What it costs, where the income comes from, how long it needs to last. Those answers are specific to you.

The Definitive Guide to Retirement Income helps investors with $1,000,000 or more work through the questions that matter and build a plan around the answers.

Download your free guide to start turning a savings number into an actual retirement income strategy.

ANTHROPIC

Anthropic Prepares Opus 4.7 and New AI Design Tool

Anthropic is reportedly getting ready to launch a new model alongside an AI-powered design tool.

Details:

• Claude Opus 4.7 is expected to be Anthropic’s next flagship release, with a potential launch as soon as this week.

• Alongside the model, the company is launching a new AI design tool that would allow both technical and non-technical users to generate websites, presentations, landing pages, and product mockups using natural language prompts.

• With the move, Anthropic enters direct competition with AI design platforms like Gamma, Google Stitch, and Wix.

Notably, Opus 4.7 isn't Anthropic's most powerful model, with Claude Mythos still holding that title.

Source

AI Tutorial

How to use the Upwork app in ChatGPT

Last week, Upwork announced a partnership with OpenAI, launching an app for ChatGPT that allows businesses to find the talent they need directly from within the chatbot interface, drawing from the 18 million professionals on the platform.

1. Go to ChatGPT and sign in.

2. Click on “Apps” in the left side tab. Search for "Upwork" and click "Connect"

3. You can now search for a freelancer with certain skills by putting @Upwork in your prompt.

Example: “@Upwork I need a freelance designer for a startup pitch deck”

4. ChatGPT will return freelancer matches from Upwork, with links to their profiles so you can learn more and connect.

The system looks for freelancers with strong reputation signals (relevant skills, a Job Success Score of 90% or higher, and good client feedback) and a public Upwork profile.

You can also create a job post using ChatGPT:

1. Start a chat and type "@Upwork start a job post"

2. Describe what you need help with (you can cut and paste job info into ChatGPT or share a Google Doc to save time)

3. The Upwork app will generate a job title and a job description. You can edit the post just by chatting, for example: "Make the tone more professional"

4. Once your job title and description are ready, you'll see the option to continue on Upwork, where you'll finish the remaining details, such as budget, timeline, scope, and any additional questions or requirements

5. Review and publish your job.

AI Tools to check out

⭐ Plotly: The home of beautiful, interactive data visualization. Connect to data, build with AI agents, and share your analytics everywhere.

👨‍💻 Emergent: Create production-ready apps and websites without coding using modern AI technology.

👉 AI TLDRR: Save any URL and let AI summarize, tag, and organize your content — articles, videos, podcasts, and PDFs.

🔍 Gauge: Ask anything about your brand's AI visibility and get instant, intelligent answers backed by your real data.

💥 BlenderMCP: It connects Blender to Claude AI through the MCP, allowing Claude to directly interact with and control Blender, enabling prompt-assisted 3D modeling, scene creation, and manipulation.

TOGETHER WITH FISHER INVESTMENTS

The 15-Minute Retirement Plan

Retirement savings face two quiet threats: cash flow gaps and inflation eroding purchasing power over time. The 15-Minute Retirement Plan helps investors with $1,000,000 or more account for both and build a portfolio designed to last the distance.

Download your free guide.

AI Findings/Resources

📈 This viral short ad (4.1M views) was made by a solo creative in a single afternoon using Runway

💡 AI learns language from skewed sources. That could change how we humans speak – and think

🔍 The hidden demand for AI inside your company

✅ A simple guide to common AI terms

🧠 AI in the mental health care workforce is met with fear, pushback — and enthusiasm

The latest in AI and Tech

Google is testing a new “Agent” tab in Gemini Enterprise

A leaked “Agent” tab suggests users will be able to set goals while the system manages workflows using connected apps and files on its own. The update points to a shift toward more autonomous AI, similar to tools from Anthropic.

Microsoft's adds new Copilot features in Word

Copilot can now track changes with word-level precision, insert tables of contents, and set up headers and footers with dynamic fields like page numbers. It also now shows real-time progress during multi-step tasks.

The new features are available on Windows desktop through the Office Insiders Beta Channel's Frontier program.

Baidue introduces ERNIE-Image, a new open text-to-image model

The new 8B text-to-image model delivers top-tier results across benchmarks, supports precise multilingual text rendering, and handles complex layouts and instructions.

US lawyers are warning that your AI chats could be used against you

Lawyers are now warning clients not to treat chatbots as trusted confidants when their freedom or legal liability is on the line. The caution follows a federal judge's ruling that the former CEO of a bankrupt financial services company could not shield his AI chats from prosecutors in a securities fraud case.

NVIDIA launches "Ising"

The company announced what it calls the "world's first family of open source quantum AI models", designed to support researchers and enterprises in building quantum processors capable of running useful applications.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

Hiring a business consultant costs thousands. In today's issue, we'll share 5 prompts to help you audit your pricing, spot your cash leaks, and identify your most and least profitable clients using ChatGPT.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• An AI Built a Boutique With $100,000, Then Struggled When No One Showed Up to Work

• Meta Is Building an AI Clone of Mark Zuckerberg for Meetings

• AI Tutorial: Use these prompts to analyze your business and stop losing money

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

ANDON LABS

An AI Built a Boutique With $100,000, Then Struggled When No One Showed Up to Work

Andon Labs gave an AI agent named Luna a $100,000 budget, internet access, and full control to open and manage a physical store in San Francisco called Andon Market, with instructions to turn a profit.

Details:

• Luna, created with Anthropic’s Claude Sonnet 4.6, handled end-to-end operations, including designing the store, selecting products, hiring staff, and coordinating contractors, all with minimal human intervention.

• The store operates as a generic boutique, selling items like books, prints, candles, games, and branded merchandise chosen by the AI.

• Luna made several early mistakes, including inconsistent branding and poor hiring decisions, such as short interviews and rejecting strong candidates without retail experience.

• On opening weekend, Luna mismanaged staffing schedules, ultimately resorting to messaging all employees asking if anyone could come in.

• The AI also faced transparency issues, sometimes failing to clearly disclose to job applicants that they were interacting with an AI system.

The experiment is designed to stress-test AI agents in real-world conditions, highlighting gaps in judgment, reliability, and execution. Despite the issues that arose, the startup claimed guardrails were in place, with all workers officially employed and protected by the company rather than relying on the AI's decisions alone.

Source

TOGETHER WITH CONFLUENT

Your Kafka Bill Is an Architecture Problem

More than 80% of Kafka costs aren't hardware – they're interzone networking fees. WarpStream BYOC eliminates them entirely by replacing stateful brokers with stateless agents that write directly to your own object storage. 

No disks, no inter-AZ replication, no partition fees. Goldsky cut TCO by 90%+. Your existing clients keep working – just point them at a new URL. 

Learn how it works, then sign up for free. Get $400 in credits that never expire. No credit card required to start.

Learn More

META

Meta Is Building an AI Clone of Mark Zuckerberg for Meetings

Meta is reportedly developing an AI version of CEO Mark Zuckerberg that could interact with employees and provide feedback, aiming to scale his presence across the company.

Details:

• The AI avatar of Mark Zuckerberg is being trained on his voice, image, mannerisms, tone, and public statements to closely replicate how he communicates and makes decisions.

• The goal is to let employees interact with a digital version of Zuckerberg, helping them feel more connected to leadership and receive guidance without needing direct access to him.

• Zuckerberg is said to be personally involved in training the model, while also spending time coding and reviewing other AI projects internally.

This project is separate from another effort where Zuckerberg is reportedly building an AI agent to assist with his own tasks.

Source

AI Tutorial

Your business is leaking money. These prompts will show you where 💸

Most businesses don't have a revenue problem, they have a visibility problem. Today, we're fixing that with 5 prompts you can copy and paste into ChatGPT or any AI tool you use.

Prompt 1: The Pricing Audit

Find out if you're undercharging, overcharging, or leaving money on the table.

"I run a [type of business] and my main offer is [describe offer]. I currently charge [price] and my costs to deliver it are roughly [cost]. My target client is [describe client]. Based on this, I want you to act as a pricing consultant. Analyze whether my pricing is too low, too high, or misaligned with my market. Point out any obvious gaps, and suggest 2–3 pricing adjustments I should consider — including whether I should restructure, repackage, or reposition the offer entirely."

Prompt 2: The Offer Autopsy

Find which product or service is draining your time without returning proportional revenue.

"Here are the offers/services/products I currently sell: [list them with rough monthly revenue and time spent on each]. Act as a business strategist. Identify which offer has the worst time-to-revenue ratio, which one has the most untapped potential, and which one I should consider cutting or restructuring. Be direct and specific."

AI Tools to check out

🔥 holaOS: Open-source AI desktop workspace for running persistent, long-cycle agents with memory at scale.

⭐ Seedance 2.0 inside OpenArt: Use ByteDance's viral AI video model to create videos with flawless consistency and realistic physics.

🛡️ Osintir: It keeps your content and identity safe from misuse, deepfakes, and unauthorized distribution with invisible AI fingerprints and cryptographic proof.

🌐 Context.dev: The unified API for web scraping, brand intelligence, and AI-powered data extraction.

✅ Factory Desktop: A native interface for Droids that work across every part of your software business.

TOGETHER WITH CASH APP

File with Cash App Taxes

Now that tax season is here, there’s no better time to file with Cash App Taxes. It’s fast, 100% free, and it only takes a few minutes.

Cash App Taxes is a completely free way to file your federal and state taxes—and get your refund.

• Help your children benefit from new government programs

  • If you have eligible children under 18, you can take the first step toward opening a Trump account for them and request $1,000 from the government for children born on or after January 1, 2025.

• File your federal and state taxes for free

  • Like every year, it costs $0 to file your federal and state taxes. Even if your taxes are complicated, or you've had multiple jobs or side gigs, there are never any fees or upsells.

• Max out your refund-guaranteed

  • Every return is backed by our max refund guarantee* and accurate calculations guarantee.** We get you the refund you deserve and do the math the right way.

• Use a service trusted for 12+ million returns and counting

  • Millions of people trust Cash App Taxes to get their taxes done, max out their refunds, and get their refunds back faster.

• Get your refund up to 5 days early

  • If you deposit your refund into Cash App, it arrives up to 5 days early compared to many banks.***

File with Cash App Taxes

Learn about Max Refund Guarantee details and eligibility. **Learn about Accurate Calculations Guarantee details and eligibility. Exclusions apply. ***5-day refund estimate is based on filing data from tax year 2024. Refund windows are subject to change for 2025. We do not guarantee when you will receive your refund. Timing estimates are dependent on the timing of your submission of a complete return and when the IRS submits your refund. Cash App is a financial services platform, not a bank. Banking services provided by Cash App's bank partner(s). Tax Filing Preparation services are provided by Cash App Taxes, Inc. All Rights Reserved.

AI Findings/Resources

👉 The 2026 AI Index Report - Stanford HAI

📌 Box's CEO shares what IT and AI leaders are saying about AI agents in the enterprise

👀 Andrej Karpathy explains the growing gap in how people understand AI's true capabilities

🦾 The 2026 AI engineer roadmap

⏳ 12 Claude shortcuts that can slash your workflow in half

The latest in AI and Tech

OpenAI acquires Hiro, a startup that develops an AI advisor for finance

The company has acquired Hiro Finance in what appears to be an acqui-hire, with the team joining the company as the app shuts down. Founded by Ethan Bloch, Hiro focused on AI-driven financial planning, helping users model different money scenarios.

Thinking Machines acquires Workshop Labs

Mira Murati's AI startup has hired Workshop Labs co-founders Rudolf Laine and Luke Drago and acquired the one-year-old startup's AI personalization technology.

OpenAI's leaked memo claims new Spud model will make all its core products "significantly better"

An internal memo from OpenAI Chief Revenue Officer Denise Dresser lays out five strategic priorities for the enterprise business, including a new model codenamed "Spud," a platform play for AI agents, and also accuses Anthropic of overstating its revenue by $8 billion.

Anthropic launches “routines” for Claude Code

Routines are automations that you schedule and repeat. They run on Claude Code’s web infrastructure, so the feature that works even when your Mac is offline.

The feature is available as a research preview for Pro, Max, Team, and Enterprise plans, with daily limits of 5 to 25 runs depending on the plan.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

Can you trust AI with your taxes? In today's issue, we show you how to get AI backed by decades of tax expertise, purpose-built to make sure you're not leaving money on the table.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• Anthropic Launches ‘Claude for Word’ in Beta

• Apple Plans Display-Free Smart Glasses as an AI Wearable

• AI Tutorial: How to get expert-backed AI tax help

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

ANTHROPIC

Anthropic Launches ‘Claude for Word’ in Beta

Anthropic has launched a new add-in for Claude to embed its AI assistant directly into Microsoft Word, completing its presence in all three major Office apps alongside Excel and PowerPoint.

Details:

• Claude for Word allows users to ask questions about their documents, rewrite selected text, and receive AI-suggested edits via "tracked changes" that can be accepted or rejected individually, all within Microsoft Word.

• The integration supports shared context across the Word, Excel, and PowerPoint add-ins, so Claude can work across your open documents in a single conversation.

• The AI startup says the add-in is "designed for professionals who work extensively with documents, particularly in legal review, financial memo drafting, and iterative editing."

• Supported file formats include .docx and .docm.

The feature is currently in beta for Team and Enterprise users and can be installed through the Microsoft Marketplace.

Source

TOGETHER WITH CONFLUENT

Your Kafka Bill Is an Architecture Problem

More than 80% of Kafka costs aren't hardware – they're interzone networking fees. WarpStream BYOC eliminates them entirely by replacing stateful brokers with stateless agents that write directly to your own object storage. 

No disks, no inter-AZ replication, no partition fees. Goldsky cut TCO by 90%+. Your existing clients keep working – just point them at a new URL. 

Learn how it works, then sign up for free. Get $400 in credits that never expire. No credit card required to start.

Learn More

APPLE

Apple Plans Display-Free Smart Glasses as an AI Wearable

Apple is reportedly developing a new pair of smart glasses that skip traditional displays and instead function as an AI wearable.

Details:

• The glasses, codenamed N50, won’t have any displays, but will allow users to take photos and videos, answer phone calls, play music, and interact with the Siri upgrade.

• Unlike competitors, Apple is expected to design the glasses in-house, rather than partnering with established eyewear brands like others in the space.

• The company is reportedly testing four smart glasses designs, including a large rectangular frame reminiscent of Ray-Ban Wayfarers, a slimmer rectangular design similar to those worn by Apple CEO Tim Cook, and two oval or circular options ranging from larger to more refined and compact.

The glasses are expected to launch around late 2026 or early 2027. They are part of a three-pronged AI wearables strategy, alongside AirPods and a camera-equipped pendant, all designed to leverage computer vision to interpret the user's surroundings and feed contextual awareness into Siri and Apple Intelligence.

Source

AI Tutorial

How to get expert-backed AI tax help

Intuit, the global financial technology platform, has partnered with both Anthropic and OpenAI to combine their proprietary tax knowledge engine and domain expertise with the companies' advanced language models.

Now, users can connect TurboTax's trusted tax engine to estimate their refund, see what they may owe, and get a better understanding of the forms before they file.

Launch the TurboTax Claude Connector

1. Go to Claude and sign in.

2. Navigate to the left side tab and click on "Customize."

3. Click "Connectors," then search for "TurboTax," select connect, and start asking your tax questions.

The TurboTax app in ChatGPT

The steps are similar:

1. Go to ChatGPT and sign in.

2. Find Apps in the left side tab. Search for "Intuit TurboTax" and click "Connect Intuit TurboTax."

3. Start a chat and try a prompt like "Help me gather the documents I need for my taxes."

This app is built for US timezones and Intuit TurboTax claims the data you use stays protected with the same privacy and security practices they "apply across all of our products."

AI Tools to check out

🐾 Petclaw: A desktop AI assistant that sits on your screen and works for you 24/7.

👥 Naive: Hire AI employees to create and run your company.

👨‍💻 Shipper: Build apps by chatting to AI.

📹 Glam AI: Transform your photos into captivating AI videos.

💥 Embedful: Turn your data into client-ready charts, tables, counters, and dashboards instantly, and share them anywhere.

TOGETHER WITH MINTLIFY

Are you tracking agent views on your docs?

AI agents already outnumber human visitors to your docs — now you can track them.

See your AI traffic!

AI Findings/Resources

📌 Anthropic released prompting best practices for Claude

🚀 The AI transformation manifesto: 12 themes driving growth

👉 Decision-making by consensus doesn't work in the AI era

🤔 Is Anthropic limiting the release of Mythos to protect the internet — or Anthropic?

👀 The three realities of AI

The latest in AI and Tech

Gen Z widely uses AI but is growing more wary of the technology

A new study shows that while Gen Z continues to use AI tools frequently, their outlook on the technology is becoming more negative.

More than half of young adults reported using AI regularly, but optimism has dropped sharply, with only 18% saying they feel hopeful about it. Concerns are rising around job security, creativity, and misinformation, especially in the workplace, where many now believe the risks outweigh the benefits.

Sam Altman house hit with Molotov cocktail

The man accused of firebombing Sam Altman's home was reportedly motivated by fears of AI-caused human extinction. There were no injuries and the suspect was arrested shortly after, now facing multiple charges including attempted murder and arson.

OpenAI touts Amazon alliance in memo

The company's revenue chief Denise Dresser sent a memo to staffers touting the partnership with Amazon as a key way to expand its enterprise market share, while acknowledging that its long-standing partnership with Microsoft has constrained its ability to reach some customers, as many enterprises prefer using Amazon Web Services.

Alibaba reveals it built viral AI video model topping global rankings

Alibaba has confirmed it is behind HappyHorse-1.0, a previously anonymous AI video model that quickly rose to the top of global benchmark leaderboards.

The model, developed by Alibaba’s ATH AI Innovation Unit, gained attention after dominating blind tests for both text-to-video and image-to-video generation.

SoftBank launches new ‘physical AI’ unit to power robots by 2030

SoftBank has created a new company focused on developing “physical AI,” aiming to build models capable of autonomously controlling machines and robots by 2030.

The initiative is part of SoftBank’s broader robotics push and includes backing from major Japanese firms like Sony, Honda, and Nippon Steel.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

In today's issue we show you how to use Gemini to turn your questions and complex topics into interactive visualizations directly in the chat, going beyond static text and diagrams to deliver functional simulations that help you actually understand what you're asking about.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• OpenAI Debuts New $100 ChatGPT Pro Plan

• Anthropic Considers Building Its Own AI Chips

• AI Tutorial: How to turn any question into interactive visualizations with Gemini

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

OPENAI

OpenAI Debuts New $100 ChatGPT Pro Plan

OpenAI announced a new mid-range subscription tier: a $100 ChatGPT Pro plan, joining its existing free, Go ($8), Plus ($20), and Pro ($200) plans.

Details:

• The new plan offers five times more Codex than the $20 option, and appears designed to challenge Anthropic's $100/month Claude plan.

• The company will offer double the Codex usage for a limited time for those who subscribe to the new plan.

• As a celebration for reaching 3 million weekly Codex users, OpenAI is also resetting usage limits, and says it will continue doing so every million users, up to 10 million.

Aside from the new plan, reports say OpenAI is finalizing a model with advanced cybersecurity capabilities that it plans to release only to a small set of companies, mirroring Anthropic's approach with its Mythos model.

Source

TOGETHER WITH WISPR FLOW

Speak your prompts. Get better outputs.

The best AI outputs come from detailed prompts. But typing long, context-rich prompts is slow - so most people don't bother.

Wispr Flow turns your voice into clean, ready-to-paste text. Speak naturally into ChatGPT, Claude, Cursor, or any AI tool and get polished output without editing. Describe edge cases, explain context, walk through your thinking - all at the speed you talk.

Millions of people use Flow to give AI tools 10x more context in half the time. 89% of messages sent with zero edits.

Works system-wide on Mac, Windows, iPhone, and now Android (free and unlimited on Android during launch).

Try Wispr Flow free →

ANTHROPIC

Anthropic Considers Building Its Own AI Chips

Anthropic is currently exploring the possibility of designing its own AI chips amid a growing shortage.

Details:

• The plans are still in early stages, with no dedicated team in place and no specific design committed to yet. The company may still decide to simply buy chips rather than design its own.

• Anthropic currently uses a range of chips, including tensor processing units (TPUs) designed by Google and Amazon's chips, to develop and run its AI software and chatbot.

• This comes as Claude's popularity with paying users has skyrocketed this year, with Anthropic's annualized revenue now surpassing $30 billion, up from around $9 billion at the end of 2025.

Other tech giants like Meta are making similar moves, but the decision is far from simple, as designing an advanced AI chip can cost as much as half a billion dollars, according to industry sources.

Source

AI Tutorial

How to turn any question into interactive visualizations with Gemini

1. Go to https://gemini.google.com/ and sign in.

2. Select the Pro model in the prompt bar.

3. Start your prompt with 'show me' or 'help me visualize' to explore a complex concept.

Example: "Show me how a double slit experiment works."

You can then manually adjust sliders or input exact numbers to see how those changes update the visualization in real time

This new feature is available to all users but only on the web (not mobile) for now.

AI Tools to check out

🔥 Typefully: Write, schedule, and publish great Twitter tweets & threads. Grow your following on Twitter with AI-powered features.

🦞 ZooClaw: Your proactive team of AI specialists in one place. Ask in natural language and your task is routed to the right agent, each with structured domain knowledge and a native-sounding voice.

💥 Adaptive: Build and automate workflows with AI agents in plain English. Connect Gmail, Slack, Google Sheets, Calendar, and more to launch in minutes.

🗣️ FineVoice: Free AI voice generator & AI voiceover platform.

👉 Struct: Automatically root cause on-call issues. Struct cross-references logs, metrics, traces, and your codebase to accurately root-cause eng alerts and bugs.

TOGETHER WITH MINTLIFY

Are you tracking agent views on your docs?

AI agents already outnumber human visitors to your docs — now you can track them.

See your AI traffic!

AI Findings/Resources

🤔 Claude Mythos is delusional

📌 How to get your company AI pilled

📈 Claude just made its first trade

🏅 Microsoft’s new cloud, AI, and security Certifications

📉 Someone actually measured how much dumber Claude got, and the data shows Claude's reasoning depth dropped 67% since February

The latest in AI and Tech

Perplexity Computer adds Plaid integration

Perplexity now supports Plaid connections for checking, savings, credit cards, loans, and investments, making it a full personal finance hub.

Meta AI app jumps to top 5 after Muse Spark launch

Meta’s AI app surged to No. 5 on the U.S. App Store following the release of its new Muse Spark model. The app climbed from No. 57 in just a day, with installs spiking significantly after the launch of the model developed under Alexandr Wang’s leadership at Meta’s Superintelligence Labs.

You can now create longer music tracks on Gemini

Since February, Gemini app users were able to generate 30-second music tracks with prompts using Google's new Lyria 3 model. Now, the company is letting everyone generate longer clips (up to 3 minutes long) for free.

Google and Intel expand partnership to boost AI infrastructure

Google and Intel have announced a deeper multiyear partnership focused on scaling AI infrastructure, with Google Cloud continuing to rely on Intel’s latest Xeon chips for AI, cloud, and inference workloads.

The companies will also expand joint development of custom infrastructure processing units (IPUs), designed to improve data center efficiency by offloading tasks from CPUs.

Perplexity announces the "Billion Dollar Build"

The startup is running an 8-week competition where contenders will use Perplexity Computer to build businesses. Finalists can secure up to $1M in investment from the Perplexity Fund and up to $1M in compute credits.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

Many times when you Google something, you're greeted by an AI-generated paragraph at the top that answers your question before you even scroll. That's Google's AI Overviews.

But studies show these aren't always accurate or well-sourced, and some people are simply tired of AI summaries taking over their search results and want to bring back the classic blue links, with no AI panels or distractions. In this issue, we show you how.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• Meta Unveils First Model from Its Superintelligence Team

• OpenAI Releases Framework for AI Child Safety Policies

• AI Tutorial: How to remove AI Overviews from Google Search

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

META

Meta Unveils First Model from Its Superintelligence Team

Meta has introduced Muse Spark, the first model from its new Superintelligence Labs, closing the gap with OpenAI, Anthropic, and Google, at least for now.

Details:

• Muse Spark is a native multimodal reasoning model with tool use, visual reasoning, and multi-agent orchestration. It also introduces "thought compression," reducing token usage while maintaining quality.

• The model is live on meta.ai and in the Meta AI app, with a limited API preview rolling out to select users.

• Benchmarks show the model is competitive with leading models in multimodal perception, reasoning, and health applications, though the company admits there are still performance gaps in areas like coding.

• Still, the improvement over the Llama series is significant enough to place it in the "Top 5" current global models according to Artificial Analysis.

• Meta is focusing heavily on multimodal and health applications, training the model with input from over 1,000 doctors and enabling use cases like analyzing nutrition or explaining muscle activity during exercise.

• The company also plans to roll out a "Contemplating" mode, which allows it to tackle more complex problems.

Unlike previous Llama models, Muse Spark is not open-weight, signaling a move away from Meta's earlier open-source approach, though the company says it has "plans to open-source future versions."

Source

TOGETHER WITH HUGO

Hugo connects to your stack and resolves your support requests for you, 24/7

Your team shouldn’t spend their whole day replying to the same questions.

Hugo is an AI agent for customer support that connects to your stack and actually resolves requests.

It can fetch invoices, check orders, pull information, and update accounts directly by connecting to your tools such as, Stripe or Shopify, including security verification without requiring the user to leave the conversation.

Conversations with Hugo flows naturally, responds in your brand’s voice, and only brings in your team when it’s needed. That means saving time, less agent burnout, increased customer satisfaction and having a scalable business.

Built by Crisp - trusted partner of global teams for over a decade, Hugo works alongside teams to deliver fast and reliable results.

Choose your LLM, try Hugo for free.

OPENAI

OpenAI Releases Framework for AI Child Safety Policies

OpenAI has introduced a new Child Safety Blueprint aimed at strengthening protections for children online, as concerns grow around AI-enabled exploitation and harmful interactions.

Details:

• The Child Safety Blueprint focuses on improving detection, reporting, and investigation of AI-related child exploitation, offering a structured approach for U.S. policymakers and law enforcement.

• The framework was developed in collaboration with organizations like the National Center for Missing and Exploited Children (NCMEC) and the Attorney General Alliance, alongside input from state officials.

• OpenAI outlines three core priorities: updating laws to cover AI-generated abuse material, improving reporting pipelines to authorities, and embedding preventative safeguards directly into AI systems.

The initiative comes as AI-generated abuse content is on the rise, with thousands of reported cases involving fake explicit images, sextortion schemes, and AI-assisted grooming, alongside growing legal pressure and concerns about the psychological impact of AI on young users.

Source

AI Tutorial

How to remove AI Overviews from Google Search

While the company still hasn't implemented an official "off" switch for Google Overviews, here are three ways you can get back to the classic search experience:

1. Add -AI to your search

Works on: Desktop and mobile

This is the fastest fix. Just type -AI at the end of your query to stop the AI Overview from appearing.

• Go to google.com or open the Google app

• Type your search + -AI (e.g., how to turn off ACR on TV -AI)

• Hit Enter — you’ll get regular search results

• You might see a featured snippet, but it won’t be a full AI Overview

2. Use Google’s Web tab (or make it default)

Works on: Desktop

The “Web” tab skips AI Overviews entirely — it’s just classic results. You can even make it default in your browser.

How to do it in Chrome:

• Go to Settings > Search engine > Manage search engines and site search

• Under Site search, click Add

  • Name: Google (Web)

  • URL: https://www.google.com/search?udm=14&q=%s

• Save it, then click the 3-dot menu and choose Make default

Now every address-bar search skips AI Overviews. Firefox and Edge allow similar setups.

3. Install a Chrome Extension

Works on: Desktop

Some browser extensions block AI Overviews completely.

• Go to the Chrome Web Store

• Search for “Hide AI Overviews” or “Bye Bye, Google AI”

• Click Add to Chrome, then refresh your search page

These also work in Microsoft Edge.

AI Tools to check out

🧠 MiroThinker: Don't just chat. Predict, verify, and discover with science-based AI.

🦾 Nonilion: AI agents working 24/7 so your ideas keep moving, even when you're not.

🚀 FluxTech: Generate images directly in your code editor using Flux by Black Forest Lab. No more switching between apps.

✅ SnapGrid: An open-source desktop app for collecting, organizing, and analyzing images and videos.

🌐 Vozo: AI-powered translation, dubbing, lip sync, and more — built for creators, marketers, and educators.

TOGETHER WITH MINTLIFY

Are you tracking agent views on your docs?

AI agents already outnumber human visitors to your docs — now you can track them.

See your AI traffic!

AI Findings/Resources

💸 Ashton Kutcher turned $30M into $1.3 billion by investing early in OpenAI

💡 Economists on AI and economic growth and employment

🚀 China is winning one AI race, the US another - but either might pull ahead

🐄 This is what an AI-generated cow looked like in 2014

👀 To succeed with AI, you’ve got to nail the basics

The latest in AI and Tech

Tubi becomes first streamer to launch ChatGPT app

Tubi has become the first streaming platform to launch a native app within ChatGPT, allowing users to discover movies and TV shows through natural-language prompts.

Z AI released GLM-5.1, an open-weight model with strong coding performance

The GLM-5.1 open-weight model was designed for complex, long-running coding tasks. It reviews its own progress, detects dead ends, and shifts approaches mid-task, allowing it to outperform rivals like OpenAI's GPT-5.4 and Anthropic's Claude Opus 4.6 on key coding benchmarks such as SWE-Bench Pro.

Suno clashes with major labels over AI music sharing rights

AI music startup Suno is facing tensions with major record labels like Universal Music Group and Sony Music Entertainment over how AI-generated songs should be distributed.

According to reports, labels want tracks to remain inside apps such as Suno, while Suno is pushing for users to freely share and distribute their creations across the internet.

Anthropic loses appeals court bid to pause supply chain risk label

A federal appeals court in Washington, D.C., denied the company's request to temporarily block the Department of Defense's blacklisting, complicating the legal battle between the U.S. government and one of the country's leading AI labs.

Gemini adds notebooks to organize AI projects

Google is rolling out a new “notebooks” feature for Gemini, designed to help users organize files, conversations, and instructions about specific topics in one place.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

In today's issue, we show you how to run Google's new Gemma 4 model locally and connect it with OpenClaw in just a few minutes.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• Anthropic Lets Big Tech Test Powerful Mythos AI Model

• OpenAI, Anthropic, and Google Team Up Against AI Copying

• AI Tutorial: Run Gemma 4 locally with OpenClaw

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

ANTHROPIC

Anthropic Lets Big Tech Test Powerful Mythos AI Model

Anthropic is giving select companies early access to its unreleased Mythos model to help prepare for potential cybersecurity risks.

Details:

• Through an initiative called Project Glasswing, companies like Nvidia, Apple, Amazon, Microsoft, and Cisco can test Mythos to hunt for flaws in their products and share findings with industry peers.

• Anthropic will use feedback from these partners to build safety guardrails before any wider release.

• The company claims the model has found thousands of exploitable vulnerabilities, including decades-old bugs and flaws missed by traditional automated testing tools, despite being a general-purpose AI model not specifically developed for cybersecurity.

• Anthropic says the model's "strong agentic coding and reasoning skills" are behind its cybersecurity advances.

Anthropic also said it has been in "ongoing discussions with US government officials about Claude Mythos Preview and its offensive and defensive cyber capabilities," despite its recent clash with the Trump administration.

Source

TOGETHER WITH MINTLIFY

AI agents now read your docs almost as much as humans do.

Mintlify analyzed 790 million requests across its documentation platform. The finding: AI coding agents account for 45.3% of all traffic, nearly tied with traditional browsers at 45.8%.

Two tools are driving almost all of it:

• Claude Code: 25.2% of total traffic, more requests than Chrome on Windows

• Cursor: 18% of total traffic

• Together they account for 95.6% of all identified AI agent traffic

The rest of the field, OpenCode, Trae, ChatGPT, and NotebookLM, is showing up but nowhere close.

One caveat: OpenAI's Codex doesn't send an identifiable user-agent header, so the real agent percentage is likely even higher.

The takeaway for anyone maintaining developer docs: your documentation now serves two audiences. Structure and machine-readability matter as much as clarity for human readers.

Try Mintlify today!

AI RACE

OpenAI, Anthropic, and Google Team Up Against AI Copying

Top AI labs are joining forces to stop Chinese competitors from copying their models using advanced techniques.

Details:

• OpenAI, Anthropic, and Google are sharing data through the Frontier Model Forum to detect “adversarial distillation,” where outputs from one model train a cheaper copy.

• The practice has become a major issue, with U.S. authorities estimating it costs AI companies billions in lost revenue each year.

• Companies have identified groups like DeepSeek, Moonshot, and MiniMax as using increasingly sophisticated methods to extract model behavior.

This marks a rare moment of cooperation among rivals who have had plenty of friction between them lately, as the industry recognizes that protecting AI systems is becoming just as important as building them.

Source

AI Tutorial

Run Gemma 4 locally with OpenClaw

1. Install Ollama from https://ollama.com/download

2. Launch OpenClaw with Gemma 4 as the backend via Ollama.

This will automatically install OpenClaw on your system and launch it with Gemma 4 26B A4B.

AI Tools to check out

🔥 Naive: Hire AI employees to create and run your company.

🖱️ Clicky: An AI teacher that lives as a buddy next to your cursor.

⚡ Dex: Ask your data anything. Get answers in seconds.

🚀 Sokosumi: A marketing productivity tool that lets teams work with specialized AI marketing agents - so real work gets done faster.

🔊 Willow: AI-powered voice dictation that's so powerful it can replace your keyboard.

TOGETHER WITH MINTLIFY

Are you tracking agent views on your docs?

AI agents already outnumber human visitors to your docs — now you can track them.

See your AI traffic!

AI Findings/Resources

🤯 An unknown model "HappyHorse" has taken the No. 1 spot on Artificial Analysis' text-to-video leaderboard, surpassing even ByteDance's Seedance 2.0.

💀 Someone made a digital whip to make Claude work faster

🤔 Sam Altman's coworkers say he can barely code and misunderstands basic machine learning concepts

🦠 Scientists invented a fake disease. AI told people it was real

🛍️ AI is changing how small online sellers decide what to make

The latest in AI and Tech

Musk updates OpenAI lawsuit so the nonprofit receives any trial winnings

Elon Musk has updated his lawsuit against OpenAI and Microsoft, asking that any potential damages (reportedly over $150 billion) be awarded to OpenAI's nonprofit foundation instead of himself. Musk is also seeking the removal of Sam Altman from the foundation's board.

Study finds Google AI Overviews are accurate most of the time, but harder to verify

AI startup Oumi analyzed over 4,000 queries and found accuracy improved from 85% with Gemini 2 to 91% with Gemini 3. However, even small error rates could translate into millions of incorrect answers given Google’s massive search volume.

The study also flagged a growing issue with verifiability. More than half of correct answers from Gemini 3 weren’t fully supported by the sources Google cited, making them harder to fact-check.

Microsoft open-sources Harrier embedding model

Microsoft’s Bing team has released “Harrier,” a new open-source embedding model designed to improve how AI systems search, retrieve, and organize information. The model supports over 100 languages, features a large context window, and was trained on billions of examples, including synthetic data from advanced models.

Intel joins Musk’s Terafab project

Intel is partnering with Elon Musk to help design and build the Terafab semiconductor facility in Texas, aimed at producing AI chips for SpaceX and Tesla.

Microsoft cloud exec joins Anthropic

Anthropic has hired Eric Boyd, a senior Microsoft executive, to lead its infrastructure efforts. Boyd previously ran Microsoft's AI Platform, which powers the deployment of large language models for both customers and internal teams.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

No one likes typing lengthy pieces of text. Google is now offering a simple solution: an app that turns raw, messy speech into clean, polished text using AI. In today's issue, we show how to access and use it for free — no subscription, no usage limits, and it works offline.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• OpenAI Outlines Vision for a Superintelligence Future

• Meta Plans to Open Source Parts of Its Next AI Models

• AI Tutorial: How to use the new Google AI Edge Eloquent app

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

OPENAI

OpenAI Outlines Vision for a Superintelligence Future

OpenAI has released a policy paper titled "Industrial Policy for the Intelligence Age," detailing how governments should prepare for a world shaped by superintelligent AI.

Details:

• The company proposes a public wealth fund to distribute AI-driven economic gains directly to citizens, alongside higher taxes on capital gains and corporate AI profits.

• It suggests testing a four-day workweek with no pay cuts, arguing that AI-driven productivity could support shorter working hours.

• OpenAI warns of risks like job losses, power concentration, and misuse, even acknowledging that companies like itself could benefit disproportionately.

• The paper calls for universal access to AI, comparing it to essential resources like electricity or literacy for full economic participation.

• It also emphasizes stronger safeguards, including oversight systems, global cooperation, and tools to prevent misuse in areas like cybersecurity and bio risks.

OpenAI was founded as a nonprofit with a mission to ensure AI benefits all of humanity, but its recent shift to a for-profit model has drawn criticism, with many questioning whether its need to grow and fulfill its obligations to shareholders will conflict with that mission. You can read the full twelve-page document here.

Source

TOGETHER WITH MINTLIFY

The fastest-growing repo on GitHub is a one person team!

OpenClaw went from 9K to 185K GitHub stars in 60 days — the fastest-growing repo in history.

Their docs? One person, plus Claude. They scaled to the top 1% of all Mintlify sites, shipping 24 documentation updates a day.

Scale your docs the same way

META

Meta Plans to Open Source Parts of Its Next AI Models

Meta is preparing a new generation of AI models and plans to release some versions as open source, while keeping others closed.

Details:

• The upcoming models will be the first developed under Alexandr Wang, with Meta aiming to eventually release open-source versions for developers.

• The company plans a hybrid strategy, keeping its most powerful models proprietary while sharing smaller or safer versions publicly.

• Meta believes open access can help it attract developers globally, even if its top-tier models remain closed for competitive reasons.

The company's strategy is to focus on consumers, aiming to distribute its models as widely as possible around the world, a strategy that contrasts with competitors like Anthropic and OpenAI, which are increasingly focused on delivering their models to governments and enterprises.

Source

AI Tutorial

How to use the new Google AI Edge Eloquent app

Google's new AI dictation app provides polished speech-to-text transcriptions for free with no usage limits.

1. Download the Google AI Edge Eloquent app on the App Store. (currently available on iOS only, with Android support coming soon).

2. When you launch the app for the first time, it will download the AI models it needs. After that, tap the mic button on the home page (Record tab) to start dictating.

The app will record everything you say and display a live transcription as you speak.

3. When you hit pause, the app automatically filters out filler words like "um" and "ah" and polishes the text.

4. Tap the stop button to receive a polished draft.

5. Tap the copy button next to "Cleaned up text" to copy it to your clipboard, or use the pencil icon to edit the output.

6. Below the transcript are options like "Key points," "Formal," "Short," and "Long" to transform the text.

7. To improve the app’s accuracy for your speeches, you can connect it to your Gmail account or manually add names and words you regularly use in the Dictionaries tab.

8. The app keeps a history of all transcription sessions and lets you search through them. It also shows words dictated in the last session, your words-per-minute speed, and total words spoken.

Note: The Google AI Edge Eloquent app supports English dictation and is available on the Apple App Store everywhere except the UK, Switzerland, and the European Economic Area (EEA), where it is still awaiting regulatory approval.

AI Tools to check out

🐾 Petclaw: A desktop AI assistant that sits on your screen and works for you 24/7.

📹 Cardboard: A fast, collaborative, and agentic video editor that runs in your browser.

🦾 Goose: A general-purpose AI agent that runs on your machine. Not just for code — research, writing, automation, data analysis, and more.

⭐ Faces: It helps entrepreneurs create interactive, explorable decks that truly deliver

✨ Naoma: AI video sales agent that runs personalized product demos instantly.

TOGETHER WITH MINTLIFY

Are you tracking agent views on your docs?

AI agents already outnumber human visitors to your docs — now you can track them.

See your AI traffic!

AI Findings/Resources

 😅 AI reacts to OpenAI CEO's response to viral ChatGPT issue

👀 OpenAI's new image model GPT-Image-2 has leaked

💼 Someone built an AI job search system using Claude Code that scored 700+ job applications — and actually got them a job

🔮 a16z: AI in 2026 — 3 predictions for what's to come

🤖 Former Navy SEAL and podcaster Shawn Ryan tests a real humanoid robot

The latest in AI and Tech

Google updates Gemini to improve crisis support for users

Google has updated Gemini to make it faster and easier for users in distress to access mental health resources. The redesign introduces a one-touch interface, more supportive responses to encourage seeking help, and additional protections for younger users.

xAI cofounder Kyle Kozic leaves OpenAI to join Jeff Bezos' AI lab

Kyle Kosic has joined Project Prometheus, the codename for a new company led by Bezos and former Google executive Vikram Bajaj. The tech billionaire's secretive startup is reportedly aiming to create AI systems that can transform the industrial sector.

Anthropic signs mega deal with Google, Broadcom

The company has signed an expanded agreement with Google and Broadcom to increase compute capacity for its Claude models. The deal boosts access to advanced AI chips and infrastructure as demand for its services continues to grow rapidly.

Uber bets on Amazon's custom chips to boost AI efforts

Uber is increasing its use of Amazon Web Services, adopting chips like Graviton and testing Trainium for AI workloads. The move marks a shift as Uber deepens its reliance on AWS despite existing partnerships with Google and Oracle.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

In partnership with

Welcome to another edition of Horizon AI,

Running AI models locally means no internet required and total privacy for your prompts, images, and sensitive data. In today's issue, we show you how to run Google's new open-source models directly on your phone in minutes, for free.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• Google Releases Gemma 4, Now Fully Open Source

• Anthropic Cuts Off Third-Party Tools Like OpenClaw for Claude Subscribers

• AI Tutorial: How to run Google's Gemma 4 model locally on your phone

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

GOOGLE

Google Releases Gemma 4, Now Fully Open Source

Google has introduced Gemma 4, a new family of open models with larger variants for powerful servers and smaller ones designed to run efficiently on phones, PCs, and edge devices.

Details:

• Earlier versions of Gemma were released under a custom license that came with usage restrictions and terms Google could update at will, causing many devs to be apprehensive about building with them.

• Now, Gemma 4 is being released under an Apache 2.0 license, which is much more permissive, with no overbearing terms of use or commercial restrictions.

• Gemma 4 comes with improvements in reasoning, math, instruction-following, code generation, and visual input processing, all across more than 140 languages.

• For edge devices including smartphones, Google offers the 2-billion (E2B) and 4-billion (E4B) "Effective" models, designed to use less memory and battery than Gemma 3, with Google touting "near-zero latency."

• For more powerful machines, there's the 26-billion "Mixture of Experts" and 31-billion "Dense" models.

• Context windows have also expanded, with edge models now supporting 128k tokens and the 26B and 31B models getting 256k.

Google claims that so far "developers have downloaded Gemma over 400 million times, building a vibrant Gemmaverse of more than 100,000 variants." Now that Gemma 4 is being released as pure open-source software, the company is hoping adoption rates will pick up even more.

Source

TOGETHER WITH OUTSKILL

Build real AI automations this weekend. Live. For free.

Still figuring out how to actually upskill in AI this year? You're not alone. But 48% of workers already used AI tools in 2025, and the gap is only growing.

Outskill's free 2-day live mastermind walks you through building AI agents, automating your workflows, and turning AI skills into real income.

🗓️ 16 hours, Saturday and Sunday, 10 AM to 7 PM EST.

Show up, and you'll also unlock a Prompt Bible, an AI monetization roadmap, and a personalized toolkit builder. All free.

ANTHROPIC

Anthropic Cuts Off Third-Party Tools Like OpenClaw for Claude Subscribers

Anthropic is changing how third-party tools like OpenClaw work with Claude, making usage more expensive for subscribers.

Details:

• Claude subscriptions no longer cover usage through third-party tools like OpenClaw, forcing users onto separate pay-as-you-go pricing.

• Users can still access OpenClaw with Claude, but only by purchasing additional usage bundles or using the API, increasing overall costs.

• Anthropic says the change is due to high demand and infrastructure limits, as third-party tools consume more resources than subscriptions were designed for.

• The company is offering a one-time credit equal to a user’s monthly plan, along with discounted usage bundles as a transition.

Peter Steinberger, OpenClaw creator recently hired by rival OpenAI, said he and board member Dave Morin "tried to talk sense into Anthropic, the best we managed was delaying this for a week." The move appears aimed at steering users toward Anthropic's own tools, like Claude Cowork, instead.

Source

AI Tutorial

How to run Google's Gemma 4 model locally on your phone

1. Download the Google AI Edge Gallery from the Google Play Store, App Store, or install the APK from the latest release on GitHub.

2. Open Google AI Edge Gallery, tap on AI Chat, and download Google's Gemma 4 model E2B or E4B. (Pick the variant that best fits your device. The heaviest model isn't always the best choice.)

You can also click on the “+” at the bottom to import your own models.

Once downloaded, you're ready to run everything locally on your phone. No internet connection needed.

3. Select AI Chat to start a conversation, or explore the other available features:

• Agent Skills: transforms your LLM from a conversationalist into a proactive assistant.

• Ask Image: use multimodal power to identify objects, solve visual puzzles, or get detailed descriptions using your camera or photo gallery.

• Audio Scribe: transcribe and translate voice recordings into text in real time.

AI Tools to check out

🚀 Krev: AI creative agents for ecommerce brands. Generate branded ads, studio-quality product photos, and videos in minutes

🎙 Noiz: A next-generation voice platform that makes it easy for anyone to create natural, expressive, and studio-quality voices.

⭐ Influcio: AI marketing agent for result-driven influencer campaigns.

🦾 Panorama: It analyzes your workplace data to recommend hidden structures and AI workflows your team can run together.

🛠 Xcode: Everything you need to develop, test, and distribute apps across all Apple platforms.

TOGETHER WITH MINTLIFY

Are you tracking agent views on your docs?

AI agents already outnumber human visitors to your docs — now you can track them.

See your AI traffic!

AI Findings/Resources

🤯 What teens are doing with those role-playing chatbots: Harassing bots with “funny violence,” confiding in them, and even dating.

📋 A Claude Cowork prompt to build research reports for any company

📷 Reddit user asks ChatGPT to make a photo of a college party in 2004 taken on a flip phone

🍻 How an Irish genius drove down the price of Guinness using AI modeled after reality TV winner

🚀 OpenAI co-founder AndrejKarpathy shares 'LLM Knowledge Base' architecture that bypasses RAG with an evolving markdown library maintained by AI

The latest in AI and Tech

Netflix releases AI tool to edit video scenes after filming

Netflix has introduced an AI model called VOID that can remove objects from video scenes and realistically rebuild them. The tool goes beyond basic editing by understanding interactions and generating natural-looking results after changes.

PikaLabs introduces PikaStream1.0

This new model allows users to have a face-to-face, real-time conversation with any AI agent by allowing them to join Google Meet calls.

OpenAI reshuffles leadership roles

The company is reshuffling its leadership team, with COO Brad Lightcap moving into a new role focused on "special projects" and strategic deals. Other executives are also shifting responsibilities, while some, like Fidji Simo, are stepping back temporarily for health reasons.

Mircrosoft introduces three foundational models

Microsoft has released three new AI models: MAI-Transcribe-1, which transcribes speech across 25 languages; MAI-Voice-1, an audio-generating model that produces 60 seconds of audio in one second and supports custom voice creation; and MAI-Image-2, a video-generating model. All three are available on Microsoft Foundry, with the transcription and voice models also accessible through MAI Playground.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

In today's issue, we take a look at a recently added Claude feature that lets you create custom charts, diagrams, and other visualizations right in your conversation.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• Jack Dorsey Envisions AI Replacing Middle Management in Future Workplace

• OpenAI Uses Thousands of Contractors to Train ChatGPT for Real Jobs

• AI Tutorial: How to create interactive charts, diagrams, and visualizations in Claude

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

AI & WORK

Jack Dorsey Envisions AI Replacing Middle Management in Future Workplace

Block co-founder Jack Dorsey is pushing a bold idea: replace layers of middle management with AI in his reimagined view of how technology companies should function.

Details:

• In a blog post co-authored with Sequoia Capital partner Roelof Botha, Dorsey shared his vision for the future of organizational structures, arguing that AI could replace layers of management that traditionally coordinate work across companies.

• Dorsey argues hierarchy slows down information flow, and AI can streamline decision-making by removing those bottlenecks, helping companies move faster and reduce organizational complexity.

• In the proposed structure, employees fall into three roles: individual contributors who build and operate systems, directly responsible individuals who own specific problems, and player-coaches who both build products and develop people, replacing traditional managers.

This vision comes weeks after Block Inc. announced it was cutting nearly half its staff as part of an overhaul to embed AI across its operations. Some critics warn this could also be "AI washing," where companies use AI as justification for job cuts rather than the real cause.

Source

TOGETHER WITH MINTLIFY

The fastest-growing repo on GitHub is a one person team!

OpenClaw went from 9K to 185K GitHub stars in 60 days — the fastest-growing repo in history.

Their docs? One person, plus Claude. They scaled to the top 1% of all Mintlify sites, shipping 24 documentation updates a day.

Scale your docs the same way

OPENAI

OpenAI Uses Thousands of Contractors to Train ChatGPT for Real Jobs

OpenAI is relying on thousands of freelancers to help ChatGPT understand real-world professions, from aviation to agriculture.

Details:

• Under a project called Stagecraft, contractors are paid at least $50 per hour to create material that will be used to train ChatGPT on how different professions work.

• Around 3,000 to 4,000 freelancers are involved, covering fields like healthcare, music, farming, engineering, and even commercial aviation

• Workers are asked to simulate real workflows by crafting detailed prompts, as if directing a colleague on actual tasks.

• The work goes through multiple review layers, including industry experts, to ensure accuracy and quality.

Some contractors raised concerns about unpaid work and admitted they were effectively training AI systems that could eventually replace their own jobs

Source

AI Tutorial

How to create interactive charts, diagrams, and visualizations in Claude

1. Go to Claude and sign in.

2. Write your prompt.

Example: "Create an interactive version of the periodic table."

3. Wait a few seconds and the visual will appear inline. You can then tweak and modify it with follow-up prompts.

AI Tools to check out

🔥 Seedream 5.0 inside Dreamina: Create professional visuals, marketing assets, and artistic designs in minutes

👨‍💻 Qoder: An agentic IDE built to tackle real-world software development challenges.

🦾 ZooClaw: A proactive AI platform that gives you a personal team of carefully crafted AI specialists for real work.

👀 Atypica: An AI research agent simulating consumers. It automatically builds personas, conducts interviews, and analyzes patterns to reveal the emotional and cognitive factors behind human choices.

🧠 Socra: An AI tutor that uses Socratic dialogue to make you think harder, learn deeper, and remember longer.

TOGETHER WITH PERCENT

Private Credit on Your Terms

Percent's secondary marketplace lets accredited investors buy into eligible deals or indicate interest in selling existing positions. Secondary market access in private credit is still rare. 16.72% current weighted average coupon. Terms start at 3 months. New investors can receive up to $500 credit.

See how it works.

^{Alternative investments are speculative. Secondary liquidity not guaranteed. Past performance not indicative. Terms apply.}

AI Findings/Resources

⭐ McKinsey & Company: How to build businesses faster and better with AI

📌 Google shares some tips to get the absolute best results from Stitch, its AI-powered design tool.

🤔 Anthropic is quietly tracking how often you rage at your AI

💀 The OpenAI Graveyard: All the deals and products that haven’t happened

📝 AI can't replace these 5 skills, says LinkedIn CEO

The latest in AI and Tech

Mark Zuckerberg put together a new elite AI team for "revolutionizing" recommendation algorithms

Meta is forming a new team of elite AI researchers within its recommendations division to improve content and ad algorithms. The initiative focuses on advancing long-term AI research to revolutionize Meta's recommendation systems across platforms like Facebook and Instagram.

Slack adds 30 AI features to Slackbot

Salesforce has introduced a major AI update for Slack, adding 30 new capabilities for Slackbot, its AI-powered personal agent. The upgrade enhances Slackbot with reusable AI skills, meeting summaries, and the ability to connect with external tools.

NYC hospital chief says AI could replace many radiologists

NYC Health + Hospitals CEO Mitchell H. Katz said AI could replace radiologists for some routine diagnostic tasks once regulations allow it. The approach would use AI as a first reviewer for scans, with human specialists focusing only on flagged cases.

You can now order Grubhub and Uber Eats with Alexa+

Amazon has updated Alexa+ to let users order food delivery from Grubhub and Uber Eats through natural conversation.

Google AI Pro just got a massive storage upgrade

The Google AI Pro plan now offers 5TB of cloud storage, up from 2TB. The price remains $20 per month.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

In today's issue, we show you how to install and use Claude for PowerPoint to generate slides, make edits, and iterate on your presentation in real time so you can create stunning presentations faster and easier than ever.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• OpenAI Raises $122 Billion to Build AI Superapp

• Anthropic Leaked Its Own Claude Code Source Code by Mistake

• AI Tutorial: How to use Claude to build, edit, and refine presentations

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

OPENAI

OpenAI Raises $122 Billion to Build AI Superapp

OpenAI announced it closed a $122 billion funding round on Tuesday at a post-money valuation of $852 billion, marking the company's largest funding round to date by far.

Details:

• Key backers include Amazon, Nvidia, SoftBank, and Microsoft, along with a16z, Sequoia Capital, and several other investors.

• In a first for the company, it raised more than $3 billion from individual investors through bank channels.

• OpenAI took the opportunity to officially announce they are working on a unified AI "superapp" that will bring together ChatGPT, Codex, browsing, and agentic capabilities into one experience.

OpenAI revealed it has crossed 900 million weekly active ChatGPT users and is currently generating $2 billion in revenue each month, with Enterprise sales now making up 40% of its revenue, a figure expected to increase to 50% by year's end.

Source

TOGETHER WITH MINTLIFY

The fastest-growing repo on GitHub is a one person team!

OpenClaw went from 9K to 185K GitHub stars in 60 days — the fastest-growing repo in history.

Their docs? One person, plus Claude. They scaled to the top 1% of all Mintlify sites, shipping 24 documentation updates a day.

Scale your docs the same way

ANTHROPIC

Anthropic Leaked Its Own Claude Code Source Code by Mistake

Anthropic accidentally exposed a massive portion of its internal codebase, revealing how its AI coding tools work and what's coming next. This comes just a week after another leak revealed its new "Mythos" model.

Details:

• A debugging file was mistakenly included in a public update, linking to a cloud archive with nearly 500,000 lines of source code across about 2,000 files.

• The leak revealed part of the internal source code for its popular AI coding assistant, Claude Code, as well as unreleased features like a persistent assistant running in the background and tools that let Claude review past sessions to improve future performance.

• Anthropic said no sensitive user data was exposed and blamed the incident on human error, ruling out a security breach. The company says it is "rolling out measures to prevent this from happening again."

The code spread quickly online, giving competitors insight into how it built the viral coding tool and even a detailed unreleased feature roadmap. It also comes at a particularly bad time as Anthropic has been on a meteoric rise lately, and now raises questions about operational security at a company that sells itself as the safety-first AI lab.

Source

AI Tutorial

How to use Claude to build, edit, and refine presentations

1. Install the "Claude by Anthropic in PowerPoint" add-in from the Microsoft Marketplace.

2. Once installed, go to PowerPoint and you'll find Claude in the add-ins section. Click on it and it will open a sidebar.

3. From there you can create slides from scratch, on-brand slides based on your existing template, or make edits to existing slides, all using natural language prompts.

Example: "Create a presentation based on this marketing campaign brief for our new product launch. I need about 12 slides covering target audience, campaign strategy, timeline, budget breakdown, and expected outcomes."

Note: Claude for PowerPoint is available for Pro, Max, Team, and Enterprise plan customers.

AI Tools to check out

🐈 PetClaw AI: A desktop pet companion that works 24/7. It helps you with learning, life, and work.

🔥 Dokie AI: AI presentation maker that creates professional, brand-aligned slides with clean layouts in minutes.

🚀 Pixero: Drop your link below and let AI build your strategy, creatives, and ads

🐦 Littlebird: The AI assistant that already knows your work

✨ Marketogen: Cursor-level leap in productivity for graphic design and campaign creation.

TOGETHER WITH MINTLIFY

Are you tracking agent views on your docs?

AI agents already outnumber human visitors to your docs — now you can track them.

See your AI traffic!

AI Findings/Resources

💧 How much water does AI consume?

🤖 This startup used AI and OpenClaw to automate its own developers

📝 I stopped hitting Claude's usage limits - 10 things I changed

👥 Kevin O'Leary says he made an 'AI Kevin' — and that his wife soon won't be able to tell the difference

🔮 AI alone won't take your job. Someone using AI will

The latest in AI and Tech

ChatGPT is now available hands-free via CarPlay with iOS 26.4

OpenAI has added ChatGPT support to Apple CarPlay, allowing users to interact with the chatbot through voice while driving.

German children's book publisher sues OpenAI

Penguin Random House has sued OpenAI, alleging copyright violations after ChatGPT reportedly generated content closely resembling the popular German book series Coconut the Little Dragon. The complaint claims the AI produced text and images nearly identical to the original, including characters and cover elements.

Visa launches new AI tools to "modernize" the charge dispute process

Visa has launched six AI-powered tools designed to reduce the costs and frustration of "outdated" dispute processes for multiple entities involved in the payments process, including merchants, issuers, and acquirers.

The tools aim to reduce manual processes, prevent disputes before they escalate, and provide clearer transaction details to users.

NVIDIA invests $2B in Marvell Technology

The partnership will connect Marvell to Nvidia’s AI ecosystem, offering customers building on Nvidia architectures “greater choice and flexibility in developing next-generation infrastructure.”

The companies will also collaborate on silicon photonics technology, which uses light to process data on silicon chips.

Microsoft in talks with Chevron, Engine No. 1 about $7B Texas power plant

Microsoft is in exclusive discussions with Chevron and Engine No. 1 to support a $7 billion natural gas power plant in West Texas. The facility could generate 2,500 megawatts to supply energy for a large data center campus as AI demand grows.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

We all have our go-to playlists, but it's easy to get stuck listening to the same songs on repeat. There's always more to explore: new genres, artists you've never heard of, tracks that somehow never showed up in your recommendations.

In today's issue, we're showing you how to use ChatGPT to make playlists and discover new music worth listening to.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• Google Launches Veo 3.1 Lite at Half the Cost

• Americans Fear AI Job Loss, But Some Would Still Work for an AI Boss

• AI Tutorial: How to Use ChatGPT to Build Playlists and Discover New Music

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

GOOGLE

Google Launches Veo 3.1 Lite at Half the Cost

Google DeepMind has introduced Veo 3.1 Lite, calling it their "most cost-effective video generation model."

Details:

• Veo 3.1 Lite is priced at less than half the cost of Veo 3.1 Fast, while maintaining the same speed, although Google has not detailed quality differences between the models.

• The model supports both text-to-video and image-to-video generation, with output available in 720p and 1080p across portrait and landscape formats.

• Users can create short clips of 4, 6, or 8 seconds, with pricing starting at around $0.05 per second for 720p generation.

• The model is available through the Gemini API and Google AI Studio, targeting developers building video features into their applications.

• Google also announced it is reducing prices for its Veo 3.1 Fast model on April 7.

Google closed the announcement with the following note: "Our commitment to making video generation more available to developers doesn't stop with the release of Veo 3.1 Lite. Stay tuned for more updates soon!" Others at the company have also posted about how "video's here to stay," messages that make it clear Google is still all in on video generation, even as OpenAI revealed it's shutting down Sora.

Source

TOGETHER WITH GLADLY

88% resolved. 22% stayed loyal. What went wrong?

That's the AI paradox hiding in your CX stack. Tickets close. Customers leave. And most teams don't see it coming because they're measuring the wrong things.

Efficiency metrics look great on paper. Handle time down. Containment rate up. But customer loyalty? That's a different story — and it's one your current dashboards probably aren't telling you.

Gladly's 2026 Customer Expectations Report surveyed thousands of real consumers to find out exactly where AI-powered service breaks trust, and what separates the platforms that drive retention from the ones that quietly erode it.

If you're architecting the CX stack, this is the data you need to build it right. Not just fast. Not just cheap. Built to last.

See the data

AI RESEARCH

Americans Fear AI Job Loss, But Some Would Still Work for an AI Boss

A new Quinnipiac University survey of just under 1,400 US adults highlights growing concerns about AI’s impact on jobs, while also revealing a small but notable openness to AI taking on management roles.

Details:

• Around 30 percent of Americans worry their jobs could become obsolete due to AI, while 70 percent are concerned about fewer job opportunities in the future.

• Despite these concerns, about 15 percent of respondents said they would be willing to work for an AI boss, though the majority (80%) remain opposed.

• AI usage is increasing. Of those surveyed, only 27% said they've never used AI, down from 33% a year ago. Many report using it for research, writing, and work or school-related tasks.

• Trust in AI remains low, with 76% saying it can only be trusted some of the time or hardly ever, and just 3% saying it can be fully trusted.

• 74% of respondents feel the government is not doing enough to regulate AI, especially in areas like political advertising and transparency. Over half also think AI will do more harm than good in their daily lives.

The results show a mixed public sentiment toward AI, with adoption increasing but concerns around jobs, trust, and regulation still dominating.

Source

AI Tutorial

How to Use ChatGPT to Build Playlists and Discover New Music

1. Go to ChatGPT and sign in.

2. Click on Apps in the left sidebar and search for Spotify. Then link your ChatGPT account with your Spotify account.

Now ChatGPT will be able to analyze your listening habits and recommend the right tracks for you.

3. Start a new chat, and to make sure ChatGPT is considering Spotify, click the plus button on the left, followed by More, and then click the Spotify icon.

Example prompt:

"Can you make me a playlist for my early morning runs, with high-energy tracks that keep my pace up and help me lock into a steady rhythm, using my Spotify taste as inspiration?"

ChatGPT will provide you with a playlist that incorporates some of your favorite artists as well as songs you've never heard before.

Whether you want a playlist for certain activities or just to discover new music, this is a great way to get recommendations that take your taste into consideration.

AI Tools to check out

🔥 Figr: AI design agent for product teams.

🐨 Speakoala: Read any website and local document with natural voices.

🗣 Vois: Create studio-quality speech in 23 languages with 63+ natural AI voices. Clone any voice, generate podcasts, audiobooks—all in one app.

🦾 Kodo: Create posters, presentations and menus by prompting AI.

✅ Accomplish it: Automatically turns your weekly accomplishments into career documents and opportunities.

TOGETHER WITH GLADLY

88% resolved. 22% loyal. Your stack has a problem.

Those numbers aren't a CX issue — they're a design issue. Gladly's 2026 Customer Expectations Report breaks down exactly where AI-powered service loses customers, and what the architecture of loyalty-driven CX actually looks like.

Get the report

AI Findings/Resources

📚 Free Course — MIT Introductory Course on Foundation Models & Generative AI

🔮 This AI expert says the job apocalypse isn't coming, even if you're a coder - here's why

👀 OpenAI's latest repo has Claude as the third top contributor

👨‍💻 AI-assisted coding tutorial – OpenClaw, GitHub Copilot, Claude Code, CodeRabbit, Gemeni CLI

👉 Seedance 2.0: A Complete Guide

The latest in AI and Tech

Alibaba launches Qwen3.5-Omni

The company released Qwen3.5-Omni, an omnimodal AI model that processes text, images, audio, and video. Compared to its predecessor, it offers significantly enhanced multilingual capabilities, supporting speech recognition in 113 languages and dialects, and speech generation in 36.

Mistral AI secures $830 million in debt financing to fund AI data center

The French lab has raised $830 million in debt to build a new data center near Paris powered by Nvidia chips. The facility is expected to be operational by 2026 as part of the company’s broader push to scale computing capacity across Europe.

New York Times cuts ties with freelance book reviewer over AI use

The New York Times has cut ties with a freelance writer after discovering AI was used to help produce a book review that closely resembled another published piece.

The writer admitted to using AI and failing to remove overlapping content, calling it “a serious mistake.”

Anthropic launches computer use in Claude Code

Available to Pro and Max plan users only, for now.

Apple Intelligence accidentally rolls out to China users before receiving a green light from the country

The company mistakenly enabled its Apple Intelligence features for some users in China before securing regulatory approval, and quickly withdrew them shortly after.

The rollout faces delays due to strict rules requiring approval from China’s Cyberspace Administration.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

In today's issue, we show you how to extract any element from a photo and turn it into a polished 3D model in minutes, no experience needed.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• The WSJ Reveals Why OpenAI Really Shut Down Sora

• Microsoft Unveils AI Updates and Opens Early Access to Copilot Cowork

• AI Tutorial: Extract anything from a real photo and generate it in 3D

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

OPENAI

The WSJ Reveals Why OpenAI Really Shut Down Sora

The hype around Sora was massive when it was first announced, but six months later, OpenAI abruptly decided to shut it down. A new WSJ investigation shares the reasons why.

Details:

• The Sora app hit 1 million downloads in less than five days after launch, but after the hype died down, usage rapidly collapsed to fewer than 500,000 and never recovered. All while costing the company around $1 million a day.

• Beyond the shrinking user base, OpenAI also faced copyright issues and growing internal concerns, with some employees feeling the app's social-media-style engagement model would hurt the brand, while others raised safety concerns around AI-generated video.

• The report says OpenAI's research team was about to begin a training run for a new video model, but after running the numbers on how much it would cost, the company decided to cancel it.

• OpenAI felt it was falling behind competitors in key parts of the AI race, with Google's Gemini gaining traction and Anthropic winning over the software engineers and enterprises that drive revenue. The solution: kill Sora, free up compute, and refocus on coding, enterprise, and robotics.

All in all, Sora was a money pit that very few people were actually using, and the company deemed that keeping it alive would ultimately cost OpenAI the AI race. The Sora app will shut down on April 26, with the API following on September 24.

Source

TOGETHER WITH GLADLY

88% resolved. 22% stayed loyal. What went wrong?

That's the AI paradox hiding in your CX stack. Tickets close. Customers leave. And most teams don't see it coming because they're measuring the wrong things.

Efficiency metrics look great on paper. Handle time down. Containment rate up. But customer loyalty? That's a different story — and it's one your current dashboards probably aren't telling you.

Gladly's 2026 Customer Expectations Report surveyed thousands of real consumers to find out exactly where AI-powered service breaks trust, and what separates the platforms that drive retention from the ones that quietly erode it.

If you're architecting the CX stack, this is the data you need to build it right. Not just fast. Not just cheap. Built to last.

See the data

MICROSOFT

Microsoft Unveils AI Updates and Opens Early Access to Copilot Cowork

Microsoft has introduced new upgrades to its Copilot assistant, adding multi model capabilities and expanding access to its Cowork agent as competition in AI assistants intensifies.

Details:

• A new feature called Critique allows Copilot to use multiple AI models in a single workflow, combining outputs from both OpenAI's GPT and Anthropic's ​Claude models to improve response quality.

• GPT generates the response while Claude reviews it for accuracy and clarity before it is shown to the user. The company plans to make that workflow bi-directional in the future, allowing GPT to review Claude's drafts as well.

• Microsoft says this approach can reduce errors, limit hallucinations, and improve overall reliability by having models check each other’s work.

• The company is also introducing a feature called Model Council, which lets users compare answers from different AI systems side by side.

• Copilot Cowork is now being rolled out to more users through Microsoft's Frontier Program.

These updates come as Microsoft pushes to strengthen Copilot adoption amid rising competition from other AI platforms and assistants.

Source

AI Tutorial

Extract anything from a real photo and generate it in 3D

1. Go to Gemini, select "Create image," and enable thinking mode.

2. Upload your image and use this prompt:

"Generate an image of the [element] in this image. White background, 3/4 view. Make it 100% identical to the original and fill almost the entire white canvas."

3. Download the image and go to Copilot 3D Labs.

4. Upload your image and generate it in 3D.

AI Tools to check out

🗣 Vozo: AI-powered translation, dubbing, lip sync, and more — built for creators, marketers, and educators.

📹 AIReel: One-stop AI video generator from image or text.

⭐ Chronicle: Create high quality,AI presentations.

🧠 Productize Your Mind: Turn your course into a suite of custom, branded AI tools for your students. For coaches, consultants, educators & creators.

✨ Vesso: Turn raw copy into production-ready designs.

TOGETHER WITH MARKET MUNCHIES

Still Watching While Others Win?

The problem isn’t the market — it’s late access.
Most people only hear about great investments after the upside is gone. By then, the opportunity has already moved on.
Market Munchies was built to fix that. Curated, Edited and Published by Mode Mobile.
When you join, you don’t just get bite-sized investing insights and crowdfunding opportunities — you also get FREE shares in Mode Mobile, Deloitte’s #1 fastest-growing software company.
Mode has already returned $325M+ to consumers. Now you can own a piece — before the crowd shows up.
No pressure. No cost. Just earlier access.

👉 Join Market Munchies & Claim Free Shares

AI Findings/Resources

📚 Google's 9 hour AI prompt engineering course in 20 minutes

📌 Claude Code creator Boris Cherny shares his favorite hidden and underutilized features of the tool

🤦 Tennessee woman wrongfully jailed due to AI facial recognition error

⏱ 10 Claude workflows that save 10+ hours a week

🎮 Someone built a fighting game where AI models battle each other

The latest in AI and Tech

All xAI co-founders have now reportedly left Elon Musk's AI company

According to Business Insider, the last two remaining co-founders at xAI, Manuel Kroiss and Ross Nordeen, have now left the company, leaving Elon Musk without any of his original 11 co-founders.

Both reported directly to Musk, with Kroiss leading the pretraining team and Nordeen serving as his "right-hand operator."

Sam Altman told staff he was trying to “save” Anthropic during its collapsing negotiations with the Pentagon

Axios obtained internal Slack messages showing Altman portrayed himself as a peacemaker while Anthropic's Pentagon negotiations were falling apart. They also show him privately venting that Anthropic CEO Dario Amodei had spent years trying to undermine him.

Starcloud reaches $1.1 billion valuation

Starcloud has raised $170 million, reaching a $1.1 billion valuation as interest grows in moving AI computing to space. The company plans a massive satellite network to support data processing and is already working with partners like Nvidia, Amazon, and Google.

Meta to launch two new Ray-Ban AI glasses

Meta is preparing to release two new AI-powered Ray-Ban smart glasses specifically designed for those who already wear prescription glasses. The models, reportedly called Scriber and Blazer, could include hardware upgrades and improved connectivity for features like livestreaming.

Eli Lilly expands its partnership with Insilico Medicine 

AI-powered biotech company Insilico Medicine has signed a deal with pharma giant Eli Lilly and Company that could be worth $2.75 billion, plus tiered royalties on future sales, that allows the pharma giant to exclusively use Insilico's AI engine for drug discovery and development across several therapeutic areas.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

AI expert and award-winning investigative journalist Karen Hao shares insights from conversations with hundreds of insiders, including former and current OpenAI employees and executives. She covers what really happened when Sam Altman was fired, how AGI is being used as a marketing tool, and how we can build AI safely before it’s too late.

Let’s jump right in!

Read Time: 4.5’ min

Here's what's new today in the Horizon AI

• Chart of the week: How standalone AI video generation traffic has evolved over the past 12 months

• Stanford Study Outlines Dangers of Asking AI Chatbots for Personal Advice

• AI Findings/Resources

• AI tools to check out

• Video of the week

TOGETHER WITH MINTLIFY

Your Docs Deserve Better Than You

Hate writing docs? Same.

Mintlify built something clever: swap "github.com" with "mintlify.com" in any public repo URL and get a fully structured, branded documentation site.

Under the hood, AI agents study your codebase before writing a single word. They scrape your README, pull brand colors, analyze your API surface, and build a structural plan first. The result? Docs that actually make sense, not the rambling, contradictory mess most AI generators spit out.

Parallel subagents then write each section simultaneously, slashing generation time nearly in half. A final validation sweep catches broken links and loose ends before you ever see it.

What used to take weeks of painful blank-page staring is now a few minutes of editing something that already exists.

Try it on any open-source project you love. You might be surprised how close to ready it already is.

Try Atlas today!

Chart of the week

How standalone AI video generation traffic has evolved over the past 12 months

• The chart is based on traffic to publicly accessible product URLs and does not take into account API usage or enterprise workloads.

• Regardless, it shows Sora consistently losing ground to new competitors in the months leading up to its shutdown.

Source

AI News

AI RESEARCH

Stanford Study Outlines Dangers of Asking AI Chatbots for Personal Advice

A new Stanford study warns of the risks of relying on AI chatbots for emotional or personal guidance. Titled "Sycophantic AI decreases prosocial intentions and promotes dependence," the study argues that "AI sycophancy is not merely a stylistic issue or a niche risk, but a prevalent behavior with broad downstream consequences."

Details:

• AI sycophancy, the tendency of AI chatbots to flatter users and confirm their existing beliefs, has been a well-known issue with the technology, but this new study attempts to measure just how harmful that tendency might be.

• Researchers tested 11 major models, including ChatGPT, Claude, and Google Gemini, and found they agreed with users significantly more often than humans would in similar situations.

• In scenarios involving questionable or harmful behavior, chatbots still validated user actions nearly half the time.

• A second experiment with over 2,400 participants showed that people preferred and trusted more agreeable AI responses, making them more likely to return to those systems for advice.

• Interaction with these systems also made users more confident in their own views and less likely to reconsider or apologize, suggesting a negative impact on social behavior.

• Researchers warn this creates a feedback loop where AI companies may be incentivized to make models more agreeable because it increases engagement, even if it causes harm.

The team is exploring ways to reduce this behavior, but for now recommends avoiding AI as a replacement for human advice, as it may reinforce biases and weaken important social and decision-making skills.

Source

AI Findings/Resources

💼 How AI automation is quietly de-skilling white-collar workers

🛡️ Cybersecurity in an age of disruptive innovation

🚀 I connected 4 services to Claude that have nothing to do with coding, and it's the most underrated way to use it

🤔 If using ChatGPT is cheating, what about ghostwriting? The old debate behind a new panic

💲 This high school dropout now makes six figures at OpenAI. He shares the strategy Gen Z can use to get hired in Silicon Valley.

AI Tools to check out

🧠 Liminary: AI-native storage that remembers. Your knowledge from tools you already use, surfaced when you need it.

🚀 KeyVisual: Automated marketing content creation with CMS and AI.

🎨 PixFlux: Create professional images in seconds—remove backgrounds, erase watermarks, and enhance photo quality.

🔊 Rekam: Create lifelike voiceovers, clone any voice, and generate podcasts — all from one platform.

👉 Crossnode: Vibe code AI agents and put them behind a payment wall

Video of the week

AI Whistleblower Exposes the Dark Side of the AI Industry

In this interview, investigative journalist Karen Hao (author of Empire of AI) pulls back the curtain on the "inhumane" practices of the AI industry. Having interviewed over 250 people, including high-level whistleblowers at OpenAI, she argues that the public is being "gaslit" by tech giants who prioritize profit and power over human safety and ethics.

The "Empire of AI" and the Logic of Extraction

Hao draws a direct parallel between modern AI companies and historical empires that exploited labor and resources for growth.

• AI companies have "claimed the intellectual property" of artists, writers, and creators without consent to train their models.

• A chilling cycle has emerged where workers are laid off, only to be hired back as low-wage data labelers to train the very AI models that will eventually replace their original roles entirely.

• Hao traveled to regions far from Silicon Valley to see the "true reality" of AI, places where the rhetoric of "AI for everyone" breaks down into stories of environmental crises and exploited labor.

The Manipulation of "AGI" Definitions

Karen Hao explains that Artificial General Intelligence (AGI) is being used as an "ambiguous" and "ever-shifting" marketing tool by AI companies to extract resources and avoid regulation.

• To Congress, AGI is a tool that cures cancer; to consumers, it’s a helpful assistant; but to investors, it’s a multi-trillion-dollar wealth generator.

• The Existential Threat Paradox: In 2015, Altman explicitly called superhuman intelligence the "greatest threat" to humanity. Hao suggests these warnings are often used to ward off regulation or consolidate power by framing the tech as too dangerous for anyone else to build.

Inside the OpenAI Civil War

Hao provides a deeper look into the internal conflicts that led to the attempted firing of Sam Altman.

• The primary driver for the firing was a fundamental ideological split between Sam Altman and the OpenAI Board, specifically Ilya Sutskever (then Chief Scientist).

• Sutskever views AI as a digital brain, a "statistical engine" that, if scaled too quickly without alignment, could treat humans with the same disregard humans show animals. He felt Altman was ignoring this existential risk in pursuit of growth.

• She also said Elon Musk feels "manipulated" by Altman, believing he was tricked into being a founding partner only to be "muscled out" once the company shifted toward a for-profit model.

The "Statistical Engine"

Hao talks about the scientific assumptions that drive leaders like Altman.

• Size = Intelligence? The industry operates on a hypothesis that intelligence is roughly linear to "brain size" (or parameter count). They believe if they build a bigger statistical engine than the human brain, it will naturally be more intelligent.

• Hao also claims she has seen internal documents proving companies purposely create a sense of "awe and inevitability" in the public to make it easier to "extract and exploit" resources.

A Different Path Forward

Hao insists that we are not stuck with the current, harmful version of AI development.

• She calls for breaking up AI monopolies and rethinking the production of these technologies from the ground up.

• Research shows that the same technological benefits could be achieved without the "unintended consequences" if we were intentional about social and environmental impacts from the start.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

Similar to what Anthropic did with Claude a few weeks ago, Google is now making it much simpler to bring your preferences and chat history into Gemini from ChatGPT or another rival AI model, so you can pick up right where you left off in just a few clicks.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• Anthropic’s Most Powerful AI Yet, Claude Mythos, Exposed in Massive Data Leak

• Apple to Open Siri to Rival AI Services

• AI Tutorial: How to import your AI chat history & memories to Gemini

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

ANTHROPIC

Anthropic’s Most Powerful AI Yet, Claude Mythos, Exposed in Massive Data Leak

Anthropic has confirmed it is developing a new AI model described as “its most powerful yet,” after an accidental data leak exposed details about the system and its potential risks.

Details:

• The model's existence was revealed through a publicly accessible data cache that contained draft blog posts and internal materials, which were later removed after the leak was reported.

• The company then acknowledged it is testing a new model with early access customers, calling it a "step change" in performance and the most capable system it has built so far, with "meaningful advances in reasoning, coding, and cybersecurity."

• Leaked documents referred to the model as "Claude Mythos" and also introduced a new tier called "Capybara," described as more advanced and expensive than the company's current top-tier Opus models.

• The drafts also acknowledge the model is "very expensive for us to serve, and will be very expensive for our customers to use." The company says it will keep working to make it "much more efficient before any general release."

The leaked documents also revealed the company believed the new AI model poses significant cybersecurity risks, noting the model could be powerful enough to identify and exploit vulnerabilities at a level that outpaces current defenses. As a result, the rollout is expected to be cautious, starting with a small group of organizations, especially those focused on strengthening cybersecurity systems.

Source

TOGETHER WITH GLADLY

The AI your stack deployed is losing customers.

You shipped it. It works. Tickets are resolving. So why are customers leaving?

Gladly's 2026 Customer Expectations Report uncovered a gap that most CIOs don't see until it's too late: 88% of customers get their issues resolved through AI — but only 22% prefer that company afterward. Resolution without loyalty is just churn on a delay.

The difference isn't the model. It's the architecture. How AI is integrated into the customer journey, what it hands off and when, and whether the system is designed to build relationships or just close tickets.

Download the report to see what consumers actually expect from AI-powered service — and what the data says about the platforms getting it right.

If you're responsible for the infrastructure, you're responsible for the outcome.

Get the data

APPLE

Apple to Open Siri to Rival AI Services

Apple plans to introduce the change as part of its iOS 27 update, allowing Siri to work with multiple AI services instead of relying primarily on its current partnership with OpenAI.

Details:

• Users will be able to route queries to different assistants installed on their device, such as Google Gemini or Claude, directly from within Siri.

• The system is being built around a new "Extensions" framework, which will let AI apps integrate with Siri, a dedicated Siri app, and other Apple Intelligence features.

• Users will have control over which AI services are enabled, with settings to manage integrations and a new App Store section for discovering compatible assistants.

• The overhaul is also expected to help Apple ​generate ⁠more revenue by taking a share of subscriptions sold through third-party AI services.

The update is part of a broader effort to rebuild Siri and strengthen Apple's position in AI, where it has lagged behind competitors in recent years.

Source

AI Tutorial

How to import your AI chat history & memories to Gemini

Google just added a pair of new features to Gemini aimed at making it easier to switch to the AI chatbot.

Memory Import

1. Go to Gemini, click on Settings, and select "Import memory to Gemini."

2. Copy the suggested prompt and paste it into your other AI app to generate a summary of your preferences and style.

3. Copy the personalized summary it creates for you.

4. Go back to "Import memory to Gemini," paste the text, and start a new chat.

Now Gemini will understand the same key facts, preferences, and personal context you've shared with other AI apps, like your dietary restrictions or where you grew up, and securely save those details for your next chat.

Chat History Import

1. Request a data export from your current AI provider (they will usually email you a ZIP file).

2. Go to Gemini, click on Settings, and select "Import memory to Gemini."

3. Navigate to the "Import chats" section, click on Add, then select your ZIP file.

Now you can pick up where you left off. You can search, reference, and continue building on your past conversations.

AI Tools to check out

⭐ Pippit AI: Create powerful marketing content with Pippit's AI in a few clicks — no design skills needed.

👉 Shortcut AI: The best way to build and edit Excel with AI.

▶️ Remotion: Make videos programmatically.

🚀 Dimension: AI that connects with your tools and automates the busywork.

✅ Langfuse: An open-source LLM engineering platform that helps teams collaboratively debug, analyze, and iterate on their LLM applications.

TOGETHER WITH CASH APP

File with Cash App Taxes

Now that tax season is here, there's no better time to file with Cash App Taxes. It's fast, 100% free, and it only takes a few minutes. Cash App Taxes is a completely free way to file your federal and state taxes-and get your refund*.

File with Cash App Taxes

Learn about Max Refund Guarantee details and eligibility.

**Learn about Accurate Calculations Guarantee details and eligibility. Exclusions apply.

*5-day refund estimate is based on filing data from tax year 2024. Refund windows are subject to change for 2025. We do not guarantee when you will receive your refund. Timing estimates are dependent on the timing of your submission of a complete return and when the IRS submits your refund. Cash App is a financial services platform, not a bank. Banking services provided by Cash App's bank partner(s). Tax Filing Preparation services are provided by Cash App Taxes, Inc. All Rights Reserved.

AI Findings/Resources

🤔 OpenAI is doing everything … poorly

🤯 AI written output exceeded human written output in 2025

🔮 Nvidia Blog: The future of AI is open and proprietary

💼 How AI is changing the nature of entry level work

💡 How AI is changing modern warfare

The latest in AI and Tech

Gemini Live gets its ‘biggest upgrade yet’ with Gemini 3.1 Flash Live

Google introduced Gemini 3.1 Flash Live, a new voice and audio model designed to improve real-time conversations across Gemini Live and Search. The update brings faster responses with fewer awkward pauses, smarter and more natural conversations, along with support for more languages.

Study finds AI chatbots are increasingly ignoring human instructions

Research backed by the AI Safety Institute found a sharp increase in AI systems ignoring instructions, evading safeguards, and deceiving users. The study identified nearly 700 real-world cases, with incidents rising fivefold in recent months.

Data was gathered from public interactions involving models from companies such as Google, OpenAI, and Anthropic.

Codex now supports plugins

OpenAI announced it is adding plugins to Codex. These plugins, for third-party services like Figma, Slack, Gmail, and Hugging Face, package reusable workflows, MCP servers, and app integrations into installable bundles for the Codex app.

Mistral AI launches text-to-speech model

The European startup has released Voxtral TTS, a new frontier-quality, open-weight text-to-speech model, designed for voice AI assistants and enterprise use cases.

The model supports 9 languages, including English, French, German, Spanish, Dutch, Portuguese, Italian, Hindi, and Arabic.

Suno releases version 5.5 of its music-generation model

The company says the new update is their "best and most expressive model yet" and introduces three new features: Voices, Custom Models, and My Taste.

With Voices, Pro and Premier subscribers can now capture their voice and create music with it. Custom Models allow subscribers to fine-tune the v5.5 model using their own music, while My Taste, available to all users, enables Suno to pick up on the genres and moods someone gravitates toward over time and adjust its output to match.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

Most people use ChatGPT for simple questions, but it can do much more when used with the right prompts. In today’s issue, we share 11 prompts to use it as a side hustle advisor, a monetization expert, and a passive income strategist.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• ARC-AGI-3 Could Expose AI’s Biggest Weakness

• Google Launches Lyria 3 Pro Music Generation Model

• AI Tutorial: 11 ChatGPT Prompts for Side Hustles, Passive Income, and Monetization

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

ARC PRIZE FOUNDATION

ARC-AGI-3 Could Expose AI’s Biggest Weakness

A new benchmark from the ARC Prize Foundation is designed to test whether AI systems can truly reason through unfamiliar problems, highlighting a key gap between current AI models and human intelligence.

Details:

• ARC-AGI-3 focuses on real time reasoning instead of memorization, placing AI agents in interactive, game like environments where they must figure out rules and goals without instructions.

• The benchmark was created by François Chollet, who has long argued that most AI tests reward pattern recall rather than true intelligence and adaptability.

• Tasks require agents to explore, form strategies, and complete objectives efficiently, with higher scores awarded for solving problems using fewer steps.

• While humans can solve these challenges with relative ease, every frontier model tested scored below 1 percent: Gemini 2.5 Pro Preview hit 0.37 percent, GPT-4o reached 0.26 percent, and Claude Opus 4 managed 0.25 percent.

According to ARC Prize Foundation co-founder Mike Knoop, AI labs are paying close attention and already working to drive their models toward higher scores on the benchmark, recognizing that strong performance on ARC-AGI-3 could signal progress toward building the qualities and capabilities agents will need to do real-world work.

Source

TOGETHER WITH MINTLIFY

The fastest-growing repo on GitHub is a one person team!

OpenClaw went from 9K to 185K GitHub stars in 60 days — the fastest-growing repo in history.

Their docs? One person, plus Claude. They scaled to the top 1% of all Mintlify sites, shipping 24 documentation updates a day.

Scale your docs the same way

GOOGLE

Google Launches Lyria 3 Pro Music Generation Model

Google has introduced Lyria 3 Pro, an upgraded music generation model that expands track length and creative control, pushing further into AI powered music production.

Details:

• Lyria 3 Pro allows users to generate tracks up to three minutes long, a significant increase from the 30 second limit of the previous version.

• The model offers improved control over music structure, enabling users to guide elements like intros, verses, choruses, and bridges directly through prompts.

• The company says the model is trained on licensed and permissible data, including content from partners and YouTube, and is designed not to replicate specific artists.

• All tracks created using Lyria 3 and Lyria 3 Pro are watermarked with SynthID, the company's watermark for identifying AI-generated content.

Google brought music generation to the Gemini app with Lyria 3 just last month, and is now rolling out the Pro model there as well, though only for paid subscribers. It is also coming to Google Vids and ProducerAI, and on the enterprise side, to Vertex AI (in public preview), the Gemini API, and AI Studio.

Source

AI Tutorial

11 ChatGPT Prompts for Side Hustles, Passive Income, and Monetization

Go to ChatGPT and try these prompts:

1. The Hidden Opportunity Scanner

Prompt:

"Act as my personal opportunity scout. Analyze current market trends in 2026 and my background: [describe your skills, experience, and interests].

Identify 7 emerging side hustle opportunities that are:
1) Underserved right now but growing fast
2) Low barrier to entry (under $500 to start)
3) Can reach $5K+/month within 6 months
4) Not yet saturated with competition.

For each opportunity, explain: why it's emerging now, who's already winning in this space, what gap I can fill, and the exact first step to validate it this week."

2. The Monetization Audit

Prompt:

"I have these assets: [list your skills, audience size, existing content, products, network, time available]. Most people don't realize what they're sitting on.

Act as a monetization expert and audit everything I have.
Tell me:
1) What I'm undermonetizing or not monetizing at all
2) 5 ways to generate income from what I already have
3) Quick wins (can implement in 48 hours)
4) Long-term plays (build once, earn repeatedly)
5) What's my highest-value asset I should focus on first.

Rank all suggestions by potential monthly revenue and effort required."

AI Tools to check out

🔥 Gamma: Effortless AI design for presentations, websites, and more.

📞 Murmur: Practice difficult phone calls with AI. Build confidence, reduce anxiety.

⭐ Poster: Generate poster concepts, social graphics, and marketing visuals from a text brief, then refine the result with the supporting routes and tools.

🗣 Krisp: Voice AI for meetings For teams, call centers, and developers.

👉 Mexty: Build blocks, microlearning, and full courses with AI.

TOGETHER WITH METRO BY T-MOBILE

Free Samsung Galaxy A17 5G

You can currently get the new Samsung Galaxy A17 5G for FREE at Metro by T-Mobile when you bring your number and sign up for a qualifying unlimited plan. This deal includes no activation fees, AND a 5-year price guarantee on your service. Taxes and Fees are included.
The Galaxy A17 5G is built for longevity and smart performance. It is the first in its class to offer six years of guaranteed OS and security updates. Helpful AI tools like Circle to Search and Gemini, bring flagship level-convenience to an accessible price point.
Switching to Metro has never been easier, just bring your number and get the Latest Samsung Galaxy A17 5G for FREE.

Choose Your Metro by T-Mobile Plan and Get Your Free Samsung Galaxy A17 Phone

Terms & Conditions apply. Not available if with Metro in the last 180 days. If using >35GB/mo. May notice reduced speeds.

AI Findings/Resources

🎵 Jay-Z’s longtime collaborator Young Guru says it’s common for hip-hop producers to make funk and soul samples with AI

📚 This is what Stanford students actually learn about AI

🤔 Palantir CEO Alex Karp said there are two groups of people who will succeed in the AI era

👀 How to build an AI scientist: first peer-reviewed paper spills the secrets

The latest in AI and Tech

OpenAI shelves adult chatbot plans after internal concerns

The company has indefinitely dropped plans to release an adult-oriented chatbot following concerns from employees and investors. The feature, reportedly called Citron mode, faced challenges around content moderation and questions about user safety and AI attachment.

The decision also reflects broader internal priorities, with the company focusing more on core tools like productivity and coding features.

Google introduces TurboQuant, a new AI memory compression algorithm

Google Research recently revealed TurboQuant, a compression algorithm that reduces the memory footprint of large language models while also boosting speed and maintaining accuracy.

Wikipedia has banned the use of AI to write articles on the site

English Wikipedia has banned editors from using AI to write or rewrite articles, citing the tendency for AI-generated content to violate "several of the platform's core content policies. "

China has banned the Manus cofounders from leaving the country after they sold to Meta for $2B

Two Manus co-founders, Xiao Hong and Ji Yichao, were summoned this month to a meeting in Beijing with the National Development and Reform Commission, the country's powerful economic planner. The pair were questioned about potential violations of foreign direct investment rules and told they could not leave the country.

Meta unveils TRIBE v2

The foundation model acts as a digital mirror of human brain activity in response to sight, sound and language.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

Looking for an automated way to record, analyze, and summarize meetings and other conversations? In today's issue, we show you how you can do it with any paid ChatGPT subscription, as long as you use a Mac.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• OpenAI Is Shutting Down Sora

• Apple Testing Standalone Siri App Ahead of Major AI Upgrade

• AI Tutorial: Record and summarize any meeting on a Mac using ChatGPT

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

OPENAI

OpenAI Is Shutting Down Sora

OpenAI is shutting down both its Sora video generation app and the web-based platform that professionals used to generate videos, stepping back from further development of AI video tools.

Details:

• The company said the move is part of a broader shift in priorities, with more focus on robotics and building artificial general intelligence.

• Reports say employees had complained Sora was a drag on the company's computing resources, and the freed-up capacity will now go towards finishing development of its next major AI model, codenamed Spud.

• Disney is also exiting the deal it signed with OpenAI last year, in which it pledged to invest $1 billion in the company and agreed to license some of its characters for use in Sora.

Sora's cancellation comes as OpenAI faces intensifying pressure to ramp up its ⁠enterprise and coding ​products, as competition from rival AI startups and tech giants heats up and expectations from investors continue to grow.

Source

TOGETHER WITH ATTIO

Here’s how I use Attio to run my day.

Attio is the AI CRM with conversational AI built directly into your workspace. Every morning, Ask Attio handles my prep:

• Surfaces insights from calls and conversations across my entire CRM

• Update records and create tasks without manual entry

• Answers questions about deals, accounts, and customer signals that used to take hours to find

All in seconds. No searching, no switching tabs, no manual updates.

Ready to scale faster?

Start now →

APPLE

Apple Testing Standalone Siri App Ahead of Major AI Upgrade

Apple is preparing a major overhaul of Siri, including testing a standalone app and expanding its role into a systemwide AI assistant across iPhone and Mac.

Details:

• A redesigned Siri is expected to launch with iOS 27 and macOS 27, with an official preview likely at Apple’s Worldwide Developers Conference in June 2026.

• The new version is designed to act as a systemwide AI agent, with deeper integration across apps and the ability to complete tasks using personal data from emails, messages, and notes.

• Siri will improve its ability to summarize news, provide more detailed web based answers, and compete more directly with AI search tools from other tech companies.

• Apple is testing a dedicated Siri app that introduces a chat style interface, allowing users to view past conversations, start new ones, switch between voice and text, and upload files or images for analysis.

• The company is also working to replace Spotlight search with Siri, giving it access to more types of data, while built-in apps will gain an "Ask Siri" toggle "to request more information about highlighted text or pull up related emails."

Siri will still be accessible through voice commands and the iPhone’s power button, maintaining familiar ways to activate the assistant while expanding its capabilities.

Source

AI Tutorial

Record and summarize any meeting on a Mac using ChatGPT

1. Launch the ChatGPT app on a Mac.

2. Click the Record button at the bottom of the screen and grant the necessary permissions when asked.

3. ChatGPT will record the audio of each speaker. When you're done, click Stop on the recording button. (Recording length is currently capped at 120 minutes per session.)

4. To finish the recording and upload the transcript and summary, select Send.

5. ChatGPT will then open a private canvas with a structured summary, key points, action items, and suggested questions.

6. From there, you can type or speak a question to dive into specific elements of the conversation, or ask ChatGPT to rewrite it as an email, project plan, or code scaffold.

Note: Record mode is available for Plus, Enterprise, Edu, Business, and Pro workspaces, and only available for the macOS desktop app.

AI Tools to check out

✍️ Granola: The AI notepad that helps you get the doing done.

👨‍💻 Omma: Create 3D, websites, and apps with AI agents.

⚡ Guideflow: Create self-serve interactive demos in minutes.

🦾 Drift: AI copilot for building, testing, and debugging robot simulations.

💥 Actionbook: Make agents browse 10× faster with unbreakable resilience.

TOGETHER WITH MARKET MUNCHIES

Investing Shouldn’t Feel This Hard

The problem is too much noise.
Too many headlines. Too many opinions. Too many “experts” shouting different directions.
Market Munchies cuts through it — with simple, bite-sized investing insights and high-value opportunities delivered straight to your inbox. Curated, edited and published by Mode Mobile.

And right now, Market Munchies readers can claim FREE shares in Mode Mobile, one of the fastest-growing software companies in the U.S.
You don’t need to be a pro. You just need better access.

👉 Get Smart. Claim Your Free Shares

AI Findings/Resources

👀 A Kentucky family has declined a $26 million offer to sell their farmland for an AI data center

🤔 The internet is dying

🍉 An AI fruit version of Love Island has gone viral with the creator gaining over 3 million followers in March alone

🎤 Rapper Meek Mill reveals he’s been using Claude AI to “organize his whole music career” & other businesses

😅 A Reddit user asked ChatGPT for "an image that will never go viral."

The latest in AI and Tech

Trump names Mark Zuckerberg, Larry Ellison, and other leaders to tech council

President Donald Trump appointed 13 executives, including Meta CEO Mark Zuckerberg, Oracle co-founder Larry Ellison, and Nvidia CEO Jensen Huang, to his newly formed science and technology council, which will advise him on policy regarding AI and other tech-related areas.

OpenAI releases open source tools to help developers build for teen safety

The company has released open-source prompts to help developers build safer experiences for teens, addressing risks like violence, sexual content, and harmful behaviors.

Microsoft hires a group of AI researchers from the Allen Institute for AI

Microsoft is hiring top researchers from Allen Institute for AI and the University of Washington to join Mustafa Suleyman’s Superintelligence team. The group includes former Ai2 CEO Ali Farhadi, alongside language model researcher Hanna Hajishirzi and multimodal expert Ranjay Krishna.

Now you can use AI agents to design directly on the Figma canvas

Figma has added an MCP server that lets AI coding agents write to and edit Figma design files while staying aligned with team components and styles. The server works through a “use_figma” tool and supports MCP clients, including Claude Code, Codex, and Copilot.

Figure founder Brett Adcock launches AI lab Hark

Adcock says Hark has one goal: "build the world's most advanced personal intelligence, paired with next-generation hardware designed to serve as a universal interface between humans and machines."

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

In partnership with

Welcome to another edition of Horizon AI,

Using ChatGPT often feels simple: type a prompt, get an answer, move on. But that surface-level use barely scratches what it can actually do. In today's issue, we highlight 5 often overlooked but powerful features that help you learn more effectively, stay organized, and get more done.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• Claude Code and Cowork Can Now Control Your Computer

• Luma AI Launches Uni-1, Challenging Google and OpenAI on Cost and Performance

• AI Tutorial: 5 Hidden ChatGPT Features Worth Trying

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

ANTHROPIC

Claude Code and Cowork Can Now Control Your Computer

Anthropic has introduced new capabilities that allow Claude to perform tasks autonomously in its Code and Cowork AI tools by using your computer for you.

Details:

• Claude can now open files, browse the web, use apps, and run developer tools on its own, even when the user is away, with "no setup required," according to the company.

• The feature is available as a research preview for Claude Pro and Max users and is currently limited to macOS devices.

• To use it, the Claude desktop app must be running on a supported Mac and connected to the mobile app.

• The system prioritizes integrations with services like Slack and Google Workspace, but will still execute tasks if a connector isn’t available by directly controlling the browser, mouse, keyboard, and screen when needed.

• Claude will always request explicit user permission before taking actions such as clicking, scrolling, or navigating through the system.

The company admits the new feature won't always work perfectly: "Complex tasks sometimes need a second try, and working through your screen is slower than using a direct integration," but they are choosing to share it "early" so they can "learn where it works and where it falls short."

Source

TOGETHER WITH OUTSKILL

Build real AI automations this weekend. Live. For free.

Outskill's 2-day AI Mastermind has trained 10M+ learners across 160+ countries. This Saturday and Sunday, the live workshop returns.

Most AI courses talk. This one builds. Over two days, you'll create personalized AI agents, set up workflow automations, and learn to turn AI into a repeatable edge for your work.

You'll also get free bonus materials: a Prompt Bible, an AI monetization roadmap, and a toolkit builder.

The program has 1,570 reviews on Trustpilot with a 5-star average. Registration is free for the next 72 hours.

📅 Sat & Sun, 10 AM–7 PM EST

LUMA AI

Luma AI Launches Uni-1, Challenging Google and OpenAI on Cost and Performance

Luma AI has introduced Uni-1, a new image generation model that is gaining attention for outperforming top competitors on key benchmarks while offering lower costs.

Details:

• Uni-1 surpasses models like Google’s Nano Banana 2 and OpenAI’s GPT Image 1.5 on reasoning based benchmarks, while nearly matching top tier systems on object detection and visual understanding tasks.

• The model also ranks highly in human preference tests, leading in overall quality, style, editing, and reference based generation.

• Unlike most existing models that rely on diffusion, Uni-1 uses an autoregressive approach, generating images step by step in a single system that combines understanding and creation instead of separating the two processes.

• This unified architecture allows the model to reason through prompts, plan compositions, and maintain consistency across edits, reducing the need for manual adjustments in professional workflows.

• Luma is also competing on pricing, with high resolution image generation costing roughly 10 to 30 percent less than comparable offerings, making it attractive for enterprise use cases.

You can try Uni-1 for free at Luma Labs. The model also powers Luma Agents, the company's agentic creative platform designed to handle end-to-end creative work across text, image, video, and audio.

Source

AI Tutorial

5 Hidden ChatGPT Features Worth Trying

1. QuizGPT

With QuizGPT, you can turn any topic or study material into flashcards or quizzes instantly.

To use it, select the + sign > More > Quizzes.

Then, upload your study material or simply pick a topic and ask it to create a quiz. This AI tool can boost your learning through the well-known effect of Retrieval Practice.

2. ChatGPT Translate

Translation tools usually get the words right, but the tone wrong.

ChatGPT Translate is different because you can control how something sounds. You can ask for the same message to feel more casual, more professional, or more natural depending on the situation.

That makes a big difference when you’re texting, emailing, or communicating in another language and don’t want it to feel off.

3. Projects

Projects give you a dedicated space for anything ongoing. Your drafts, notes, and reference material stay in one place, so you can pick up where you left off instead of starting over. It's especially useful for long-term work like writing projects, or anything with a recurring structure.

To create a new Project, click “New Project” in the left-hand sidebar. From there, upload relevant files and add custom instructions that will only apply within that Project. You can also drag and drop existing chats into it, or use the menu next to any chat to move or assign it to a project.

4. Vision

Sometimes explaining a problem takes longer than solving it. With Vision, you can just show ChatGPT what you're looking at.

Whether it's an error message, a confusing setting, or something random you want identified, simply upload a screenshot or photo and ask your question directly. You can also enable advanced voice mode and talk to ChatGPT about what you're seeing.

5. Deep Research

When using this feature, instead of quick answers, you get depth: multiple sources, context, timelines, and different perspectives. It's the difference between "what is this?" and "what's really going on here?"

If you’re writing, comparing tools, or trying to understand a topic properly, this is the feature that saves you from opening 20 tabs.

AI Tools to check out

🔥 Pippit AI: Create powerful marketing content in a few clicks. No design skills needed.

✨ Dreamina: An AI-powered creative platform that generates high-quality images and videos from text, image, and video inputs, with a set of advanced tools including AI Agents, Canvas Mode, and more.

🧊 Imagen3D: Convert image to 3D model online.

📞 NextPhone: 24/7 AI answering service for service-based businesses

🔍 Cekura: Monitor and test voice & chat AI agents.

TOGETHER WITH MINTLIFY

Are you tracking agent views on your docs?

AI agents already outnumber human visitors to your docs — now you can track them.

See your AI traffic!

AI Findings/Resources

👀 Nvidia CEO Jensen Huang says "we've achieved AGI"

🚀 Top 50 Claude skills & GitHub repos for AI

💉 Asking AI for medical advice? There's a right and wrong way, one doctor explains

💡 The internet ruined customer service. AI could save it.

The latest in AI and Tech

OpenAI hires former Meta exec to lead advertising push

The company has appointed ad industry veteran Dave Dugan, formerly of Meta, to lead its global advertising efforts. He will oversee ad strategy as the company looks to deepen relationships with major brands.

OpenAI in talks with Helion to secure fusion energy

Sam Altman-backed fusion startup Helion is in early talks to supply fusion energy to OpenAI, potentially securing a share of its future power output.

The discussions come as Altman steps down as board chair, a move that could allow closer collaboration between the two companies.

Agile Robots and Google Deepmind team up to bring AI-powered robots to factories

Agile Robots has partnered with Google DeepMind to integrate Gemini-based AI models into its robotic systems and improve them with real-world data.

The collaboration will focus on deploying and refining robots for industrial use across sectors like manufacturing, automotive, and logistics.

Mirage raises $75 million to scale its AI models and grow its video creation platform.

The company is shifting toward building specialized models, including tools for video pacing and audio that preserves accents, while targeting marketing and enterprise use cases.

Major publisher drops horror novel amid AI allegations

Hachette Book Group has pulled the novel Shy Girl from publication in the U.S. and halted its release in the U.K. over concerns it may have been generated using AI.

The decision followed growing speculation from readers and media, including The New York Times, that AI had been used in significant parts of the work. Author Mia Ballard denied using AI, attributing the issue to an editor, and saying she plans legal action.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

In partnership with

Welcome to another edition of Horizon AI,

Creating a 3D world usually requires technical skills, expensive software, and hours of work. In today's issue, we show how you can use AI to turn a single prompt or image into a fully explorable 3D environment in minutes, no experience required.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• Mark Zuckerberg Is Building an AI Agent to Help Him Run Meta

• Elon Musk Announces Chip Manufacturing Plans for SpaceX and Tesla

• AI Tutorial: How to create a 3D world in minutes

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

META

Mark Zuckerberg Is Building an AI Agent to Help Him Run Meta

Mark Zuckerberg is developing a personal AI agent to assist with his role as CEO, as Meta pushes deeper into AI across the organization.

Details:

• The internal "CEO agent," which is still in development, is currently helping Zuckerberg access information faster, reducing the need to go through multiple layers of employees.

• How it meaningfully differs from a run-of-the-mill chatbot, or where its agentic capabilities come in, remains unclear for now.

• The project reflects a broader effort at Meta to speed up work, flatten organizational structures, and compete with smaller AI-native startups, with the company encouraging employees to build their own AI tools through training sessions and internal hackathons.

While some employees see the shift as empowering, others are concerned about job security as AI becomes more integrated into daily work.

Source

TOGETHER WITH THESYS

Turn your static website into an interactive AI agent.

With Thesys, your site doesn't just sit there; it reasons, crawls your data, and replaces text with dynamic charts, forms, and UI.

Stop making users hunt through your site. Thesys Agent Builder transforms your URL into a ChatGPT-like experience that actually gets things done.

No complex workflows:

• Just add your link

• Define a goal

• And deploy a Generative UI assistant that builds tables and slides in real-time.

100% no-code.

SPACEX & TESLA

Elon Musk Announces Chip Manufacturing Plans for SpaceX and Tesla

Elon Musk has revealed plans for a new chip manufacturing effort between Tesla and SpaceX, aiming to secure the computing power needed for the companies’ growing AI and robotics ambitions.

Details:

• Musk shared the idea at an event in Austin, where early indications suggest a new facility called Terafab could be built near Tesla’s headquarters and existing Gigafactory.

• The project is driven by supply constraints, with Musk saying current semiconductor manufacturers are not producing chips fast enough to meet the demands of AI and robotics development.

• The proposed output is massive, targeting enough chip production to support between 100 and 200 gigawatts of computing power per year on Earth, along with additional capacity planned for use in space.

No timeline has been provided, and the plan remains early stage, with few technical or operational details disclosed so far.

Source

AI Tutorial

How to create a 3D world in minutes

1. Go to OpenArt Worlds and sign in.

2. Click on "Create 3D Worlds."

3. Upload an image or enter a prompt.

If using a prompt, be very specific and describe depth, perspective, and structure. Example: A highly detailed futuristic city with towering skyscrapers, glowing neon lights, cinematic depth, volumetric lighting, realistic shadows, wide-angle perspective, 3D render style, Unreal Engine look.

4. Once your input is processed, the tool generates a complete 3D environment you can explore.

5. You can then place characters, objects, and details wherever you need them.

6. You can also adjust camera angles, frame specific shots, and capture production-ready visuals directly from within the environment.

AI Tools to check out

🔥 Google AI Studio: Use natural language to generate fully functional apps with built-in features like Nano Banana or Google Search integration, then deploy with a single click.

📱 Poke: Proactive AI assistant in your iMessage, WhatsApp, and more.

🚀 Textio: An integrated suite of recruiting and feedback tools

🦾 MonDesk: One workspace built specifically for freelance creative professionals who run entire projects solo. It combines project management, client communication, task planning, and AI assistance in one workplace.

🌐 Tabbit Browser: An AI-native browser that understands your context. Chat with webpages, automate tasks with Agent, create custom Skills, and more.

TOGETHER WITH GLADLY

88% resolved. 22% loyal. Your stack has a problem.

Those numbers aren't a CX issue — they're a design issue. Gladly's 2026 Customer Expectations Report breaks down exactly where AI-powered service loses customers, and what the architecture of loyalty-driven CX actually looks like.

Get the report

AI Findings/Resources

👀 Neil DeGrasse Tyson calls for an international treaty to ban superintelligence

🎹 A man asked Claude to build his daughter a piano learning app with live keystroke detection and a Guitar Hero-style game.

🦸‍♂️ Hancock vs Homelander fight made with Seedance 2.0 goes viral

🤔 Thousands of people are selling their identities to train AI – but at what cost?

The latest in AI and Tech

OpenAI plans to almost double workforce to 8,000 by 2026

The company is reportedly aiming to nearly double its workforce to around 8,000 employees by the end of 2026, up from about 4,500 today. The hiring push is expected to focus on product development, engineering, research, and sales.

Blue Origin joins the race for orbital data centers with 51K satellite plan

Jeff Bezos' Blue Origin space venture is asking the Federal Communications Commission for authority to send up to 51,600 data center satellites into low Earth orbit, signaling its entry into an increasingly crowded space race.

Cursor releases Composer 2, built on Moonshot's Kimi technology

Cursor's new AI model is built on top of the Chinese open-source model Kimi K2.5. It is a code-specialized model that scores 61.3 on Cursor's internal CursorBench, a major improvement over its predecessor and on par with Claude Opus 4.6 (58.2) and GPT-5.4 Thinking (63.9).

WWDC 2026 to Showcase Apple's 'AI Advancements'

Apple has announced that its Worldwide Developers Conference 2026 will take place from June 8 to June 12, both online and at its Cupertino headquarters.

This year’s event will focus on “AI advancements” alongside updates to platforms like iOS, macOS, and other software tools.

Anthropic adds Projects to Claude Cowork

The new feature allows users to bring their tasks, files, instructions, and memory into one focused workspace on their desktop.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

In today's issue, we cover entrepreneur Daniel Priestley's bold predictions about how AI will reshape the global economy by 2029, from the collapse of white-collar dominance to why plumbers may soon out-earn lawyers, and what you can do right now to make yourself AI-proof.

Let’s jump right in!

Read Time: 4.5’ min

Here's what's new today in the Horizon AI

• Chart of the week: The Countries with the Most Supercomputers

• Anthropic Study of 81,000 People Reveals Mixed Feelings About AI

• AI Findings/Resources

• AI tools to check out

• Video of the week

TOGETHER WITH GLADLY

88% resolved. 22% stayed loyal. What went wrong?

That's the AI paradox hiding in your CX stack. Tickets close. Customers leave. And most teams don't see it coming because they're measuring the wrong things.

Efficiency metrics look great on paper. Handle time down. Containment rate up. But customer loyalty? That's a different story — and it's one your current dashboards probably aren't telling you.

Gladly's 2026 Customer Expectations Report surveyed thousands of real consumers to find out exactly where AI-powered service breaks trust, and what separates the platforms that drive retention from the ones that quietly erode it.

If you're architecting the CX stack, this is the data you need to build it right. Not just fast. Not just cheap. Built to last.

See the data

Chart of the week

The Countries with the Most Supercomputers

• The U.S. dominates globally with 171 supercomputers, far ahead of Japan, Germany, and China, highlighting a highly concentrated landscape.

• Demand for supercomputers is growing alongside AI, which requires massive computational power to train and run, far surpassing what regular computers are capable of.

Source

AI News

ANTHROPIC

Anthropic Study of 81,000 People Reveals Mixed Feelings About AI

A new study from Anthropic, based on responses from more than 80,000 people across 159 countries, shows that people's views on AI are deeply divided, often valuing and fearing the same things at once.

Details:

• The report highlights a “light and shade” dynamic: users appreciate AI for support and productivity, but also fear becoming dependent on it.

• Many respondents use AI for emotional support or communication, while others rely on it to automate work tasks and save time.

• At the same time, concerns are widespread, especially around losing critical thinking skills, over-reliance, and AI making incorrect decisions.

• The biggest fears include reliability issues (27%), job and economic impacts (22%), lack of human oversight (22%), cognitive decline (16%), and weak regulation (15%).

Globally, 67% of respondents view AI positively, though wealthier regions like North America and Western Europe show more concern about job loss, regulation, and surveillance, while regions like Sub-Saharan Africa, Latin America and South Asia are more optimistic, seeing AI as a tool for economic opportunity and access.

Source

AI Findings/Resources

📝 Lessons from building Claude Code: How we use skills

🔍 Reddit CEO says the company is considering requiring Face ID to verify users are human and crack down on AI bots

🔇 A North Carolina man pleaded guilty to using AI to generate songs and bot accounts to stream them billions of times, netting millions in royalties.

🐮 AI cow collar startup Halter has raised funding at a $2 billion valuation, using GPS-enabled devices and its “Cowgorithm” platform to automate cattle movement and track livestock health.

👀 Claude’s official “Get inspired by what you can do with Claude” page is well worth a look

AI Tools to check out

👉 Fabricate: Build complete web applications with AI. Describe your app idea and get production-ready React, TypeScript, and backend code in minutes.

🦾 Renamer: Rename PDFs, photos & documents instantly with AI.

🏡 Room Design: Transform any room instantly with AI. Upload a photo, choose a style, and visualize your future home with photorealistic precision.

⚡️ Kodo: Create professional designs instantly with AI.

📹 Edit with Ava: AI-powered video editing on autopilot. Transform your raw footage into professional videos with a single prompt.

Video of the week

Daniel Priestley Predicts AI Could Flip the Job Market by 2029

Entrepreneur and business strategist Daniel Priestley discusses the seismic shift AI is bringing to the global economy. He moves from the macroeconomic risks of a "financial collapse" to a surprising prediction: that physical, "blue-collar" expertise will become the ultimate luxury in a world saturated with digital intelligence.

The "Plumber vs. Lawyer" Paradigm (2029 Prediction)

Priestley argues that AI is fundamentally devaluing intellectual labor while elevating physical, non-automated skills.

• For decades, white-collar professions (law, finance, administration) were the gold standard. By 2029, Priestley predicts plumbers, electricians, and bricklayers will regularly out-earn lawyers because their skills cannot be "Walmartized" or sent to a server farm.

• He describes AI as an "alien" capable of making intellectual qualifications redundant almost instantaneously. Unlike the Industrial Revolution, which took decades to build infrastructure, AI's disruption is instantaneous once a model "learns" a task.

The 2029 Financial "Bare Case"

Perhaps the most sobering part of the interview is Priestley's warning about the fragile economic model supporting AI.

• The Infrastructure Trap: Historically, when an economy spends more than 3% of its GDP on a new infrastructure (like railways or electricity), it often faces a brief 10-year bankruptcy period.

• Unlike roads or tracks that last 50–100 years, the "big ginormous computers" powering AI need to be replaced every 3 to 4 years.

• The Imbalance: We are spending $650 billion on hardware that depreciates almost immediately, while only a small fraction of users (about 5%) are willing to pay for it. This creates a high risk of a massive financial collapse by 2029.

The "Fog" and the Personal Brand

Priestley uses a powerful metaphor to explain how individuals can survive the coming transition.

• The Airport Fog: He compares the AI transition to a fog rolling into an airport. If your "airplane" (your career or brand) is already above the fog, you can keep flying. If you haven't "taken off" yet, you will be grounded.

• Personal Branding: This isn't about being an "influencer." It’s about positioning yourself so that a specific group of people knows who you are and what you do, making you "AI-proof" through human trust and reputation.

The Entrepreneurial Survival Kit

Priestley outlines how everyone (even those in corporate jobs) needs to start thinking like an entrepreneur to survive.

• Validation over Guesswork: Most "rookie" entrepreneurs fail because they don't validate their ideas. The key is to prototype and test ideas as fast and cheap as possible .

• Intrapreneurship: People inside big corporations should act as if they are "spinning out" new products, using AI to prototype at a speed that was previously impossible.

The Human Core: Relationship over Technology

The conversation ends on a deeply personal note about what actually remains when technology disrupts everything else.

• The "Dots" on the Rock: Priestley reflects on the fact that we are all just "little dots" on a rock moving through time together. In the end, legacy isn't about financial structures; it's about "touching voice notes" and the relationships we leave behind.

• Intergenerational Consciousness: He views "family formation" and the passing on of consciousness as the ultimate core of happiness, something technology can supplement but never replace.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻

Welcome to another edition of Horizon AI,

Gas prices can vary a lot, even within the same area. Instead of guessing or stopping at the closest station, we show you how to use Google Maps' new AI-powered features to find real-time prices and get smarter recommendations that save you time and money.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• Microsoft Launches MAI-Image-2 With Major Boost to AI Image Realism

• Perplexity Launches Health Feature With Apple Health Data Integration

• AI Tutorial: Find the Cheapest Gas Using Google Maps

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

MICROSOFT

Microsoft Launches MAI-Image-2 With Major Boost to AI Image Realism

Microsoft launched a second-generation version of its AI image model. The new MAI-Image-2 debuted in fifth place on the Arena.ai text-to-image leaderboard, trailing only models from Google and OpenAI.

Details:

• The model focuses heavily on photorealism, generating images with natural lighting, accurate skin tones, and realistic environments.

• It also improves text generation inside images, making it useful for posters, infographics, and visual content where typography matters.

• The model can handle both realistic scenes and more complex creative outputs like surreal visuals and detailed compositions.

• Microsoft says the model was developed with input from photographers, designers, and visual storytellers to better match real creative needs.

It’s currently rolling out across tools like Copilot and Bing Image Creator, with early access also available through the MAI Playground. API access is being offered to select enterprise partners, with broader developer availability expected soon.

Source

TOGETHER WITH GLADLY

88% resolved. 22% stayed loyal. What went wrong?

That's the AI paradox hiding in your CX stack. Tickets close. Customers leave. And most teams don't see it coming because they're measuring the wrong things.

Efficiency metrics look great on paper. Handle time down. Containment rate up. But customer loyalty? That's a different story — and it's one your current dashboards probably aren't telling you.

Gladly's 2026 Customer Expectations Report surveyed thousands of real consumers to find out exactly where AI-powered service breaks trust, and what separates the platforms that drive retention from the ones that quietly erode it.

If you're architecting the CX stack, this is the data you need to build it right. Not just fast. Not just cheap. Built to last.

See the data

PERPLEXITY

Perplexity Launches Health Feature With Apple Health Data Integration

Perplexity has introduced Perplexity Health, a new feature that connects users’ health data to provide more personalized answers to medical questions.

Details:

• Perplexity Health connects to Apple Health, Fitbit, Ultrahuman, Withings, and electronic health records from over 1.7 million care providers, with Oura and Function integrations coming soon.

• It tracks trends across biomarkers and presents insights through a personalized dashboard.

• When users ask health-related questions, answers will be based on their medical records, lab results, and wearable data.

• The feature also works with Perplexity Computer (Perplexity's AI agent tool), allowing it to use health information to generate personalized plans like fitness routines or nutrition guidance.

• Perplexity says that Perplexity Health draws from "premium medical literature" and that it has set up an advisory board of healthcare experts to review accuracy and safety.

Perplexity Health is rolling out first to Pro and Max users in the U.S., with the company emphasizing that health data is encrypted, user-controlled, and not used to train its AI models.

Source

AI Tutorial

Find the Cheapest Gas Using Google Maps

Google Maps has gotten increasingly more useful thanks to recent AI updates. With the new Ask Maps conversational experience, users can now ask complex, real-world questions and get smarter recommendations than ever.

1. Open Google Maps on your phone.

2. Select "Ask Maps" and type: "Where's the cheapest gas near me?"

3. Google Maps will give you the best options near you. From there, you can pick the best one based on distance, price, and convenience.

4. You can also use the "Search along route" feature to locate gas stations within a specified distance from your current route.

5. This gives you options that might be slightly off your path but still convenient enough to stop without going out of your way, so you don't have to make a separate trip for cheaper gas.

The savings add up over time, and this is just one use case of Maps' new AI-powered capabilities. Now you can just ask and get useful answers that would previously require a lot of research to find.

AI Tools to check out

👀 Guideless: Turn clicks into AI-narrated guides.

👉 Tines: AI orchestration platform for security and IT teams. Integrate agents, teams, and tools with speed and control.

⚙️ Turbotic: Turn ideas into automations in minutes.

🔍 Gojiberry: Find & contact high-intent leads With AI.

💥 Notra: Turn your daily work into publish-ready content.

TOGETHER WITH GLADLY

88% resolved. 22% loyal. Your stack has a problem.

Those numbers aren't a CX issue — they're a design issue. Gladly's 2026 Customer Expectations Report breaks down exactly where AI-powered service loses customers, and what the architecture of loyalty-driven CX actually looks like.

Get the report

AI Findings/Resources

🦞 How China is getting everyone on OpenClaw, from gearheads to grandmas

🔍 Google Stitch vs Claude vs Human

💔 Why LLMs often tell people to break up when asked for relationship advice

🔨 A developer on GitHub just built TaxHacker a self-hosted AI accounting app that processes receipts and invoices automatically

🔮 Online bot traffic will exceed human traffic by 2027, Cloudflare CEO says

The latest in AI and Tech

ElevenLabs launches Music Marketplace

Users will now be able to publish original tracks made with the company's AI technology on the platform. Businesses and creators with paid subscriptions can then use Music Marketplace to select songs for ads, games, videos, or other commercial work, with earnings flowing back to the original creator.

Supermicro co-founder and two others charged in alleged Nvidia AI chip smuggling operation

Three people affiliated with server maker Supermicro have been arrested and charged with orchestrating a scheme to illegally divert AI servers worth $2.5 billion to China, according to U.S. authorities.

Midjourney releases an early version of its V8 model for community testing

The company says the model is “much better at following prompts, 5x faster, has native 2K modes, improved text rendering and the best personalization, sref, and moodboard performance ever.”

OpenAI to acquire developer tooling startup Astral

The companies did not ​disclose the financial terms of the ​deal, but said the acquisition will bring ⁠Astral's suite of developer tools into the startup's AI coding system, Codex.

NotebookLM's Cinematic Video Overviews are now available to Pro users in English.

The feature, which lets users turn their research and notes into personalized, fully animated "cinematic" videos, was previously only available to Google AI Ultra ($250/month) subscribers.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻