Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

📰 This was the week that was...

This was the week the tools got sharper and the advice got more careful - inside the same seven days.

Two frontier labs converged on the same move this week: agents that pick up a task and carry it through to done. On 7 July, Anthropic's Claude Cowork - the mode where you hand Claude a task and it works across your files, calendar, email and the web until it's done - expanded from desktop-only to mobile and web. Two days later, on 9 July, OpenAI shipped GPT-5.6 and relaunched ChatGPT Work the same day, describing it as a tool that gathers context, plans the approach and takes action across your tools, files and desktop apps. Two labs, two "do the actual work" products, 48 hours apart. For scale on how real that appetite already is: OpenAI's Codex has gone from a niche coding tool to more than 5 million weekly users, a sixfold rise since February.

Then came the more interesting story. On 14 July, OpenAI published "How to manage AI investments in the agentic era" - a frontier lab telling its own customers to get a proper grip on their AI spend. That's OpenAI itself, in the same month it shipped its most capable and most agentic model yet, saying that capability and cost discipline need to move together.

That advice lands on ground already well prepared for it. Research published in March by the British Chambers of Commerce and Atos found UK SME AI adoption at 54%, up from 35% the year before - genuinely mainstream now. And TEKsystems' State of Digital Transformation 2026 report found enterprise-wide implementation had doubled year on year, from 12% to 24% - while confidence that AI spend will pay off within six months fell from 42% to 27% over the same period, and nearly a third of organisations reported costs quietly overrunning budget. I see the same pattern everywhere, across clients of every size and industry: the tool goes in after a lot of hard work deciding on platforms and then getting people to use it is ever harder. On the ground right now, two clients: both enabled AI tooling in January, both switched it on properly, and both are asking me some version of "how do we get people to use it?" and “Where’s the return”.

Enablement solved one problem. This week, both the tools and the labs that build them told us which one it opened. AI adoption.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week
✅ The adoption number everyone's quoting measures enablement - tools bought and switched on. Whether people actually changed how they work is a separate, much slower question
✅ Worth asking now: is your AI rollout horizontal (everyone, find your own use) or vertical (one process, one owner, one metric) - it decides whether you'll ever be able to see the payback

No panic needed this week. What it does call for is a proper look at whether what you enabled this year actually changed how anyone works, or just how fast they did the same thing.

🎯 Strategic Insight

Tension: 54% of UK SMEs now use AI, up from 35% a year ago, and enterprise-wide adoption has doubled too - both frontier labs even shipped agentic tools within 48 hours of each other this week. On paper, the adoption question is solved. In the same data, ROI confidence is falling and spend is quietly overrunning budget. A 54% figure that comfortable should have settled the question. It hasn't - the number counts something narrower than everyone assumes.

Optimistic insight: Here's the split I keep seeing, across clients of every size and industry: what gets reported as AI adoption is almost always AI enablement - the tool bought, the licence issued, the assistant switched on. Adoption is a different, much harder job: people actually changing how they work. Enablement is a purchasing decision, done in an afternoon. Adoption is a people-change project, and people-change is always slower than procurement. That gap is exactly why ROI confidence is falling in the same data that shows adoption "doubling": enablement is easy to buy and easy to count, adoption is neither, and most of what's being measured right now is the easy half.

There's a mechanical reason this gap matters: as AI pushes the cost of doing toward zero, doing stops being what makes you different, because your competitor's doing got cheaper at exactly the same rate as yours. Enablement buys everyone that same falling price. Adoption is the only lever left that's still yours to pull.

What's shifting: How the rollout happens explains why adoption is harder to see than enablement. In what I see across clients, roll AI out horizontally - Copilot or Claude switched on for everyone, find your own use - and the benefit is real but gets absorbed into the business: hard to trace at the task level, more a 3-5% movement on function KPIs than a headline win, which is exactly what "we can't see the payback" looks like from the CFO's chair. Roll it out vertically - one named use case at a time, run as a product with an owner, a metric and a cost line - and the benefit is designed in at the start, so it can be measured at the end. Most of what's been enabled this year went in horizontally, because horizontal is the easy purchase. Vertical is the slower, people-change work - which is exactly why it's the one still undone.

Takeaway: You already enabled the tool - that part's done. What's still undone is the room adoption needs. This week, give it to someone else: pick one team, hand them an afternoon with no task list, the AI you've already switched on, and one question - what would you build with this if speeding up the same report wasn't the brief? That's the adoption job: the people-change work the numbers don't measure, and it only moves when you give people room to explore what you've already enabled and imagine something more ambitious with it.

🤓 Geek Out

1️⃣ A £40 computer just proved you don't need the biggest model

Cambridge's Frugal AI Hub, working with the nonprofit Saving Voices Project, built a working speech-to-text system for the Soliga people of southern India from just five hours of recorded voice data. It runs offline on Raspberry Pi hardware costing about £40, needs no cloud connection, and is helping preserve a language that was never going to get frontier-model attention or a training-data budget from anyone in Silicon Valley. The wider project has similar ambitions for close to 500 million indigenous people across 90 countries, all built on the same principle: small, task-specific, and cheap enough to actually deploy. Rest of World has the full story.

Why it matters: This is proof that "good enough, cheap, and built for one specific job" beats "biggest, priciest, does everything" for most of what businesses actually need AI to do. If you're paying for headline capability you never touch, this is the nudge to check.

Action: Before your next AI tool renewal, list the three tasks you genuinely use it for, and check whether a smaller, cheaper, task-specific tool would do those three jobs just as well.

2️⃣ The whale-spotting AI that changed a shipping lane

Happywhale, a computer-vision platform that identifies individual whales and dolphins from photos submitted by anyone with a camera, won the AI for Planet category at this year's AI for Good Impact Awards, announced during the AI for Good Global Summit in Geneva earlier this month. It's cut identification time from over an hour per photo to seconds, catalogued 1.5 million photos with help from 53,000 citizen scientists in over 100 countries, and its data has directly influenced a new 40,000 square kilometre slow-navigation zone near Antarctica - a policy outcome that started with holiday snapshots.

Why it matters: This is what AI is actually good at - doing the tedious pattern-matching so humans can spend their time on the judgement calls and the decisions that follow. Most businesses have a version of this exact task sitting in their operations right now.

Action: Find the one task in your business that's essentially pattern-matching with photos or documents - invoice checking, enquiry sorting, compliance review - and ask whether just that piece, not the whole job, could be handed to AI this month.

3️⃣ The password problem nobody had solved until this week

Hand an AI agent your password so it can log into a site and finish the job, or stop and do the task yourself - until this week, that was the whole choice, and it's why most people won't let an agent near anything with a login screen. On 16 July, 1Password shipped 1Password for Claude, which lets Claude log in and complete tasks that need passwords or two-factor codes without the credential ever entering Claude's context or memory. When Claude reaches a login page, 1Password shows you exactly which credential is being asked for and why, you approve with Touch ID, and 1Password injects it straight into the page - Claude never sees the password or the code, and access is scoped to that one task and expires the moment it ends. A companion "Agentic Mode" locks the extension down the instant any AI agent takes the wheel of a browser, hiding its interface and limiting it to only the logins you approved - and it isn't Claude-only. 1Password calls this "zero-exposure architecture"; in plain English, the password never actually passes through the AI. As 1Password's CTO Nancy Wang put it: "The answer isn't handing agents your secrets. It is to let a user give an agent permission to use a credential without letting the agent see it."

Why it matters: Enabling an agent takes an afternoon; trusting it with something that matters is the slower part, and security has been the blocker. This closes that gap: the scaffolding for trusting an agent with real access arrived the same week both labs shipped agentic tools that need exactly this access to be useful.

Action: It's available now, not a beta - but Mac only at launch, and you'll need the 1Password desktop app and browser extension alongside the Claude desktop app and Claude-in-Chrome extension, all running on your existing 1Password plan. If you've got that stack, try it on one low-stakes login this week before you trust it with anything that matters. On Windows, this is one to watch for, not one to chase yet.

🎨 Weekend Playground

This week's Takeaway asked what your people would build with the room you gave them. Here's an hour to try the same question on yourself first, away from a screen for most of it - Hugo calls this a Wonder Ponder.

Why this matters: it's the smallest possible version of the adoption job - taking something you've already enabled and actually giving yourself the room to imagine what you'd build with it. (Quick disclosure: I'm a cofounder of Sherpas AI, and this exact exercise - walk, no screen, envision something bigger - is what we ask teenagers to do on our own programmes. If you want the fuller version, it's at www.sherpas-ai.com.)

👉 Mission:

• Go for a 20-minute walk with no phone in hand and no task list - just the question of what you'd build with an extra afternoon a week, if you let yourself be properly ambitious about it

• On Android: when you're back, record a two-minute voice note in the free Google Recorder app - it transcribes on-device as you talk

• On iPhone: record it in the built-in Voice Memos app, then use the dictation button in whatever free AI chat tool you already use to turn it into text. (I use Wisprflow)

• Read it back once tomorrow morning. Keep whatever's still true, bin the rest.

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else trying to handle the shift.

And here's the question I'm genuinely curious about this week: which process in your business have you actually redesigned with AI this year, versus just sped up? That question has a cost dimension too: the bill balloons because you're paying frontier prices to run the old job faster. I've just finished a briefing on exactly this, covering both ways I see AI roll out - horizontal and vertical - and the people side nobody budgets for, called "AI Adoption Without the Ballooning Bill." Reply and tell me what you're working on. I'll send you the briefing alongside a proper answer - I read every message and come back to everyone personally.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

4700 business leaders read this newsletter every week because someone they trust shared it with them. Together we can create the tomorrow we all want to live in, so please do share this with someone you trust your future with.

Time sensitivr: Sherpas AI just launched Future Founder - teenagers use real AI skills to explore career paths while building their own startup, not a simulation. It's what last week's Knowledge Was Power report highlighted gives teenagers a strong start to their career. Applications are open now at sherpas-ai.com. (Disclosure: I'm a co-founder of Sherpas AI. It is a social enterprise with a mission to create a generation of Ai fluent changemakers who can solve the worlds hardest problems.)

You can watch my podcast for AI for Equity with Jenny Garrett OBE and Lead-Sunshine Garrett to hear more about why I do this work. It was a wonderful conversation and great to sit down with a very cold drink in a cool place!

📰 This was the week that was...

This was the week the token bill came due - and the industry finally admitted the maths was never about price.

Uber burned through its entire 2026 AI budget by April, Microsoft revoked developer licences it had granted only months earlier, and one company reportedly ran up a $500 million bill after forgetting to set usage limits. Yet per-token prices keep falling - a study of 2.4 billion enterprise API calls found blended cost per million tokens down 67% year on year, from $18.40 to $6.07. Volume outgrew every budget model.

GitHub Copilot's move to token-based billing closed its first full metered month this week, and developers running autonomous coding sessions reported bills 10 to 50 times their old flat-rate plans - not from misuse, but because each agentic session makes dozens of model calls, resending the accumulated conversation every time.

Culture is catching up faster than tooling: Every reports that bragging rights have flipped from maxing out consumption to efficiency, with "revenue per million tokens" now circulating as a successor to revenue-per-employee. The FinOps Foundation finds 98% of organisations now actively manage AI spend, up from 31% two years ago.

None of that is a reason to slow down - it's a sign the discipline is arriving on schedule. More on the fix below.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week

✅ If you run agentic tools (Copilot, Claude Code, Cursor or similar), check whether a hard spending cap is switched on - several vendors now bill by default with no ceiling unless you set one

✅ Time to ask a procurement question you've probably never asked your AI vendor - this week's Strategic Insight, below

This isn't a week for panic. It's a week for a five-minute check of your billing settings.

🎯 Strategic Insight

Who in your AI arrangement is actually incentivised to bring your bill down?

Tension: Most leaders read this month's shock bills as "AI got too expensive" - and quietly assume their frontier lab will help them use less of what it bills them for.

Optimistic insight: That assumption was never true - and it isn't a crisis, it's a design problem. A frontier lab makes money on tokens, so it has no structural reason to want you using fewer of them; that is simply their incentive. The fix isn't a discipline memo asking your team to be careful - it's contractually and operationally separating who owns capability (the lab) from who owns cost (you). None of the three lanes below is universally right - picking one, on purpose, is the whole game.

• Frontier Partner - go direct to a lab's frontier model where raw capability moves the P&L, and accept being at the cutting edge requires an acceptance of spiraling costs.

• Orchestrated - put a routing layer between your business and the models, so cheap models handle routine work and expensive ones are saved for what actually needs them.

• Sovereign - run models on infrastructure you control, trading capex spend for total visibility (the sovereign end of last week's local-AI spectrum).

What's shifting: The Orchestrated lane has gone from theoretical to routine in about a year - tools like LiteLLM and OpenRouter now sit between businesses and models as standard, and a free tool that strips 60-95% of wasted spend before it reaches the model (see Geek Out) means the case for doing this builds itself. Strategist Stuart Winter-Tear has separately argued that the industry spent two years blaming the model for disappointing results, and has now converged on the same conclusion: the answer was in the organisation, not the model. Choosing a lane is that kind of decision, not a technology purchase.

Why this matters now: Successful AI adoption increases usage - the bill grows fastest exactly when your rollout is working. Have this sorted before you discover you need it.

Takeaway / Action: Before you sign or renew any frontier AI deal, ask your vendor and your own team one question: "who in this arrangement is incentivised to reduce my token consumption?" If the answer is nobody, that's your case for the Orchestrated lane - not a bigger budget line.

🤓 Geek Out

The Free Tool That Shrinks Your AI Bill Without You Lifting a Finger

Most of what you pay for isn't the AI's answer - it's everything it has to read first: old messages, files, logs, pasted-in reference material. Headroom is a free tool that quietly shrinks all of that before it reaches the model, cutting 60-95% off your token bill for the same answers. Nothing is lost - the full originals stay saved, and the model can pull them back if it genuinely needs to. Independent tests show accuracy holds up, and on some measures it actually improves.

Why it matters: This is the Orchestrated lane made real - a way to cut your bill without touching accuracy. If you're paying per token for repetitive tool calls or long logs, this is what catches it before the invoice does.

👉 Action: Run this on one repeated AI workflow: ask it to break down what it's spending at each stage - which model, roughly how many tokens in and out, what each step is for, and whether that spend was actually necessary (flag anything it can't be sure of as an estimate). Find the step where the spend doesn't match the difficulty of the work, then change one thing - trim what you're feeding it, tighten your instructions, or switch that step to a cheaper model - and try again. It's Every's method for auditing agent spend; you still need a human to review it, because AI isn't good at spotting its own waste.

👉 Read about Headroom

A Shorter Way to Feed Data to AI - Same Information, 40% Fewer Tokens

If your business feeds AI things like reports, spreadsheets or system data, it's probably being sent in a much wordier format than it needs to be. TOON is a free way to write the same information in far fewer words - it strips out the repeated formatting that computer systems normally pad every entry with. Independent tests show it uses around 40% fewer tokens than the standard format most systems already use, with no drop in accuracy - if anything, slightly better. I've used it myself on data going into prompts, and the saving is real: a rare case where the cheaper option is also the better one.

Why it matters: If your business feeds structured data - reports, tables, system exports - into AI, the format you use is a lever on your bill you're probably not pulling yet.

👉 Action: Next time you're building a prompt around structured data, try TOON instead of the usual format and compare the token count - it's a straight swap, not a rebuild.

👉 Read about TOON

Feeding Text to a Model as a Picture, Not a Sentence

DeepSeek's OCR research asks an odd question: what if a model saw a page as an image instead of reading typed words? DeepSeek-OCR compresses roughly ten text tokens into a single "vision" token while holding 97% precision, and DeepSeek-OCR-2 - the next version - was open-sourced in January 2026. It's the furthest-out signal here: if reading is one of the biggest silent costs in a long AI conversation, re-engineering how the model reads may matter more than any prompt trick.

Why it matters: This is a preview of where the token meter itself may be rebuilt from the ground up - worth knowing about now, well before it reaches your everyday tools.

👉 Action: Nothing to implement yet - file this one under "watch," and expect optical compression to start showing up inside mainstream tools within the year.

👉 Read about DeepSeek-OCR

🎨 Weekend Playground

This weekend, open PlotLines - ten out-of-copyright classic novels, from Frankenstein to Mrs Dalloway, played out journey by journey on a genuine period map. Chris Moran (LinkedIn), the Guardian's editorial innovation lead, vibecoded it over a weekend while his daughter studied David Copperfield.

Why this matters: Moran's own reflection is the more interesting bit - the map was nice, but the real learning was the back-and-forth with the machine and the rabbit holes it opened (he lost an afternoon to Victorian coaching routes). That's a better template for a weekend with a young person than the tool itself: build something together, then chase one tangent as far as it goes.

👉 Mission:

• Pick a novel on PlotLines with a young person in your life and follow its journey across the map together

• Ask what surprises them about the route - distances, place names, how people actually travelled then

• Chase one rabbit hole as far as it goes - a place, a mode of transport, a historical detail neither of you knew - and see where an hour of curiosity takes you

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift.

And here's the question I'm genuinely curious about this week: if you had to name the one AI cost in your business that nobody is currently accountable for, what would it be? Reply and tell me - I read every message and I'll come back to you personally.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

📰 This was the week that was...

This was the week two frontier labs ran into the same government wall - and open, local AI quietly had a big week of its own.

Two weeks after a US export-control order forced it offline, Claude Fable 5 returned worldwide on Wednesday - Anthropic's own testing showed less capable models could already do what triggered the ban. The same week brought Claude Sonnet 5, running close to Opus 4.8 on reasoning and tool use at a fraction of the price, and Claude Science, a research workbench for genomics, drug discovery and structural biology. Anthropic wasn't the only one hitting government limits: OpenAI's GPT-5.6 Sol launched the same week, restricted to around 20 US-government-approved companies while wider access gets worked out.

Away from the frontier labs, local AI had a big week of its own - a food-delivery company trained a trillion-parameter model on domestic chips, and finding a model that actually runs on your own hardware just got a lot easier. Full story in Geek Out.

None of that was the week's most important release, though. Sherpas AI launched Knowledge Was Power, a report asking what actually survives in knowledge work as AI advances - and finding the answer is most of what mattered in the first place. (Disclosure: I wrote it, and I'm a co-founder of Sherpas AI.) It gets the full treatment below.

From the Humancraft report: To answer the question ‘How does AI impact my role, you have to understand what work is, what humans are in the context of work and what AI is. hat’s the research and analysis conducted for this report.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week

✅ Sonnet 5 landed close to Opus-level agentic performance at a fraction of the price - worth trialling on a workflow you've been running on a pricier model

✅ Time to think about who you're hiring for, not just what you're automating - this week's Strategic Insight, below

This isn't a week for panic. It's a week for redesigning the entry-level.

🎯 Strategic Insight

If AI can only replace 5% of a human's role, what does that mean for your AI adoption plans?

📊 Stat of the week: 65% of the 67 human capabilities behind knowledge work are structurally durable against AI for the next five to ten years, according to Sherpas AI's new Knowledge Was Power report - and that durability sits inside just one of the seven things a person brings to a role. Do the maths, and the theoretical ceiling on what AI can replace in any given role is about 5%.

Tension: Most AI adoption plans quietly assume a bigger prize than the evidence supports - the unstated logic behind a lot of restructuring is "AI will eventually do most of this role", and headcount plans get built on it.

Optimistic insight: The report took knowledge work apart into 67 human capabilities and rated each one for how much ground AI is likely to take by the early 2030s. 65% hold firm for five to ten years - and that durability sits inside just Skill, one of seven things a person actually brings to a role. Wisdom, Energy, Identity, Standing, Connection and Presence make up the rest, and AI barely touches them. If you're planning redundancies in the thousands on the promise that AI does most of the job, that begs a real question: what exactly are you telling yourself you're losing?

What's shifting: The tasks inside that 5% - drafting, sorting, summarising, formatting - are exactly the tasks most junior roles were built around, so entry-level work is disproportionately exposed to the one thing AI can actually do. That's not a case for removing those roles; it's a case for redesigning them around the 95% AI can't touch, to protect your talent pipeline. The harder part sits outside any one employer's walls: the traditional bridges from education into work - placements, internships, apprenticeships - are eroding faster than replacements are being built, which means employers also need to push externally for a better bridge between education and government, not just fix it inside their own business.

Why this matters now: If you employ anyone under 25, that overexposure is either being designed around or ignored. Employers who redesign entry-level work now build the senior talent they'll need in five years; employers who freeze hiring and wait are betting someone else fixes the bridge for them.

Takeaway / Action: Pick one entry-level role this month and redesign it around Humancraft, not just skills - give it real client contact, real accountability, and a genuine reference at the end. Sherpas AI and Startup Sherpas are two working templates for what that looks like; the design pattern - paid, structured, real client, referenced - isn't proprietary, and is worth copying whatever size business you run.

(Disclosure: I'm a co-founder of Sherpas AI and Startup Sherpas, and the report's author.)

🤓 Geek Out

This Was the Week Local AI Grew Up

A Beijing food-delivery giant just trained a frontier-scale model: Meituan open-sourced LongCat-2.0, 1.6 trillion parameters, built entirely on domestic Chinese chips - the first model that size to train and run without Nvidia hardware. The same week, Hugging Face shipped a hardware filter for its Models page: set your chip once, see only the local models that actually fit. A Stanford study, covered by Let's Data Science, shows why that matters - models under 20 billion parameters now match frontier cloud models on 88.7% of everyday queries, and the efficiency gap has closed 5.3x since 2023.

Why it matters: Three signals, one direction: frontier AI no longer needs a chip supply chain two or three companies control, a genuinely capable model now runs locally on most everyday tasks, and finding the right one for your hardware just got easier. For a business wary of vendor lock-in or losing a tool overnight (see this week's Fable story, above), that's real optionality.

👉 Action: If your team has been assuming "local AI" means compromising on capability, try Hugging Face's hardware filter against your actual laptop spec and see what's realistic today. 👉 Meituan's LongCat-2.0 · Hugging Face's hardware filter · the Stanford study

What Worldview Is Your AI Quietly Feeding You?

The Economist tested 25 frontier models against the World Values Survey, the decades-long global poll of human belief, and found every one sits in the same rich-country corner - several more extreme than any actual nation surveyed. GPT is more secular than any country on earth; Gemini weighs individual freedom more heavily than any population surveyed. Ask the same political question in a different language and the answer often shifts, because a model trained on a repressive country's internet inherits that country's slant in that language, whatever the lab's own politics. Chinese models parrot the state line on sensitive topics; Western models are just as opinionated, only less visible about it, since alignment happens behind closed doors.

Diagram from the Economist article

Why it matters: Every model you deploy is quietly making subjective calls - what to include in a summary, how to frame a sensitive question, which side of a contested issue to lean toward. Multiply that by the roughly one billion people already using generative AI regularly, and a model's "personality" stops being a curiosity and starts shaping outcomes.

👉 Action: Next time AI helps with something that has a judgement call in it - hiring language, a comms draft, a market read - run it through two different models and see where they diverge. The gap is the thing worth interrogating. 👉 Read the Economist's investigation

AI Fiction Has a Tell

Researchers at Maryland and Google DeepMind built StoryScope, a tool that ignores AI writing's surface style and measures the narrative decisions underneath: how morally messy the plot is, whether the timeline jumps, whether the story explains its own meaning or trusts the reader to get it. Five leading models - Claude, GPT, Gemini, DeepSeek and Kimi - each rewrote the same 10,000-plus human short stories; a detector trained on structure alone spotted the AI version 93% of the time, even after the AI rewrote its own prose to sound less robotic. AI narrators spell out the theme outright 77% of the time versus 52% for humans, and human stories land in the rarest 10% of narrative territory nearly three times as often.

As Eamon, director of engineering at Proton, put it: "Turns out that AI slop looks like poop 💩😂"

Why it matters: Once you can name what makes writing feel machine-made - tidy, over-explained, allergic to moral ambiguity - you can train that quality up or down on purpose, whether you want a voice that reads unmistakably human or a genuinely surprising co-writer.

👉 Action: Next time AI drafts something for you, check whether it's over-explaining its own point rather than trusting you to get it - that's the tell this research keeps finding. 👉 Read the paper

🎨 Weekend Playground

This weekend, read Knowledge Was Power - Sherpas AI's new report on what survives AI in the world of work - with a teenager in your life, then drop it into NotebookLM and turn your conversation into a podcast you can both listen back to.

Download the full report.

Why this matters: The report's own advice for parents is to protect space for a teenager to be present with an adult outside the immediate family, talking about something real. Turning the report into a five-minute podcast (or a short video, using NotebookLM's video overview) is a way to have that conversation without it feeling like homework.

👉 Mission:

• Read the report together, or just the two-page summary at the front if you're short on time

• Drop the PDF into NotebookLM and generate an Audio Overview (or Video Overview) - let your teenager pick the format

• Ask them one question from the report: of the seven things a person brings to a job - skill, wisdom, energy, identity, standing, connection, presence - which do they already have, and which do they want to build this year?

• If they're 13 to 25, help them apply to Sherpas AI's next paid AI training cohort - applications are open now - and if you know another family with teenagers, forward this on to them too

(Disclosure: Sherpas AI and Startup Sherpas are Hugo's own programmes.)

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift.

And here's the question I'm curious about this week: of the seven things you bring to your role - skill, wisdom, energy, identity, standing, connection, presence - which one do you value most in your colleagues ? Reply and tell me - I read every message and I'll come back to you personally.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

📰 This was the week that was...

This was the week three governments drew three different maps of the AI future - and one company decided to ask a much bigger question.

The UK is heading into a new political chapter. Keir Starmer resigned on Monday, with Andy Burnham on course to become the UK's sixth Prime Minister since the Brexit vote. What the new government chooses to do with AI policy matters - a new hand is on the tiller at exactly the moment the rules are being written.

Meanwhile Brussels and Buenos Aires are going in opposite directions. The EU's Council and Parliament agreed to simplify and streamline the AI Act - a signal that the original rules were too complex for the businesses they were meant to protect. President Milei took the opposite view, using an FT op-ed to invite AI companies to Argentina with the promise of non-human corporations run by AI, minimal regulation, and low taxes. Three governments. Three very different bets.

Midjourney announced Midjourney Medical - full-body ultrasonic scanners, 60 seconds, built into spas. It earns its own space in Geek Out.

Access to Fable remains suspended.

Let's get into it.

🔥 Urgent Priorities

✅ The regulatory map is being drawn in three different directions simultaneously - EU simplifying, Argentina deregulating, UK in political transition. Know where your AI stack's compliance exposure sits, especially if you operate across borders

✅ The Fable question from last week is still live - "what would we do if our most-used AI tool disappeared on Monday?" still deserves an answer

✅ Prediction is a core business capability now, not a strategic luxury - this is the week to start practising it deliberately

This isn't a week for panic. It's a week for looking around corners.

🎯 Strategic Insight

The week the moat question changed.

Tension: Most organisations use AI to move faster - more of what they already do, in less time. Bain & Company is now vibe-coding functional replicas of software acquisition targets - building working imitations of entire companies in days - to find out whether the code is actually the defensible part of the business. In most cases, it isn't. The code is replicable. The moat is somewhere else.

Optimistic insight: This is clarifying, not frightening. As AI drives the cost of doing toward zero, what rises in value is what cannot be replicated at zero cost. Natural intelligence - your people's judgement, relationships, and lived experience - supported by silicon intelligence, not replaced by it. Four things keep appearing at the top of the genuinely defensible list: meaningful human connection, holding risk wisely by predicting forward, proprietary data, and clarity of purpose. Each goes up in value as AI commoditises everything else.

What's shifting: If you can vibe-code a replica of any software business in a weekend, the bottleneck in M&A shifts from due diligence to imagination. The question moves from "is this defensible?" to "where will value concentrate in the next 12 months - and how do we get there first?" That is a prediction question. And the discipline of making deliberate bets about the future is becoming a genuine competitive advantage - not for M&A firms alone, but for every business navigating a market this fast.

Why this matters now: Every business is making bets about where value will land. Most are making them reactively - waiting for the market to confirm before they move. The firms who look back on this period as their inflection point will be the ones who built a view of the future and committed to it early.

Takeaway: Take one strategic question your business has been deferring because you "need more data." Reframe it as a prediction: what do you believe is true about the next 12 months? Write it down. Test it. Prediction is not certainty - it is the discipline of looking further out than the next quarter.

🤓 Geek Out

Midjourney Goes to the Doctor

👉 Read the announcement here

Midjourney has announced Midjourney Medical: full-body ultrasonic scanners that image your entire body in 60 seconds, delivered through spas you'd want to visit anyway. Half a million tiny sensors build a 3D body map to a fraction of a millimetre - comparable to MRI quality, at a fraction of the cost and time. First spa opens in San Francisco in 2027. Goal: a billion scans a month by 2031. No investors. Community-funded.

Why it matters: Midjourney was losing the image generation race. Rather than fight harder on the same ground, they looked at their core capability - turning complex sensor data into detailed images - and asked: what is the biggest problem we could solve with this? Healthcare data scarcity was the answer. Most people never get a whole-body scan unless something is already wrong. A 60-second spa visit changes that. This is what raising your ambition looks like: not incremental improvement, but taking your existing assets to a much harder, more important problem.

👉 Action: Run the same question on your own business: what capabilities do we already have that could solve a much bigger problem than the one we're currently applying them to? 

The Frugal AI Switchboard

👉 Explore the Switchboard

The Frugal AI Switchboard - built by James Martin at BetterTech using Mistral - lets you find the right model for each type of task: coding, data analysis, meetings, writing, image generation, customer support. Alongside the capability comparison, it shows the energy use and carbon footprint of each choice.

Why it matters: Defaulting to frontier models for everything is the AI equivalent of taking a Rolls-Royce to the corner shop. The Switchboard makes the trade-offs visible - and the frugal path is usually greener, cheaper, and just as capable for the job in hand.

👉 Action: How are you ensuring your AI strategy is appropriately frugal? The Switchboard is a good place to start the audit.

The Administration of Skills Goes to Zero

👉 Explore Airskills

I had a conversation this week with Chris Parsons, who has been working on a quiet but consequential problem in AI adoption: you write a skill once, it lives on one machine, in one tool, and quietly diverges from everyone else's version. His tool, Airskills, backs up, publishes, and syncs your skills across every agent and machine - Claude, Copilot, Cursor, ChatGPT - in a standard open format with no lock-in. MIT licence. Open source.

Why it matters: The cost of writing a skill is falling fast. The cost of administering them - keeping them current, consistent, and distributed across tools and teams - is still surprisingly high. Airskills drives that cost toward zero, freeing time back for harder problems and bigger ambitions.

👉 Action: If your team is building AI skills, try Airskills for your next one. If you're not yet building skills - a conversation with Chris is a good place to start.

🎨 Weekend Playground

Warp - built by my friend Chris Barry - is a calmer kind of workspace where conversations turn into briefs, dashboards, forms and small tools, without ever leaving the thread. This weekend: try it on a family project.

Why this matters: Most AI tools are built for work. Warp works just as well for the stuff outside it - and using it on something personal is often the fastest way to understand what it can really do.

👉 Mission:

• Pick a real family project - a trip, a gathering, a home plan - and open a Warp conversation about it

• Let the conversation shape itself into something useful: a brief, a checklist, a small tool

• Works on Android, iPhone, or browser

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift.

And here's the question I'm genuinely curious about this week: if you could predict one thing about where AI will take your industry in the next 12 months, what would you bet on? Reply and tell me - I read every message and I'll come back to you personally.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

📰 This was the week that was...

This was the week a government reached in and switched a frontier AI model off.

Last Friday, the US government issued an export control directive ordering Anthropic to suspend all access to Fable 5 and Mythos 5 for any foreign national, inside or outside the United States. The practical effect was blunt: to comply, Anthropic had to disable both models for everyone. Here's the silver lining - Fable 5 had only been public for three days, so almost no one had built anything serious on it yet. This was a warning shot, not a wound. Anthropic's other models were untouched, and the company says it disagrees with the order and is working to restore access - but the precedent is now set, and it arrived at the one moment it cost businesses nothing to notice. A single government decision, citing national security, can remove a tool that hundreds of millions of people were using the day before.

In the same week, the counter-story arrived from the opposite direction. Z.ai released GLM-5.2 - an open-weights model under a permissive MIT licence, with a one-million-token context window - and it went straight to number one on the Design Arena leaderboard, ahead of Fable 5. Open weights means you can download it and run it on your own hardware. Frontier-level capability that no government can switch off remotely, because the model lives on your machine, not someone else's.

Two stories, one week, pointing in opposite directions: the most capable models can be taken away from you, and frontier-class capability is becoming something you can hold yourself. That tension is what this whole edition is about - and it's a conversation Ben and I have been having since the very first night of AI Night School, under a name we kept coming back to: digital self-sovereignty.

Let's get into it.

🔥 Urgent Priorities

✅ If anything in your business ran on Fable 5 or Mythos 5, you lost it last week - check what depended on it, and what your fallback was

✅ Map your single points of failure: which workflows assume one provider, one model, one country?

✅ This is the week to treat "what if our AI access disappeared on Monday?" as a real planning question, not a hypothetical

This isn't a week for panic. It's a week for treating AI like the critical business system it now is.

🎯 Strategic Insight

The week's real lesson isn't about politics. It's about maturity.

Tension: Most leaders choose AI tools the way they'd choose a SaaS subscription - on capability and price - and assume access will always be there. Last week proved that access is a variable, not a constant. A model you depend on can be removed by a decision made in a room you'll never sit in.

Optimistic insight: Here's the part that should steady your nerves. You already own every discipline you need to handle this. Governance. Risk frameworks. Business continuity planning. Procurement. Technical architecture. These are the muscles every well-run organisation already has for its other critical systems - and AI is now one of those systems. You don't need a new skillset. You need to point the operational maturity you already have at the AI stack. This is natural intelligence - your existing judgement and process - supported by silicon intelligence, not replaced by it.

What's shifting: The cleanest way to think about it is this: every AI use case in your business is now an AI product. And products need product thinking - what is it built on, what happens if that foundation disappears, and how do we make it leaner and more resilient? That last move is what I've been calling frugalising your AI: designing each AI use case into the component parts and developing using the smallest, cheapest, most replaceable model that still meets the bar (including turning AI into code), rather than reflexively reaching for the most expensive frontier model for everything.

Why this matters now: The US government has shown it will step in. Play the tail risk forward: what would actually happen to your operations if US-based AI companies were prohibited from serving your market entirely? For most businesses the real answer is "we haven't thought about it" - and that's the gap to close. Self-sovereignty here isn't anti-American or anti-frontier. It's resilience. It's knowing your business keeps running whoever controls the model.

One sharp aside, because it matters: any company betting it can ride AI to growth while making mass redundancies is in for a hard lesson. Running AI as critical infrastructure takes the breadth and depth of your people's operational maturity. Hollow out the organisation and you lose the very capability you need to make the transformation work.

Takeaway: Block 30 minutes this week for an "AI continuity" check. List your top three AI use cases. For each one, write down what it's built on (which model, which provider, which country) and answer one question: if this vanished on Monday, what's our fallback? The blanks you can't fill are your real priorities.

🤓 Geek Out

The AI That Runs On Your Own Machine - For Free

👉 Read it here

Investor Tomasz Tunguz this week captured where "local" AI has got to: free models, small enough to run on an ordinary laptop or office server, are now good enough to handle a real share of everyday work - with nothing ever leaving your building. They're not as fast or as capable as a paid frontier model like Claude (picture a willing junior who needs clear instructions, rather than a senior who thinks alongside you), but they cost nothing to run, work with no internet connection, and answer to no one but you. The catch worth knowing before anyone gets excited: a raw model is not a finished product. What makes "Claude" feel effortless is a layer of polish and safety the big labs build on top - and if you run a model yourself, you take that work on, which often costs three to five times the hardware bill once you count the staff time.

Why it matters: For a growing slice of routine work, "good enough, free, and entirely under your control" is now a real option - and a tool you own outright is the strongest protection there is against any single supplier being switched off.

👉 Action: Put one question to whoever handles your tech - "is there a job we do every day where a model running on our own kit would be good enough, and remove our reliance on an outside provider?" The answer tells you how exposed you actually are.

The Tiny Model That "Beat" The Giants - And Why The Quotes Matter

👉 Read it here

A nine-person team at Weibo released a model, VibeThinker, around a hundred times smaller than the big frontier models - and it scored level with them on certain maths and logic tests. Cue a familiar argument across the industry, because the headline ("tiny model matches the giants") is true and misleading at the same time. On the narrow tasks it was built for, it really does keep up. On general knowledge and judgement, it falls a long way behind. The team say so plainly; the viral version of the story does not.

Why it matters: This is one of the most useful habits a leader can build this year - reading any AI performance claim with one raised eyebrow. A tool that's brilliant at one narrow thing is not a tool that can do everything, and the gap between those two sentences is where a lot of budget gets wasted on the wrong purchase.

👉 Action: Next time a supplier quotes you an impressive score, ask the one question that cuts through it - "impressive at what, exactly, and how does it do at everything else?" Watch how the room changes.

Sovereign AI Stops Being A Slogan

👉 See the UK programme

While US frontier models dominate the headlines, the rest of the world has been building alternatives - and this is the week it stopped looking optional. Here at home, the UK's £500 million Sovereign AI programme is now putting real money into British AI companies, backed by a national supercomputer and a billion pounds of public compute. In Europe, Mistral - already shipping capable open models from Paris - is reportedly raising around €3 billion at a €20 billion valuation to build out European AI infrastructure. The picture that's forming is one where capable AI comes from several places, governed by several rule-books.

Why it matters: US frontier models are no longer the only show in town. For any leader thinking about resilience, a stack that can draw on more than one country's models - and more than one company's goodwill - is a stronger stack. Optionality is the new sophistication.

👉 Action: Add one non-US model to your awareness list this quarter - a Mistral or a GLM - and have someone trial it on a real task, just to know it's there if you ever need it.

🎨 Weekend Playground

James passed me A rational conversation on where AI is actually going - Benedict Evans on Lenny's Podcast - with a strong recommendation, and it earns it. The line that's stayed with me: the right question about your job isn't "what percentage of this can AI do?" but "is this a task, or a job?" It's a good companion to this week's edition - a calm, sceptical hour on where the value actually lands.

Why this matters: Your instinct for where AI fits - and where it doesn't - is a business asset. It develops by listening to the clearest thinkers, not by chasing the loudest headlines.

👉 Mission:

• Find a walk, a commute, or an hour where you're not also doing something else

• Put the episode on (free on Spotify - app or browser, Android or iPhone)

• When it ends, write one sentence naming a task in your business AI could take, and a job that will always need a human

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift.

And here's the question I'm curious about this week: if your most-used AI tool disappeared on Monday morning, how long would it take your business to recover - and what's your fallback? Reply and tell me - I read every message and I'll come back to you personally.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

📰 This was the week that was...

This was the week AI became two completely different things at once.

Claude Fable 5 launched publicly on 9 June - the first Mythos-class model available to everyone, not just the fifty research partners who had early access. State-of-the-art across software engineering, knowledge work, vision, and scientific research. Stripe completed a 50-million-line Ruby codebase migration in a single day, against an estimate of two months. I've used it. It is a clear step change from Opus 4.8.

At the same time, Perplexity launched Search as Code - not a new model, but a rearchitecture of how AI agents do search. Composable building blocks that agents can orchestrate per task, rather than a black-box API call. On a benchmark of 230+ cybersecurity vulnerabilities, it hit 100% accuracy against under 25% for competing systems including GPT-5.5 with web search - at 85% lower token cost. The kind of infrastructure shift that quietly changes what's economically viable to build.

And elsewhere, 25 open-weight models dropped in a single week - chat, image, speech, music, video, 3D. The densest open-weight release window on record.

The question the week raises isn't whether AI is extraordinary. It's whether you need extraordinary.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week

✅ Fable 5 is genuinely worth benchmarking against your highest-value work - the Stripe migration isn't a demo, it's evidence

✅ One note for enterprise teams: Anthropic's mandatory 30-day data retention now applies to all Fable 5 traffic, including accounts with zero-retention agreements. Worth a quick check with your legal or compliance team.

This isn't a week for panic. It's a week for making one clear decision about where in the AI stack your business actually needs to sit.

🎯 Strategic Insight

The week's real story is the 1000x price spread between the most powerful model and the cheapest.

Many AI leaders will see Anthropic launch Fable 5 and assume they should be benchmarking toward it. That makes sense - it is the most capable public model ever released. But the interesting question this week isn't what Fable 5 can do. It's the gap between what Fable 5 costs and what capable AI costs at the other end of the market.

Before this week, the spread between the most expensive public AI model and the cheapest capable one was around 300x. Fable 5's arrival doesn't narrow that gap - it widens it. The model sits at $10 per million input tokens, $50 per million output. Compare that to Gemini Flash-Lite at $0.10 input and $0.40 output - 125 times cheaper on output. Or Qwen3.5 0.8B at $0.05 per million output tokens - a factor of 1,000. These aren't toy models. They handle most standard business tasks capably.

At the same time, the top of the market is clustering surprisingly close together. Fable 5 at $50 output, GPT-5.5 at $30 - they're nearer to each other than either is to Gemini's mid-tier. Two distinct markets are forming: a premium tier where the frontier labs compete on raw capability, and a commodity floor racing toward zero. You can track this live at artificialanalysis.ai and pricepertoken.com.

The optimistic insight is that this is good news for your business. You don't need to solve "which model is best." You need to solve a much smaller problem: which of the Three Jobs - building headspace for AI in your organisation, making room in the P&L through AI-driven efficiency, or genuinely reinventing your business model - actually requires frontier intelligence?

For most headspace and efficiency work, mid-tier models are already extraordinary. For discovering genuine reinvention plays and orchestrating complexity, Fable 5 earns its premium. The strategic mistake is applying frontier-tier thinking to every job on the list.

This is why I have built the frugaliser - a tool that takes an AI use case I have developed in Claude and asks “What could be turned into code? What could be done with a smaller AI model? What could be removed?”

Takeaway: Take 20 minutes this week to map your current AI usage against the Three Jobs. For each one: frontier problem or commodity problem? The answer will probably surprise you - and might unlock a significant budget reallocation.

🤓 Geek Out

Less Memory, Same Intelligence

While Anthropic was asking "how much can we pack in?", a team of researchers at Tencent were asking the opposite question. Their paper, published this week, proposes Lookahead Sparse Attention - a new inference architecture that compresses the memory footprint of running a frontier long-context model down to just 13.5% of the original, while maintaining or slightly improving accuracy. At 500,000-token context scales, it eliminates over 90% of the memory overhead. The paper's own term for the approach: "less is more."

Why it matters: The frugal AI thesis is the direct counterweight to the Fable 5 race. For every week that frontier capability pushes upward, a team like this is working out how to deliver comparable results at a fraction of the cost. The mid-tier of the AI market is not standing still. The intelligence you can afford is improving faster than most businesses realise.

👉 Action: Read the abstract - four minutes, and it reframes the week's biggest story from a different angle. If your team runs long-context tasks - document analysis, knowledge retrieval, codebase work - this is the design philosophy that will define the mid-tier market within two years. 👉 Read it here

The CEO of Anthropic Wants His Own Model Regulated Like an Airplane

AI moves at startup speed. Policy moves like Treebeard from Lord of the Rings - a sentient tree who takes a full day just to say hello. That's the opening image in Policy on the AI Exponential - an essay from Dario Amodei passed to me by Stanislaw Wozniak that's worth 30 minutes of your time.

The essay covers five areas where AI demands urgent policy action: safety regulation, macroeconomics and jobs, scientific acceleration, civil liberties, and geopolitics. The most striking single point: Dario is calling for FAA-style pre-release testing requirements for frontier AI models - including Fable 5 - with the power to block deployment if a model fails. Alongside the essay, Anthropic has released a formal legislative proposal.

Why it matters: This is the person who built the week's most powerful model saying it now belongs in the same category as airplanes - powerful, essential, and dangerous if poorly governed. Not a PR exercise. A legislative proposal. For any leader thinking about AI strategy, this is the clearest signal yet of where the regulatory landscape is heading.

👉 Action: Read the policy architecture sections - the safety regulation framework and geopolitics especially. The macroeconomics and jobs section is worth reading with a critical eye: technologists consistently overestimate how much of work AI can displace, because they see the world through what's technically possible rather than what people actually do all day. My own data analysis on this lands soon. 👉 Read it here

The Real AI Frontier: Systems, Not Models

The businesses that extract the most value from AI in the next 18 months won't necessarily be the ones that upgraded to Fable 5. A survey published this week on arXiv maps what's actually happening in serious AI deployment - and it's about compound AI systems: architectures that combine multiple models, retrievers, agents, tools and orchestrators into coordinated workflows. The paper defines the field, proposes a taxonomy across four paradigms (RAG, LLM agents, multimodal LLMs, orchestration), and maps the design trade-offs.

Why it matters: The most capable AI deployments right now compose models together - including cheap mid-tier ones - into systems that do things no single model can. This survey is a map of that territory, and it's where the competitive advantage will actually sit.

👉 Action: Spend ten minutes on the taxonomy section. It will sharpen every AI conversation you have with vendors, developers, or your own team. 👉 Read it here

🎨 Weekend Playground

Sophie Tidman at Mayvin passed me So You Want to Be a Sorcerer in the Age of Mythic Powers... (the AI Episode) this week, and she was right to. It's about a specific idea in storytelling: untethered magic. Magic that exists in the world without a system - without rules, without a spellbook, without points to spend. You encounter it in Miyazaki films, in the stranger corners of Ursula Le Guin, in folk tales where the magic is never explained, only experienced. Just the world being strange and alive.

The recurring figure in stories like this is the sorcerer's apprentice - someone who releases magic they don't fully control, then has to call the sorcerer in to fix it. This week, Dario Amodei spent 6,000 words calling for FAA-style regulation of his own model. That's not a coincidence. That's the sorcerer's apprentice, in real time. And it means the untethered magic frame isn't just a beautiful idea in stories - it's the most accurate description we have of where AI actually is right now.

Why this matters: How you design your AI governance processes need to match what you are governing and this is a good analogy for what we are trying to govern when we talk about AI.

👉 Mission:

• Find a walk, a commute, or 45 minutes when you're not also doing something else

• Put the podcast on

• When it's done, write one sentence about something your business can now do that would have seemed like magic two years ago

Available on Spotify - free tier works. Android or iPhone app, or open in your browser.

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift.

And here's the question I'm curious about this week: Which AI job in your business are you overinvesting in - applying frontier-model thinking to something that commodity intelligence would handle just as well? Reply and tell me - I read every message and I'll come back to you personally.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

The next AI Leaders Fellowship cohort is forming - a six-week programme for leaders who want to stop chasing AI headlines and start building the headspace, habits and confidence to put it to work. If that is the shift you are trying to make in your own organisation, come and have a look.

📰 This was the week that was...

The models keep getting better. The biggest wins now are the ones you already have the tools to grab.

Since the last edition landed, the frontier moved again. Anthropic shipped Claude Opus 4.8, now its most capable model, at the same price as the version before it. Google made its fast, cheap Gemini 3.5 Flash the default for billions of users. And OpenAI quietly upgraded the everyday ChatGPT engine with GPT-5.5 Instant, claiming far fewer made-up answers. Four frontier-class releases in barely three weeks.

For most UK businesses, none of that changes Monday morning. The capability already sitting on your desk outruns how you use it. The bottleneck has quietly moved from the lab to the office. What is in it for a UK leader this week? Permission to stop chasing model names, and start closing the gap between what your tools can do and what your people actually do with them.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week

✅ The model race is now a backdrop, not a deadline

✅ Time to plan for adoption, not acquisition

This isn't a week for panic. It's a week for picking one team and helping them actually use what you already pay for.

🎯 Strategic Insight

Tension: It is tempting to treat AI as a buying decision. Pick the best model, sign up, wait for the results. But capability is no longer the scarce thing. Four frontier models launched this month and the price of intelligence kept falling. When the thing everyone is buying becomes cheap and abundant, owning it stops being an advantage.

Optimistic insight: As the cost of doing collapses towards zero, the value of the human system that absorbs and directs it rises. The new scarce resource is absorption: how quickly your people turn raw capability into changed habits. Absorption does not scale with budget. It scales with leadership attention and a willingness to change how work gets done. And size is not the head start many assume. The larger the organisation, the more habits there are to re-wire, the more teams to bring along, and the more work adoption takes. This is a leadership challenge that belongs to every business, whatever its scale.

What's shifting: The smart question is no longer "which model should we buy?" but "how fast can our people absorb what we already have?" The frontier will keep leapfrogging itself, probably monthly. You cannot win that race, and you do not need to.

The barriers that remain are human, not technical. Beyond the training gap sits a quieter one: the rise of the conscientious objector. These are often capable, thoughtful people with genuine reservations about using AI - about quality, about ethics, about what it means for their craft. The instinct is to mandate past them. The better move is to hold space for the candid conversation about what actually worries them. Adoption is culture change, and culture change runs on honesty, not compliance.

Why this matters now: The competitive bar for adoption is still embarrassingly low. The data shows most firms have bought the tools but barely begun the harder work of helping people change how they work. In our own work we keep seeing two patterns. The horizontal rollout hands everyone a licence, a Microsoft Copilot seat for all, and hopes adoption follows. It spreads thin and rarely changes behaviour. The vertical rollout picks a few specific, high-value use cases and makes them genuinely work. That second path produces what the first never does: success stories, measured ROI, and proof. And proof is what moves a sceptical culture, including those conscientious objectors. People change when they see a colleague's result, not when they are told to.

👉 Takeaway: This month, go vertical with one real use case rather than wide with a tool:

• Pick one team and one specific, repeatable task worth doing well.

• Set a clear expectation for what good looks like, and protect a little time each week for it.

• Name one enthusiast as the champion and let them teach the others.

• Measure the hours saved and the ROI, then tell that story widely.

Finding the use cases worth backing is the hard part. Most organisations have no shortage of ideas; they have no reliable way to choose the right ones. That is exactly why we developed Differential with Onepoint, an approach that takes a long list of AI ideas and narrows it to the handful worth backing, governance-ready, in weeks rather than months. If you would like help finding yours, that is where to start.

🤓 Geek-Out Stories

1️⃣ The leaky bucket: most AI spend is quietly draining away

One analysis of 178 companies found a striking mismatch. Around sixty per cent had bought AI tools for their people, yet only one in four employees felt supported in changing how they actually work. Three-quarters wanted training; barely a tenth got it. The piece, which you can read here, calls this the leaky bucket, and offers a simple fix: treat AI as a change programme, not a tool purchase.

Why it matters: If you have bought licences and seen little change, you are normal, and the problem is fixable. The money was never the hard part. The change management is.

👉 Action: Pull your AI subscription list and ask one question per tool - "who was actually shown how to use this, and when?"

2️⃣ What the AI-forward companies actually do

A widely-read study of six AI-forward firms - including Shopify, Zapier, Duolingo and Intercom - pulled out the tactics that genuinely drive adoption. The standouts are refreshingly human: explain the how and not just the why, cut the red tape around experimentation, track usage openly, and turn your in-house enthusiasts into teachers rather than leaving their know-how hidden.

Why it matters: None of these need a big budget or a new platform. They are leadership moves, which puts them squarely within reach of any business, at any size.

👉 Action: Find the person in your team who is quietly brilliant with AI already, and give them thirty minutes to show three colleagues one thing that saves them time.

3️⃣ Governance as an accelerator, not a brake

A common fear is that getting serious about AI rules will slow everyone down. A recent piece on AI governance argues the opposite. There is no such thing as "no AI" - your people are already using it. Clear, simple guidance on which tools and which data are fine, with light-touch standards behind it, lets people experiment with confidence instead of hesitating in the grey.

Why it matters: For any leadership team, a one-page "what is fine, what needs a check" note removes the fear that quietly stalls adoption, without the cost of a formal compliance programme.

👉 Action: Write a single page answering two questions for your team: which AI tools are approved, and what kinds of information should never go into them.

🎨 Weekend Playground

This weekend, try NotebookLM, Google's free research assistant that only ever works from sources you give it.

Upload a few of your own documents - a strategy paper, a messy set of meeting notes, a long report you have been avoiding - and click to generate an Audio Overview. Two AI hosts will turn your material into a surprisingly listenable "deep dive" you can play in the car. It is grounded entirely in your sources, so it is your own thinking handed back to you, clearer.

Why this matters: It is the whole theme of this week in one experiment. The magic was never in the model. It shows up when you point a capable tool at your own material and your real problems.

👉 Mission:

• Create a free notebook and upload three documents you actually need to understand.

• Generate an Audio Overview and listen to it on a walk this weekend.

• Then open the chat and ask it the one question you were dreading having to read forty pages to answer.

• Notice how it feels to have your own work explained back to you, and where that could save you an hour next week.

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift. If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

A quick share from home: Harpenden-AI is a community project upskilling our whole town in AI fluency - training sessions, drop-ins, and live projects using AI to solve real local problems. Weeks old, already gaining momentum, and built as a model any town can adopt. If you'd like to lead something similar in your community, reach out and join the fun.

📰 This was the week that was...

This was the week AI video grew up. Google launched Gemini Omni Flash on 19 May, putting conversational video editing into the Gemini app, Google Flow and YouTube Shorts. Describe a change in plain English, watch the video update, repeat. Free, at scale, for billions.

Underneath that, the market shifted. OpenAI's consumer Sora app went dark on 26 April. Alibaba's HappyHorse-1.0 has quietly become the favourite when real people rate AI video clips blind - and Alibaba has promised to release it for free. The spectacle phase is ending. What is arriving is more useful: a production layer that sits inside how you already work, bringing video within reach of every business, not just the ones with a creative budget.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week ✅ AI video has quietly crossed from demo to deployable ✅ Time to plan for personal video in your customer relationships, not a one-off experiment

This isn't a week for panic. It's a week for opening Google Flow on your phone and seeing what a personal video to your best customer could look like in five minutes.

🎯 Strategic Insight

This week I've been working with Duncan Nicholls, fellow AI Optimist and world-renowned sports photographer, on how AI can enhance creativity rather than replace it. Then Ben Relles dropped into my DMs with what he's building at Make Believe - a brilliant example of exactly what Duncan and I have been exploring. This insight comes straight from that work.

Tension: Most UK SMEs don't make video - not because they didn't want to, but because it was too expensive. A 60-second customer story used to need a crew and a four-figure invoice, so leaders used text, stock photography, or nothing.

Optimistic insight: That floor has just lifted. Personalised video - made for one customer, one prospect, one team member - is now within reach of every business. Here the economics get interesting: as the cost of doing collapses to near zero, the value of being genuinely human rises. When everyone can produce video, what differentiates is whether yours makes the person on the other end feel actually seen. Natural intelligence, supported by silicon intelligence, makes that possible at human scale.

What's shifting: The smart question is no longer "Can we afford to make a video?" but "Which of our relationships would deepen if we showed up on video, personally, at the moment they needed us?" That is a relationship question, not a marketing one - and the answers are very different.

Why this matters now: Plan for "cheaper marketing videos" and you will get cheaper marketing videos. Plan for "personal, human, video-first relationships" and you get three things competitors can't quickly copy: deeper retention, stronger word-of-mouth, and a brand that feels like a person rather than a logo.

👉 Takeaway: Between now and the end of Q3, ask one question of each customer touchpoint: would this moment land better as a short, personal video?

• The welcome after a first purchase

• The follow-up after a sales conversation

• The quarterly check-in with your best 20 customers

• The thank-you to someone who referred business

• The internal "well done" to whoever closed the deal

Pick one. Make ten this month. Watch what happens to the replies.

To explore what personal video could do for your relationships, speak to Duncan - or to Jenny Law and Jim Birdsell, two other creatives at the cutting edge of AI video.

🤓 Geek-Out Stories

1️⃣ AI is now generating tornadoes to train self-driving cars

In February, Waymo unveiled the Waymo World Model, built on Google DeepMind's Genie 3. It generates photorealistic 3D driving scenarios - tornadoes, flooded streets, vehicles going the wrong way, even an elephant on the road - to train autonomous vehicles on rare events before they happen in real life. Engineers summon a new edge case from a text prompt.

Why it matters: This is AI video doing serious industrial work, not making marketing clips. For UK SMEs, the principle transfers: anywhere you train people on rare-but-critical events (safety drills, awkward customer escalations, equipment failures), AI-generated walkthroughs are a much cheaper alternative to scripted role-play.

👉 Action: Pick one rare scenario your team struggles to train for. Could an AI-generated walkthrough give them a useful first encounter before the real one arrives?

2️⃣ ElevenLabs has paid creators $11 million in voice royalties - and just launched a music marketplace

ElevenLabs, the London- and New York-headquartered AI audio company, has paid more than $11 million to creators through its Voice Marketplace, where voice actors licence consent-based AI clones of their own voices and earn each time a paid subscriber uses them. In March, it launched a Music Marketplace on the same model. The company is now valued at $11 billion.

Why it matters: This is the optimistic counter-narrative to "AI replaces creatives". A consent-based, royalty-paying marketplace respects rights, creates new income streams, and gives UK SMEs one of the safest commercial sources for AI voiceovers and background music.

👉 Action: Next time you need a voiceover or background track for a video, source it through a consent-based marketplace. Your future legal team will thank you.

3️⃣ The clock is ticking on AI video transparency rules in Europe

The EU AI Act's Article 50 transparency obligations come into force on 2 August 2026 (the European Parliament's draft Digital Omnibus may push watermarking to December). Providers must embed machine-readable marks on every AI output. Any business deploying AI to create deepfakes or public-facing AI content must visibly label it. UK businesses trading into the EU are in scope.

Why it matters: Labelling AI content is about to become a cost of doing business in Europe. Most major tools (SynthID, OpenAI, ElevenLabs) are already building it in. The risk is generating AI video with no provenance trail and finding in August that none of it is compliant.

👉 Action: Add one line to your AI policy this quarter: "All public-facing AI-generated content must carry visible disclosure and machine-readable provenance." Then check that your tools support it.

🎨 Weekend Playground

This weekend, watch Ben Relles' short film announcing his new AI lab, Make Believe. Present-day Ben appears alongside himself at a 1980s school dance, a 1990s startup office, a 2000s viral video shoot. Same person, three decades, one seamless cut. Backed by Reid Hoffman, Make Believe is building interactive video that talks back to viewers - cooking shows where the chef answers your questions, fitness creators who critique your form.

It is strange, beautiful, and a small window onto where personal AI video is heading: synthetic memories made with consent, by the person they are about.

Why this matters: It is the most concrete answer to "what is AI video actually for?" that has shown up in months - a new format entirely, personal and intimate.

👉 Mission:

• Watch the film twice - once for the spectacle, once for the emotional beat

• Pick one moment from your business's own past and ask: what would it mean to bring it back to life?

• Notice your reaction - excitement, unease, both? That is useful data about how customers will feel

• Resist the urge to commission one immediately

Before you go: which one customer relationship in your business would change if you sent them a short, personal video this week?

One more invitation before you head into the weekend: applications are open for the next AI Leaders Fellowship. Join a community of leaders learning to use AI for positive impact in their own work and across their organisations. The journey kicks off on 4 June and runs for six weeks, Thursdays at 8am. If you've been waiting for the right moment to take this seriously, this is it.

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift. If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

📰 This was the week that was...

This was the week AI stopped predicting text and started predicting the world. At Google I/O 2026, Sundar Pichai put it plainly: "With world models, AI is moving from predicting text to simulating reality." Google unveiled Gemini 3.5 Flash, a 24/7 cloud-based agent called Spark, and a research model that helps forecast hurricanes. The throughline was clear: prediction is becoming the product.

At the same time, Cursor launched Composer 2.5, an agentic coding model built on Kimi K2.5 that matches Claude Opus 4.7 at a fraction of the cost. Frontier capability keeps cascading downwards into cheaper, more accessible tools. What's in it for a UK business leader? The AI era is quietly asking you to make bolder decisions, more frequently, with less data, against a more uncertain world. And the same AI is finally giving you the tools to do exactly that.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week

✅ Prediction and simulation moving from research labs into mainstream tools

✅ Time to plan how your business will build a prediction capability, not just chase cost savings

This isn't a week for panic. It's a week for asking which of your decisions could be sharper if you had a better view of what's coming next.

🎯 Strategic Insight

Tension: Most SME leaders are looking at AI through a cost-saving lens - automate the email, summarise the meeting, draft the proposal. That captures real value. But it misses the deeper shift AI is creating: the world is now moving faster, with less certainty and more cadence than your existing planning rhythms can handle. Cost-saving alone leaves you better at running yesterday's business in a world that no longer behaves like yesterday.

Optimistic insight: The AI agenda is asking every leader to make bolder decisions, more frequently, with less data, against a more uncertain world. That sounds like a problem - until you notice that AI is also producing the tools to answer it. LLMs now match expert human "superforecasters" on standard forecasting benchmarks. Weather and demand can be simulated 1,000 times faster than a year ago. Customer reactions can be tested against synthetic populations in hours rather than weeks. This is natural intelligence supported by silicon intelligence in its purest form: humans bring context and judgement, AI brings breadth, speed and consistency. Together they outperform either alone.

What's shifting: Prediction is becoming a core organisational capability, not a function reserved for finance teams, strategy consultants or hurricane forecasters. Every business now needs a way to forecast faster, stress-test more often, and decide with confidence in the face of incomplete information. The smart question is no longer "How do we save costs with AI?" but "What predictions do we need to make better, and how often?"

Why this matters now: If you only plan for "AI as cost-saving", you'll free up some hours and miss the bigger shift. If you instead plan for "AI as a prediction capability", you get three benefits: faster decisions, lower decision fatigue, and a team that learns alongside the tools. In a world where decisions arrive thicker and faster, the businesses that build a prediction muscle will pull ahead - quietly, compoundingly, and without burning out their leaders.

👉 Takeaway: Pick one decision you make every month - a sales forecast, a hiring call, a stock order, a campaign budget - and build a simple prediction loop between now and the end of Q2:

• Ask Claude or Gemini to forecast the outcome, with reasoning

• Have your team make their own forecast, independently

• Average the two, or take the higher-confidence one with the other as a sanity check

• Track which approach is closest at the end of the quarter

If you'd like help designing your own prediction capability, reply and we'll send you our one-page template.

🤓 Geek-Out Stories

1️⃣ Weather forecasting just got 1,000x faster - and open source

NVIDIA's Earth-2 platform launched in January 2026 at the American Meteorological Society meeting. It bundles three AI models that produce 15-day global forecasts, kilometre-scale local storm predictions, and high-resolution downscaling - all roughly 1,000 times faster than traditional physics-based models. The Israel Meteorological Service is already running it with a 90% reduction in compute; TotalEnergies, AXA and S&P Global use it for energy and insurance risk. The models are free on Hugging Face and GitHub.

Why it matters: Forecasting capability that used to belong to governments and supercomputers is now sitting on commodity GPUs. If your business depends on weather, supply chains, energy use or seasonal demand, you can run thousands of scenarios for the cost of one.

👉 Action: Ask your operations team: "What's one weather-sensitive decision we make every week, and could a probabilistic forecast change it?"

2️⃣ AI now forecasts world events as well as expert humans

A November 2025 paper from Bridgewater AIA Labs introduced the AIA Forecaster, an LLM system that combines agentic search, a supervisor agent and statistical calibration. On ForecastBench - the standard academic benchmark for predicting real-world events - it achieved Brier scores statistically indistinguishable from human superforecasters, the top 2% of trained forecasters in the world. On harder live prediction markets, market consensus still won. But here's the twist: a simple ensemble of the AI plus market consensus outperformed both alone.

Why it matters: This is the first credible evidence that off-the-shelf AI can match world-class human forecasters on geopolitical, economic and policy questions. The implication for business planning is significant - and the implication for combining AI with human judgement even more so.

👉 Action: Take a key prediction you're already making (next quarter's pipeline, churn, raw material prices) and ask Claude or Gemini to forecast it independently. Compare the answers. The disagreement is the most interesting data point.

3️⃣ Synthetic populations are now predicting how millions will behave

Stanford's study of 1,052 simulated people - each interviewed for two hours and turned into an AI agent - reproduced participants' answers to the General Social Survey with around 85% of the accuracy of the participants themselves. Toluna has since scaled this to over 1 million synthetic personas across 15 markets and 9 languages, each with a memory that keeps them consistent. Colgate-Palmolive published a methodology that recovers 90% of human test-retest reliability on purchase intent.

Why it matters: You can now stress-test a marketing campaign, pricing decision or product concept against a synthetic audience in hours rather than weeks. The accuracy is good enough to inform decisions, but not good enough to replace real customer conversations - so use it as a fast first filter, not a final answer.

👉 Action: Before your next campaign or product decision, ask Claude to simulate three target customer personas reacting to your concept. Treat their responses as hypotheses to test with real people, not as conclusions.

🎨 Weekend Playground

This weekend, try Manifold Markets, a free play-money prediction platform where you can bet (with virtual currency) on real-world questions across politics, business, science and culture.

It takes about two minutes to sign up. There's no real money involved, but the experience of putting a number on your beliefs - and watching the market move - is a brilliantly humbling way to practise forecasting. You can also create your own markets and invite your team.

Why this matters: The best forecasters in the world don't have crystal balls - they have calibration. They know when they're 70% sure versus 90% sure, and they update fast when new evidence arrives. This is a skill you can practise, and it's becoming a core leadership capability in an AI-rich world.

👉 Mission:

• Sign up and place virtual bets on five questions you have a view on

• Pick at least one outside your usual expertise (try a science or sports market)

• Note your confidence level for each

• Check back in seven days and see what moved

• Bonus: ask Claude to forecast the same five questions and compare

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift. If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

📰 This was the week that was...

This was the week learning stopped being something we finish, and became something we all do, all the time.

The signs were everywhere in education. Top grades are suddenly flooding US universities, with researchers linking the surge directly to the arrival of ChatGPT. An MIT fiction professor wrote movingly about catching his students using AI, and turning their confessions into a lesson about why the struggle to write is the point. A serious question is being asked out loud too: in an AI world, does the credential of the future belong to the institution, or to the individual teacher (one provocative framing doing the rounds calls it "OnlyProfessors"). AI tutors are already being trialled in classrooms, and new models like Alpha School and Startup Sherpas are showing what learner-led looks like.

The same shift is playing out well beyond the classroom. Consumers are teaching themselves to lean on AI for health worries, mortgage decisions, even weaning a baby. Business leaders picked up a new toolkit too, with Claude for Small Business launching in the US. Everyone, everywhere, is learning how to work with AI at the same time. That is a daunting thought. It is also a wonderful one.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week ✅ The big shift is in how people learn, and how your customers now seek advice - steady, not sudden ✅ Time to plan for an organisation that notices change early, not one that reacts to it late

This isn't a week for panic. It's a week for designing how your business keeps learning.

🎯 Strategic Insight

Tension: The new wave of AI tools makes it easy to hand over your work and pocket the time saved. That is tempting. It is also the moment where a lot of value quietly leaks away.

Optimistic insight: Letting AI do your work is a different thing from working with AI to do your work. The first hollows out your understanding. The second deepens it. At AI Night School we talk about the Three Jobs of an AI leader: Job 1 is to create the headspace to understand AI, Job 2 is to use AI to make room in the P&L, Job 3 is to reinvent the business. This week's tools are brilliant at Job 1 - they hand you back hours. The opportunity is to spend those hours thinking, not to stop thinking. Natural intelligence, supported by silicon intelligence.

What's shifting: The move is from outsourcing tasks to building capability. From consuming the answer AI gives you, to owning the thinking behind it. The leaders who win won't be the ones who automate the most. They'll be the ones who stay curious about their own business.

Why this matters now: Tools like Claude for Small Business make handing over work almost frictionless, and your customers are already outsourcing their advice-seeking to AI. When everyone can get the answer instantly, the advantage moves to the people who still understand why the answer is right.

👉 Takeaway: Pick one task you're tempted to fully hand to AI this month, and run it as a partnership instead:

• Let AI do the first draft or the heavy lifting

• Then spend fifteen minutes interrogating it - what did it assume, what did it miss, what would you do differently?

• Bank the time saved as thinking time, not just time off

If you'd like help designing that habit across a team, reply and we'll share how we run it in our workshops.

🤓 Geek-Out Stories

1️⃣ Your customers are already asking AI about you

Consumer behaviour is shifting fast. One in seven people in the UK have now used an AI chatbot for health advice instead of contacting their GP. Nearly a quarter have turned to AI for mortgage guidance. And three in four parents are using tools like ChatGPT for help with weaning. People know the advice isn't perfect - only 7% feel very confident in it - but they ask anyway, because it's instant and judgement-free.

Why it matters: Whatever you sell, some of your customers are forming views and making decisions through AI before they ever reach you. If you give advice for a living, AI is now in the room with you. If the advice people get elsewhere shapes what they expect from you, that's your concern too.

👉 Action: Ask a couple of AI tools the questions your customers would ask about your sector, your product, or your category. Read what they say. That's increasingly the first impression you don't control.

2️⃣ AI that checks your thinking, not just does it

Researchers at Cornell have built a decision-making tool with an unusual twist. Instead of you checking the AI's answer, the AI checks you. You rank what you value - cost, reliability, whatever matters - then the tool spots where your actual choices quietly contradict your stated priorities, surfacing the bias you didn't know was there. A human still makes the final call.

Why it matters: This is a lovely example of natural intelligence supported by silicon intelligence. The highest-value use of AI often isn't handing over the decision - it's having something rigorous hold up a mirror to how you're making it.

👉 Action: On your next significant decision - a hire, a supplier, a priority - write down your criteria first, then your gut choice. If they don't match, you've found something worth examining.

3️⃣ The plumbing of more sustainable AI

While the headlines chase models, the quiet investment is going into infrastructure. UK company Iceotope builds liquid cooling for data centres, the unglamorous "plumbing" that lets AI hardware run far more efficiently, with significant reductions in the power and water a system needs. As demand for AI grows, this is the layer that decides whether it stays sustainable - for the planet, and for the balance sheet.

Why it matters: Sustainable AI and economical AI are becoming the same conversation. The cost and footprint of running AI is turning into an operational line item, not just an environmental one.

👉 Action: Next time you scope an AI project, ask your provider a simple question: where does this run, and what does it cost to run? Efficiency is quietly becoming a competitive edge, not just a green one.

🎨 Weekend Playground

This weekend, go and learn something - with AI as your study partner rather than your ghostwriter.

Two easy places to start. Google's NotebookLM lets you upload your own documents - a report, a few articles, meeting notes - then ask questions, get summaries, even generate an audio discussion of the material. And many AI tools now have a "learn mode" or "study mode" that, instead of just handing you the answer, walks you through it like a patient tutor.

Why this matters: This is the difference at the heart of this week's newsletter, made small and playful. One mode does the work for you. The other helps you actually understand. Feeling that difference for yourself is the best way to know which one your business should lean on.

👉 Mission:

• Open NotebookLM and feed it something real from your work - a long report, a cluster of articles

• Ask it three questions you'd normally not have time to explore, and try its audio overview

• Switch on learn mode or study mode in your favourite AI tool

• Ask it to teach you one thing you've been meaning to understand, and notice how different it feels from being handed an answer

• Decide which mode deserves more room in your week

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift. If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here – designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

🎒 Exclusive early-bird: Summer Sherpas AI

If you have a teenager you want to thrive with AI, this one's for you. Summer Sherpas AI is now open for applications – a paid summer programme where teenagers learn AI skills by doing real work for real UK companies (Accenture, HMRC and Aviva have all been involved). Newsletter subscribers get early-bird access before it opens to the wider list. Disclaimer: I'm a co-founder of Sherpas AI, a social enterprise teaching teens to thrive in the AI economy.

📰 This was the week that was...

This was the week the law caught up with the spreadsheet. A Chinese court ruled it illegal to fire a worker on the grounds that AI would be cheaper, ordering a Hangzhou fintech firm to pay over 260,000 yuan in compensation. The judge's line was striking: AI hasn't reached the point where it can substantially replace human workers, so cost-cutting alone isn't a lawful reason to terminate a contract.

This is a quiet earthquake. Society has checks and balances, and legal innovation can shape technology adoption just as powerfully as the technology itself. Western jurisdictions will be watching closely, and so should you. The optimistic flip is brilliant: if cost-cutting can't carry your AI business case, you're forced to lean into the more interesting question – how does AI help us grow, reinvent and solve bigger problems?

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week

✅ A new legal signal has reshaped the risk profile of "AI as cost-cutting"

✅ Time to plan for growth-led AI adoption, not just headcount-led

This isn't a week for panic. It's a week for re-grounding your AI business case in ambition rather than reduction.

🎯 Strategic Insight

Tension: The fastest, easiest AI business case is the wrong one. Most boards approve AI investment because the numbers say "fewer people, lower cost". After this week's Chinese court ruling, that case is structurally weaker – and frankly, it always pointed in a smaller direction than the technology deserves.

Optimistic insight: Take headcount reduction off the table and a more powerful business case appears. This is exactly the third job we need to do: reinvent the business model. Natural intelligence supported by silicon intelligence isn't about doing the same job with fewer people – it's about doing bigger jobs with the same people. AI lets ambition scale where it couldn't before. That's where the real value lives.

What's shifting: The smart question is no longer "how many roles can we automate?" but "what superhuman problems can we now take on?" Things you couldn't tackle before – complex systemic challenges, market expansions, new product lines, deeper customer service – become reachable when execution capacity is no longer your bottleneck. I've been building an AI tool that maps very hard systemic problems, identifies leverage points and helps changemakers focus their efforts where they actually matter, drawing on decades of systems-thinking practice. That's the kind of capability ambition unlocks.

Why this matters now: Boards that anchor AI adoption to redundancy are about to find their case exposed on three fronts: legally, reputationally, and strategically. Boards that anchor AI adoption to growth, role redesign and organisational reinvention build a far more durable case – and one that motivates rather than threatens the people they need to make it work.

👉 Takeaway: Before your next AI investment paper, run it through three growth-side questions:

• What new revenue, product or market does this AI capability open up that we couldn't reach before?

• How does this redesign roles to make our best people more powerful, rather than fewer?

• What "too hard" problem in our business or sector could we now genuinely take on?

If you'd like help building the growth-led version of your AI business case, reply and we'll set up a call.

🤓 Geek-Out Stories

1️⃣ The rise of the FreeDreamer

A new generation of founders is emerging who treat AI as the manifestation layer for their ambition. Cofounder.co is the latest example – an agent orchestration platform that runs engineering, sales, marketing, design, finance and ops, so a single founder can focus on the dream rather than the ops. It joins Paperclip (a human control plane for managing AI employees with org charts, budgets and governance) and Bella's Strategy OS (an open source operating system for solo founders, which Kathy Walker has been using to launch Startup Sherpas). What unites them is a shift from minimising your dreams to maximising them – holding a Maximum Ambition Dream that ladders from vision to mission to goals to tasks.

Why it matters: For SME leaders, this is the stack your competitors will use to enter your market with a fraction of the headcount and twice the ambition. Understanding the FreeDreamer toolkit early gives you a map of where competitive pressure will come from – and a template for how your own teams might work in 18 months.

👉 Action: Write down your own Maximum Ambition Dream this week – the biggest version of what your business could be in three years if execution were near-free. Don't edit it. Then ask which AI capability gets you there fastest.

2️⃣ Differential: from blueprint to backlog in 8-12 weeks

Most large organisations have hundreds of candidate AI use cases and no defensible way to choose between them. Pilots get funded by enthusiasm. Governance arrives late. Value stays hazy. Differential, launched by OnePoint and written up this week on techUK by their CEO Shashin Shah, is built to fix that. It systematically discovers 150+ candidate use cases across an organisation, runs them through compliance, risk and technical-readiness gates upstream, and produces a value-ranked, risk-adjusted shortlist in eight to twelve weeks rather than nine to twelve months. Disclaimer: I'm a co-creator of Differential.

Why it matters: If your AI portfolio is a parade of pilots rather than a prioritised programme, you're paying for motion rather than progress. Differential turns "everything is possible" into "these are the right ten things to do next" – with the governance baked in, not bolted on.

👉 Action: Audit your live AI pilots this week. Count them, list their owners, and estimate their realised value. If the answer makes you wince, book a Differential demo.

3️⃣ The Imaginarium: when AI does the typing and you do the dreaming

I wrote up a piece this week on a personal experiment that's reshaped how I think about AI in the work itself, not just around it. The Imaginarium is what happened when I stopped trying to make AI fit my existing proposal-writing workflow and instead asked what version of that work would actually energise me. The answer was the imagining – walking, voice-noting, picturing what we could build together. The bit that drained me was the typing. So I separated them: I do the dreaming, a set of Claude skills and folders does the typing. What comes out the other side isn't a proposal. It's an invitation to partnership – with an interactive value model the recipient can tweak before we even begin. I'm calling this emerging discipline wishcraft – the artscience of manifesting your dreams using AI.

Why it matters: This is the real reframe of AI for SME leaders. We were taught to minimise our dreams in a scarcity economy, hence "Minimum Viable Product". The bottleneck has shifted. Execution is no longer scarce – the size of the dream you can articulate is. Start with the maximum ambitious dream, then work down to the minimum meaningful jump.

👉 Action: Pick one task on your to-do list you keep walking past. Don't push through it. Ask why it drains you, and what version of that work would energise you instead. That gap is where your own Imaginarium lives.

🎨 Weekend Playground

This weekend, try Penumbra, an AI workspace built for domain experts who are tired of re-explaining their world to a fresh chat every Tuesday.

I referenced Penumbra in this newsletter about a year ago and it's grown up beautifully. Where most AI tools forget everything between sessions, Penumbra lets you build a persistent context baseline – your terminology, your frameworks, your documents – so the AI starts every conversation already knowing your world.

Why this matters: This is what holding a Maximum Ambition Dream looks like in software. Once your context is loaded, you're no longer prompting an AI from scratch – you're reasoning together about your actual business. It's a small taste of what natural intelligence supported by silicon intelligence really feels like.

👉 Mission:

• Sign up for the free trial and create a workspace for your domain

• Upload three documents you reference often (a strategy doc, a key contract, a customer research file)

• Describe your domain to the Shape Designer in your own words

• Ask a question that would normally need three tools and a colleague

• Notice how much faster you get to a useful answer

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift. If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

Something I made this week, that I'd love to show you.

Last weekend I went for a walk through the woods around Harpenden, in the sun, talking into my phone. By the time I was back home, Harpenden.AI existed - a vision for making Harpenden the UK's first truly AI-fluent town. I did no machine work on it, only human imagination. The tools I've built in the last six months - what I've started calling my Imaginarium - did the rest.

I'm sharing it because it's a small, real proof of where the cost of doing has actually got to. A vision document, a town strategy, a council pitch and a public artefact - shaped on a forest walk. If you want to see what that looks like in practice, the page is here. If you want one for your business, your school, your trust, your team, hit reply.

📰 This was the week that was...

This was the week the question of "whose AI" stopped being theoretical. China dropped DeepSeek V4, an open-weight, MIT-licensed model with 1 million tokens of context as standard, V4-Flash priced at roughly $0.28 per million output tokens. On the same day OpenAI launched GPT-5.5, and the day after Microsoft exclusivity ended, OpenAI's models landed on AWS Bedrock. Suddenly the menu of "where can I run a serious model" went from short to long.

The geopolitical weather changed too. Beijing ordered Meta to unwind its $2 billion acquisition of Manus. France told every government ministry to plan its exit from Microsoft to Linux by autumn, covering 2.5 million civil servants. The signal is steady: the cheapest, most powerful tools are arriving fast, and where they run, on whose hardware, under whose laws is becoming a board-level question rather than a technical one.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week ✅ Frontier models are getting cheaper at the bottom, more expensive at the top, and trickier to plan against ✅ A new question is about to land on every CFO's desk: what does AI actually cost, and what's it actually worth?

This isn't a week for panic. It's a week for a conversation you may have been quietly avoiding: does our AI stack make money for us, and does our pricing match how the AI actually creates value?

🎯 Strategic Insight: the unit economics of AI just got sharp enough to answer

For two years, "what does AI actually cost, and what's it actually worth?" has been a vibe more than a number. This week the bill arrived - in two visible ways - and two serious pieces of work landed that close the gap between vibes and a real answer.

Tension: The cost of running AI workloads, after two years of being treated as a rounding error, has become a headline number. OpenAI doubled its API price for the new GPT-5.5 model that launched on Thursday: $5 per million input tokens, $30 per million output, up from $2.50 and $15. OpenAI argues the model is more token-efficient, so the effective cost rises by closer to 20% than 100%, but GitHub Copilot is launching it with a 7.5x premium request multiplier. On the other side of the market, Anthropic is so compute-constrained that since 27 March it has been quietly draining Claude session limits faster on weekday afternoons (1pm-7pm GMT). Some Max subscribers ($200/month) have reported their full daily allowance gone in 19 minutes. As Vin Vashishta puts it in The Tokenomics of Agentic Commerce, "as models scaled in size, inference costs scaled in silence." The silence has ended.

Meanwhile most SMEs are still paying for AI on a per-seat basis - the same way they pay for Microsoft 365 - while the AI doing the most useful work is no longer a tool that helps a named user but an agent that does the work autonomously while no one is logged in.

Optimistic insight: Two complementary frameworks have arrived at exactly the right moment.

On the buying side, Michael Mansard at Zuora's Subscribed Institute has just published the COMPASS framework (Choice of Optimal Metrics for Pricing Agentic Systems and Solutions). Two questions get you to the right pricing model:

• What's the AI's job? Is it a worker (always-on capacity), a service (specific tasks done), a utility (raw consumption), or a partner (delivering business outcomes)?

• How attributable is its value? Can you cleanly link what the agent does to a number on your P&L - or is it tangled up with other factors?

Those answers point to one of four pricing structures: per-agent, per-activity, per-output, or per-outcome.

On the operating side, Vashishta's piece introduces Tokenomics - the discipline of designing AI workloads that actually make money. His argument is sharper than the usual ROI hand-wave: every agent has to scale on three dimensions, not two. Reliability and utility are the ones everyone tracks. Profitability is the one that quietly kills initiatives. He cites OpenAI's decision to shut down Sora and Microsoft's pullback on free Copilot features as the same story: even the giants are hitting workloads where the maths doesn't work.

But the upside, when it does work, is significant. Macy's launched "Ask Macy's", an AI shopping assistant built on Gemini, in March. Customers who use it spend 4.75 times more than those who don't. Vashishta says his retail clients are seeing 3x to 6x lifts on similar tools - but only when they're built as proper assistants, not chatbot-shaped marketing.

What's shifting: The smart question is no longer "should we deploy AI?" but two harder, more useful ones:

• On every workload we're paying for: what's the agent's job, how attributable is its value, and does the pricing model match? (COMPASS)

• On every workload we're building or running: do reliability, utility AND profitability all scale together as we grow, or does one of them quietly fail? (Tokenomics)

This is Job 2 of the Three Jobs Framework - making room in the P&L using AI - meeting reality. Job 2 isn't "use AI to save costs". It's "use AI in places where the unit economics work, and ruthlessly stop using it in places where they don't". Most SMEs haven't yet drawn that distinction.

Why this matters now: If you only plan around today's invoice, you'll discover the real cost in the renewal. If you only plan around the upside ("Macy's got 4.75x!"), you'll deploy something that costs more to run than it earns. If you instead bring both sides of the question to the same table - what we're paying and whether the workload actually pays - you walk into 2027 with an AI stack that compounds, not one that quietly leaks margin.

And notice how this loops back to the question we started with. You can't be sovereign over a tool that doubles its price overnight, or that throttles you between 1pm and 7pm on a Tuesday. The tokenomics question and the sovereignty question are the same question, asked from two angles.

👉 Takeaway: Before your next AI contract renewal, run this thirty-minute exercise with your CFO and whoever owns operations:

• List your top three AI vendors and your top three AI workloads (the workloads might be internal, not bought-in).

• For each, name the agent's job: worker, service, utility, or partner.

• For each, mark value attribution as low, medium or high.

• For each, mark profitability honestly: is the workload generating more than it costs to run, including inference? Don't know? That's the answer.

• Identify one workload to renegotiate (using COMPASS lens), and one workload to either redesign or retire (using Tokenomics lens).

You'll know more about what your AI is actually doing - and earning - than 80% of your peers walking into the same renewal conversations.

If you'd like a hand running this on your own AI stack, hit reply.

🤓 Geek-Out Stories

(Three stories the headlines didn't bring you. The world behind the world.)

1️⃣ AI is starting to learn from physics, not just text

Nature this week covered the rise of world models - AI systems trained not on more text from the internet, but on data about how physical environments actually behave. The argument is that today's large language models are statistical pattern-matchers; world models are the beginning of something different - AI that holds an internal representation of space, time and consequence. NVIDIA's Cosmos, DeepMind's DreamerV3 family, and a wave of new robotics research are all converging on this idea.

Why it matters: This is, in plain English, AI moving from intellectualised observation (predicting the next word in a sentence) to embodied participation (predicting what happens when you push the cup). For an SME leader, the practical implication is not next year - it's three to five years out. But it's the right time to start thinking about what your business looks like when AI can model your operations, your supply chain or your customer journey as a living physical system, not as a spreadsheet.

👉 Action: Pick one part of your business that's currently modelled in a spreadsheet. Ask: if a system could simulate this as a living thing - with cause and effect, not just averages - what decisions would we make differently? File the answer somewhere you'll find it again in eighteen months.

2️⃣ Sony's table-tennis robot beat elite players this week

Sony AI's Project Ace was published on the cover of Nature on 23 April. It's the first robot to compete with elite and professional human table-tennis players in a real, regulation match. It returned spins up to 450 radians per second. It scored 16 direct points after serving against elite players, who managed eight against it. The interesting bit is not the score - it's that it perceives, plans and acts at the edge of human reaction time, in the real world, with real spin and real noise.

Why it matters: Most AI we work with lives in pure text. This is AI in a body, in a room, against a person who is also adapting in real time. That's the same physics your delivery vans, your warehouse, your service engineers, your customers actually inhabit. The cost of doing physical work is going to fall the same way the cost of doing knowledge work has - just three years behind. The leaders who notice now will quietly position before their sectors do.

👉 Action: Ask your operations lead a single question this week: "Where in our business does a person currently bridge the gap between a screen and the physical world?" Write down what you learn. That list is your physical-AI roadmap, three years early.

3️⃣ AI is being used to keep dying languages alive - and it's quietly the most optimistic story in the field

The UN reports an indigenous language disappears every two weeks. This year, research from IBM and the University of São Paulo showed that fine-tuning state-of-the-art translators with surprisingly tiny amounts of data is now producing high-quality machine translation for languages spoken by only a few thousand people. Brazilian indigenous communities are using it to build spell-checkers, next-word predictors, and writing tools in their own language. Project CETI, meanwhile, has been using AI to map sperm-whale "phonetic alphabets" - 18 rhythms, 5 tempos, optional rubato, optional ornamentation, hundreds of distinct codas - opening the question of what whales might actually be saying to each other.

Why it matters: This is what AI for good looks like when it isn't a press release. Tools optimised for the giants have a side effect: when you turn them sideways, they become unreasonably useful for the small, the rare and the threatened. The same architecture that helps a Fortune 500 do call-centre triage helps a community of 20,000 speakers keep their grandmothers' language alive. There's a quiet lesson for SME leaders here too: the best uses of AI are often the small, specific, deeply contextual ones - not the generic enterprise rollouts.

👉 Action: Find one thing in your business that has been "too small to bother automating" - a niche process, a small customer segment, a piece of institutional knowledge held by one person. Spend an hour this week asking what AI could do for just that thing. The answers tend to be surprising.

🎨 Weekend Playground

This weekend, take your phone for a walk and try Merlin Bird ID, the free app from Cornell Lab of Ornithology. Hold it up to a tree, hit Sound ID, and stand still. Within seconds it will tell you which birds are singing around you. Photo ID does the same for what you can see. It works offline. It's free. It's powered by machine learning trained on millions of recordings collected by ordinary people who love birds.

It's the most quietly profound piece of AI I know. The reviews say it better than I can - one user wrote that Merlin gave her a daily mindfulness practice she didn't know was missing.

Why this matters: This is AI doing what AI is best at - quietly, in the background, connecting a person to the living world. No screens to scroll. No prompts to engineer. No models to choose. Just you, the morning, and the birds you didn't realise were already there. Natural intelligence, supported by silicon intelligence. As lived practice.

👉 Mission:

• Download Merlin before Saturday morning

• Walk somewhere green - your garden, your local park, the woods, the river

• Stand still for five minutes, hit Sound ID, and just listen

• Notice what was there the whole time

• Bonus: do it before you check your phone for any other reason on Saturday

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift. If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

📰 This was the week that was...

This was a week of big moves, and a simple message underneath them. Anthropic launched Claude Design, a new way to create prototypes, slides and one-pagers by chatting with Claude - Figma's share price dropped 7% in hours. Beijing's Moonshot AI open-sourced Kimi K2.6, a model built to run autonomously for hours on complex tasks. And today OpenAI rolled out GPT-5.5, pitched as "a new class of intelligence for real work".

At the same time, Anthropic's own customers started noticing their bills creeping up - same headline price, new tokeniser, more tokens per request. The message for SME leaders cuts through all of it: move forward step by step, focused on unlocking real value from careful, measured adoption. Don't chase the launches. Don't panic about the pricing wobbles. Choose one workflow, improve it, measure what you saved.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week ✅ Foundation model launches are becoming routine - expect one a month ✅ Pricing is shifting quietly beneath headline rates - worth a look at your monthly invoice

This isn't a week for panic. It's a week for step-by-step value capture - one workflow at a time.

🎯 Strategic Insight

Tension: The label "Artificial Intelligence" does us no favours. It conjures Terminators, dystopian futures and replacement anxiety. It makes the technology feel foreign, threatening and hard to comprehend. And that means business leaders are being asked to make big decisions about something their nervous system is telling them to fear.

Optimistic insight: I spent last weekend on retreat at The Dreaming in the Welsh mountains, and one thought kept surfacing: we've named this technology badly. What we actually have, in our hands today, is closer to Easy Magic. Do you want easy magic? Of course. Do you want your ideas to come to life in minutes instead of months? Yes. Natural intelligence supported by silicon intelligence - that's what this is. It's a creative amplifier, not a replacement. The label should match the lived experience.

What's shifting: The teams getting the most value right now are the ones who've stopped treating AI as a scary new technology and started treating it as a useful new capability. They describe what they do in human terms - "it drafts my proposals", "it tidies my data", "it writes my first pass". The mystique falls away. The value becomes obvious.

Why this matters now: If you only plan for "rolling out AI tools", you'll likely hit internal resistance, vague scepticism and quiet non-adoption. If you instead plan for "giving people easy magic that saves them hours", you get three benefits: faster uptake, clearer ROI, and teams who actually enjoy the change. Language shapes decisions. Leaders who reframe this shape adoption.

👉 Takeaway: This week, try the language experiment with your team:

• Stop saying "AI" internally for a fortnight. Replace it with what the tool actually does ("the drafter", "the summariser", "the researcher").

• Ask your team what they'd like their "easy magic" to be - one task they dread that they'd love to hand off.

• Pick the top three answers and find out which tool already does each one well.

• Measure the time saved over four weeks. Share the numbers internally.

If you'd like a structured session to run this with your leadership team, reply and we'll send you our half-day workshop outline.

🤓 Geek-Out Stories

1️⃣ AI needs a sense of smell

A thoughtful piece in Noema argues something most of the industry has quietly ignored: that real intelligence might need olfaction. Between 2015 and 2025, research papers on machine smell stayed flat while vision, language and audio grew exponentially. But smell is our third-largest sensory input by volume, and it's how organisms navigate, detect disease, and build memory. Researchers like Kordel France are now calling for a standardised data format for smell - a "JPG for odours" - to finally let AI learn the world the way living things do.

Why it matters: This is a reminder that the "intelligence" in AI today is narrow. It's very good at text, decent at images, increasingly good at sound - and completely blind to taste, smell, touch, and embodied experience. Knowing where the gaps are is as strategically useful as knowing what the technology can do. It keeps you honest about what to automate and what to keep human.

👉 Action: Next time you're asked whether AI can replace a role, list the senses that job actually uses. If smell, touch or embodied judgement are central (think hospitality, healthcare, food, care work), the honest answer is "not anytime soon".

2️⃣ A frugal-AI breakthrough: 100x less energy, better accuracy

Researchers at Tufts University published a paper earlier this month showing a neuro-symbolic approach that cuts AI energy use by up to 100x while actually improving accuracy on structured, long-horizon tasks. The trick: combining neural networks (pattern recognition) with symbolic reasoning (rules and logic), so the machine doesn't have to brute-force its way through every problem. It's the clearest signal yet that "bigger model = better" is not the only path forward.

Why it matters: Frugal AI is moving from academic interest to strategic necessity. As training costs rise and regulators sharpen their focus on energy use, efficiency is becoming a genuine competitive advantage. Smaller, more targeted systems can often do the job of a giant model at a fraction of the cost and footprint.

👉 Action: For each AI use case in your business, ask: "What is the smallest, most specific model that would solve this well?" The answer is rarely the most famous one.

3️⃣ AI as a quiet ally for nature

CBC reported last week on a growing set of conservation projects where AI is doing the unglamorous work humans can't: watching thousands of hours of salmon migration footage to count populations, mapping Canada's vast wilderness from satellite imagery, tracking species recovery in near-real-time. Canada's new national nature strategy explicitly names AI as part of the toolkit. Similar work is underway in UK peatland restoration and marine biodiversity monitoring.

Why it matters: The best AI stories aren't always the loudest ones. This is AI doing patient, repetitive, valuable work at a scale humans physically can't - and it's a reminder of what "AI for good" actually looks like in practice. For leaders thinking about sustainability reporting, ESG, or community impact, there's a credible story to tell.

👉 Action: Ask your team one question: "Where are we drowning in data we can't review fast enough?" That's often where AI creates its quietest, biggest wins - and where an ESG or impact narrative may already be hiding.

🎨 Weekend Playground

This weekend, try Claude Design, Anthropic's new visual creation tool - it turns a chat prompt into polished slides, mockups and one-pagers in minutes.

You'll need a Claude Pro, Max, Team or Enterprise subscription (no free tier for this one). Once you're in, click the palette icon in the left-hand sidebar to start a project. Upload a sketch, a reference image, or just describe what you want.

Why this matters: This is a direct taste of what "easy magic" feels like. No training, no design background, no waiting for a creative brief to come back. A short conversation, a polished output. It's the clearest example this week of where the labels fall short - this isn't artificial anything, it's useful magic.

👉 Mission:

• Rebuild one slide from your most recent deck - ask it to improve the visual design

• Prototype a landing page for an idea you've been sitting on

• Create a one-pager explaining your business to a new hire

• Make a simple infographic from last quarter's results

• Show it to someone at home and watch their face

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift. If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

A quick note from Hugo:

If your team is feeling the pace of AI and wondering where to start, I run workshops that don't focus on AI at all. We use the "superpower" framing - telepathy, teleportation, time travel - to unlock what problem is actually worth solving, then pair it with AI. Teams leave with momentum and a plan.

If you'd like your team to think bigger, get in touch.

Shoutout to my friend John Mulholland for the superpower framing.

📰 This was the week that was...

You could feel the pace of change accelerating this week. Three major launches landed on the same day, and each one tells you something different about where business is heading.

Anthropic released Claude Opus 4.7, a step-change in long-running agentic work - the kind of model that quietly does hours of senior engineering without supervision. Canva unveiled Canva AI 2.0, reimagining itself as a conversational, agentic system at the centre of how work gets done, with Slack, Gmail, Calendar and more baked in. And OpenAI gave Codex control of your desktop, so its coding agent can now open and operate apps on your Mac while you get on with something else.

What's in it for a UK business leader? The tools your teams used yesterday are measurably more capable today. The ones that were "nice to have" a quarter ago are becoming the place where work actually happens. And the gap between "what's possible" and "what we're doing" is widening faster than most plans assume.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week ✅ Capability is accelerating - plan for the tools your team has in 90 days, not today ✅ Time to budget for ambition, not just adoption

This isn't a week for panic. It's a week for sitting with a bigger question: who are we becoming?

🎯 Strategic Insight

Tension: The pace of capability is accelerating, and yet the people who most need to believe AI is for them increasingly don't. New exclusive data from Startup Sherpas (shared below) shows girls in the UK's most deprived schools approaching "strongly disagree" that AI will create more opportunities than it takes away. At the same time, Canva's AI 2.0 launch makes a point worth pausing on: even as their platform becomes fully agentic, they insist humans remain at the heart of design for most organisations.

Optimistic insight: The acceleration is real, and so is the choice about what to do with it. The cost of doing is collapsing to near zero. As execution becomes abundant, it becomes commoditised. What rises sharply in value is why you exist, who you serve, and what you stand for. Natural intelligence, supported by silicon intelligence, is the competitive lens - not silicon on its own. The leaders who'll compound advantage from here are the ones who use the speed to create more room for judgement, creativity and connection, not less.

What's shifting: The smart question is no longer "which tool should we buy?" but "who on our team - and in the communities we serve - needs to believe this is for them?" AI is becoming the substrate, not the stand-alone product. That makes belief, confidence and access the new bottleneck.

Why this matters now: We've been writing about the three jobs for a while - create headspace, make room in the P&L, and reinvent the business. This week is squarely a Job 3 week. If you only plan for Job 2 (tool adoption, cost take-out), you'll likely discover the pipeline of AI-ready talent is narrower than you thought, and the people you most want to develop feel least included. If you plan for inclusion alongside reinvention, you get three things: a deeper talent pool, a team that raises rather than lowers its ambition, and a brand that stands for something.

👉 Takeaway: Between now and summer, pick three practical moves:

• Invite one underrepresented group inside the business to co-design an AI workflow that matters to them

• Pair every AI tool rollout with a clear "what's in it for you" story for the people using it

• Set an explicit ambition target for AI usage, not just a cost target

If that sparks a question you're sitting with, reply and tell me - I'll come back to you personally.

🤓 Geek-Out Stories

1️⃣ Canva AI 2.0: the agentic platform that still puts humans at the heart

Canva has announced its biggest evolution since launching in 2013. Canva AI 2.0 brings conversational design, agentic orchestration, living memory across your projects, and connectors into Slack, Gmail, Zoom, Notion, Google Drive and Calendar - plus a deepened partnership with Anthropic that lets you move Claude artifacts straight into Canva to edit and publish. It now serves more than a quarter of a billion people every month, and their in-house models run up to 7x faster and 30x cheaper than comparable frontier alternatives.

👉 Read the announcement

Why it matters: For any SME that runs on decks, social, sheets and internal comms, Canva is about to stop being "a design tool" and start being "the place work happens". The signal underneath the feature list: Canva explicitly positions humans as still at the heart of design. That's a permission slip for your teams to lean in confidently without worrying they're losing the craft.

👉 Action: Nominate one team - marketing, HR, sales ops - to pilot Canva AI 2.0 for 30 days. Ask them to log three things weekly: hours saved, quality gained, and where a human had to step in. That last column is where your real competitive advantage lives.

2️⃣ Exclusive: Startup Sherpas data shows who AI is - and isn't - reaching

Startup Sherpas surveyed 70 young people at the start of their AI programmes. Students at independent schools hold a net-positive view of AI at +0.51 on a 5-point scale. Students at the most disadvantaged school in the study - Walkers Riverside Academy in Newcastle - scored -1.61. Girls in disadvantaged state schools scored -1.41 net, the lowest in the dataset. When asked whether "AI will create more interesting opportunities than it takes away", WRA students averaged 1.50 out of 5 - approaching "strongly disagree" as a group.

👉 Contact Startup Sherpas for the full report

Why it matters: This is the pipeline of AI-confident young people you'll be hiring from over the next decade, and it's narrower than it looks. It skews male. It skews affluent. If your talent, product or customer strategy assumes broad AI enthusiasm among the next generation, the data says you'll miss the majority. The good news: this is a fixable perception gap.

👉 Action: If you run a schools partnership, apprenticeship scheme or early-careers programme, ask one question this quarter: "what are we doing specifically to make girls from disadvantaged backgrounds feel AI is for them?" If the answer is "nothing targeted", that's your next investment.

3️⃣ AI optimism travels further than you think: Nepal and Sri Lanka

A reporter's notebook from Computerworld cuts against the Western AI conversation in a way worth reading. In Kathmandu and Colombo, the chatter isn't hallucinations, rogue agents or data leaks - it's empowerment, upward mobility and women-led entrepreneurship. Local founders, without the burden of legacy systems or bloated IT budgets, are building AI-native from day one. Free-tier tools and open-source small language models mean, in the words of Momo VC's Preeti Adhikary, "AI is truly democratised now."

👉 Read the notebook

Why it matters: Two stories run alongside each other at the moment. One says AI concentrates power in a handful of US hyperscalers. The other says AI lowers the bar for anyone, anywhere, to build something. Both are true, and the second is the one most underweighted in UK boardrooms. Your competition is no longer just the incumbent down the road - it's the AI-native founder in Colombo with no technical debt and nothing to unlearn.

👉 Action: Ask one thing in your next strategy meeting: "if we were starting our business today, in a market with no legacy, what would we build?" Then listen for which of your current constraints are actually habits.

🎨 Weekend Playground

🧪 Take Opus 4.7 on your "I'll get to it" pile

Claude Opus 4.7 landed today with a meaningful step up in long-running agentic work. Early testers report it handles the kind of tasks that previously needed close supervision - and its vision has more than tripled in resolution.

Pick the one thing that's been sitting on your "I'll get to it" pile for months. The messy spreadsheet. The half-written strategy. The policy doc that needs a refresh. The diagram you keep meaning to decode. Hand it to Opus 4.7 on Saturday morning with a cup of coffee, and see how far it gets before you finish it.

Why this matters: The best way to recalibrate your sense of what AI can do is to give it the task you assumed it couldn't. This isn't a productivity hack. It's a calibration ritual.

👉 Mission:

• Pick the one task you've been avoiding because it's "too fiddly"

• Paste it into Opus 4.7 with the context you'd give a new hire

• Ask it to tell you what's unclear before it starts

• Review the output the way you'd review a junior colleague's - give feedback, not rewrites

• Notice what surprised you. That's where your 2026 leverage lives.

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift.

And here's the question I'm genuinely curious about this week: what AI question are you holding right now? The one you haven't quite asked anyone, or the one you've asked and not had a good answer to. Reply and tell me - I read every message and I'll come back to you personally.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

🎓 Learn how to use AI to save 1 day a week in just 6 weeks: next cohort starts Tuesday 14th April

The next cohort of the AI Leaders Fellowship kicks off Tuesday 14th April. It's where SME leaders come to accelerate their AI skills, build their own AI strategy, and learn the core practical skills that save you one days a week - every week. If you've been meaning to get properly fluent rather than dabbling around the edges, this is the one.

👉 Grab your seat here

📰 This was the week that was...

This was the week we got a preview of AI's next chapter - and it's less about raw power, more about foundations. Anthropic unveiled Claude Mythos Preview, a model that can autonomously find and exploit software vulnerabilities across every major OS and browser - and in the same breath launched Project Glasswing, bringing Amazon, Apple, Google, Microsoft and Cisco together to harden critical software before attackers get similar tools.

At the same time, Google DeepMind shipped Gemma 4, putting frontier-class intelligence onto phones that work entirely offline, and Meta's Superintelligence Labs entered the race with Muse Spark, a multimodal "personal superintelligence" trained with over 1,000 physicians. Meanwhile the plumbing got easier: Claude Managed Agents handles sandboxing, sessions and permissions so teams like Notion, Rakuten and Asana ship agents in days rather than months.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week

✅ Infrastructure foundations shifting beneath the capability headlines

✅ Time to think about sovereignty and security, not just performance scores

This isn't a week for panic. It's a week for mapping your AI foundations.

🎯 Strategic Insight

Tension: It's tempting to focus on which AI provider offers the smartest responses or highest accuracy scores. Most procurement conversations still centre on feature comparisons and capability demos, rather than whether you have the foundations to use AI effectively when business continuity depends on it.

Optimistic insight: The smart money is moving to foundations: security, sovereignty, and systems that work when connectivity fails. Project Glasswing shows that leading technology companies are thinking defensively first - securing software before offensive capabilities proliferate. Gemma 4's edge focus demonstrates that the future of AI isn't just bigger models in bigger data centres, but capable models running where you need them. And Claude Managed Agents shows the same pattern one layer up: the winning question is no longer "can we build an agent?" but "can we run it safely, reliably, and at scale?"

What's shifting: The conversation is moving from "Which AI gives the best answers?" to "Which approach gives us control when it matters most?" Between on-device models, security-hardened deployment, managed agent runtimes and proper governance frameworks, the organisations that thrive will be those that built robust foundations rather than chased the latest capability demos.

Why this matters now: Anthropic's own research shows AI capabilities in cybersecurity have reached a threshold where vulnerabilities found by the model often survive decades of human review and millions of automated tests. The same acceleration that makes AI powerful also makes security and sovereignty urgent, not optional.

👉 Takeaway: Between now and Q2, audit your AI foundations across three dimensions:

• Security: How quickly could you patch vulnerabilities if AI-powered attacks accelerate?

• Sovereignty: Which of your AI use cases need to work without internet connectivity or external API dependencies?

• Governance: Where do you need human oversight that can't be bypassed by social engineering or prompt injection?

If you'd like help stress-testing your AI strategy against these shifts, book a call and we'll work through it.

🤓 Geek-Out Stories

1️⃣ Keeping AI agents on a leash

AWS published a detailed guide for controlling which websites your AI agents can access, responding to enterprise customers who need domain-based filtering and egress control for agents that browse the web. The approach uses network controls to block unauthorised destinations while maintaining audit logs for compliance reviews.

Why it matters: This addresses a core business risk with AI agents - you want them to access enough to be useful, but not so much that they become a liability. Imagine an agent meant to research competitors accidentally sharing sensitive data with external sites, or being tricked into visiting malicious domains.

👉 Read it here: https://aws.amazon.com/blogs/

👉 Action: Review any AI systems you use that access external websites or APIs. List which domains they genuinely need versus which should be blocked.

2️⃣ Practice makes perfect - even for AI

Research published in Communications of the ACM argues that user simulation - creating virtual customers that interact with AI systems - is a critical catalyst for overcoming bottlenecks in AI development. The technique enables user behaviour modelling, synthetic interaction data, and system evaluation at scale.

Why it matters: Instead of learning from real customer interactions (and potentially frustrating them with early mistakes), you can simulate thousands of realistic sessions to test and improve your AI systems. Faster improvement cycles, without risking actual customer relationships.

👉 Read it here: https://cacm.acm.org/opinion/the-indispensable-role-of-user-simulation-in-the-pursuit-of-agi/

👉 Action: Pick one customer journey where you'd like to test AI improvements. Consider whether virtual customer testing could help you validate changes before deploying to real customers.

3️⃣ AI that works without Wi-Fi

Developer Fikri Karim used Google's new Gemma 4 model to build an English learning app that looks at objects around you, hears conversations, and provides real-time language coaching - all running entirely on a phone with no internet connection. The app shows how frontier AI capabilities are now available for local deployment on everyday devices.

Why it matters: This solves fundamental business challenges around connectivity, data privacy and service reliability. UK businesses in farming, construction, field services, or any operation away from reliable internet can now consider AI applications that work regardless of signal strength or data centre availability.

👉 Read it here: https://deepmind.google/

👉 Action: Identify three business processes that happen in locations with poor connectivity. Consider whether locally-running AI could enhance these workflows without requiring infrastructure investment.

4️⃣ Giving your AI a memory of its own

Milla Jovovich has released MemPalace, an open-source system that lets your AI remember every conversation you've ever had with it - across tools, across sessions, across models. A separate project from Andrej Karpathy takes a different angle: a persistent wiki that compounds knowledge as you use it, so every new conversation builds on what came before rather than starting from scratch.

Why it matters: As you lean harder on AI, the frustration of re-explaining yourself every single session becomes the single biggest drag on productivity. Persistent memory is also the unlock for more sophisticated agent workflows - an AI that forgets can't plan, can't learn your preferences, and can't compound value over time. This is a hot space precisely because whoever solves it well changes the ceiling on what AI can do for you.

👉 Read them here: https://github.com/milla-jovovich/mempalace https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f

👉 Action: Notice the next three times you have to re-explain context to an AI tool. That friction is telling you where persistent memory would pay you back fastest.

🎨 Weekend Playground

This weekend, pick your path. If you've got an Android, try the Google AI Edge Gallery app from the Play Store - it lets you run advanced AI models entirely on your phone with no internet connection required. Experiment with processing text, images and audio locally. The real insight isn't just that it works offline - it's that you can have private conversations with frontier-class AI without sending any data to external servers.

If you're on iPhone or would rather stay in the browser, spend an hour with Claude instead. Push it on something real from your working week - a messy decision, a document you need to make sense of, a strategy you're wrestling with - and see how far a proper thinking partner can take you.

Why this matters: Both paths point at the same shift - from AI-as-a-service to AI-as-infrastructure, and from AI-as-novelty to AI-as-thinking-partner. One puts the model on hardware you control. The other puts a frontier model to work on the problems actually on your desk.

👉 Mission:

• Android: download the Google AI Edge Gallery app and try the image analysis feature with a photo from your camera roll

• Android: ask it to explain a complex document without any internet connection

• Claude: bring it the hardest decision on your plate this week and ask it to stress-test your thinking

• Claude: upload a document you've been meaning to read and ask for the three things that matter most

• Either path: notice where on-device or frontier intelligence could quietly enhance your business operations

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift.

If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

📰 This was the week that was...

This was the week the AI world went quiet - and that's fine. It's Easter. The inbox is lighter. The calendar is thinner. The news cycle slowed to a murmur. If you're reading this with a cup of tea and nowhere urgent to be, good. That's exactly where you should be.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week

✅ The AI landscape hasn't changed since last Friday - it will still be there on Tuesday

✅ Time to think, not time to react

This isn't a week for panic. It's a week for a long walk and a clear head.

🎯 Strategic Insight

Tension: We talk a lot about the three hardest jobs in AI adoption: creating headroom in the diary to understand AI, making room in the P&L to fund it, and finding the courage to reinvent the business around it. But there's a job that comes before all three - and it's the one most leaders skip.

Optimistic insight: Making headspace. Not headroom. Not budget. Headspace. The quiet, unscheduled thinking time where strategy actually forms. Most leaders are so busy reacting to AI news, trialling tools, and fielding vendor pitches that they never sit still long enough to ask the only question that matters: what do we actually want AI to do for this business?

What's shifting: The leaders pulling ahead aren't the ones consuming the most AI content. They're the ones who've carved out space to think clearly about what they've already learned. The shift is from "keeping up" to "making sense".

Why this matters now: Easter gives you something the calendar rarely offers: four unscheduled days. That's not a gap. That's a gift. The leaders who use this weekend to think - not scroll, not trial, just think - will come back sharper than the ones who spent it catching up on demos.

👉 Takeaway: Before Tuesday, ask yourself three questions:

• What have I already learned about AI that I haven't yet acted on?

• Where is AI already saving us time or money that I could redirect?

• If I could only make one AI bet this year, what would it be?

Write the answers down. On paper. No chatbot required.

🤓 Geek-Out Stories

1️⃣ Google's "Pied Piper moment": AI memory compression just got serious

Google Research unveiled TurboQuant, a compression algorithm that could shrink AI's working memory by at least 6x without losing accuracy. The internet immediately nicknamed it "Pied Piper" after the fictional startup in HBO's Silicon Valley. It's still a lab breakthrough - not deployed anywhere yet - but Cloudflare's CEO called it Google's DeepSeek moment: proof that efficiency gains are becoming just as important as raw power.

Why it matters: For UK business leaders, this is another signal that the cost of running AI is heading down, fast. You don't need to understand vector quantisation. You need to understand that the models you're paying for today will cost a fraction to run tomorrow. Plan accordingly.

👉 TechCrunch has the full story

2️⃣ Anthropic accidentally revealed its most powerful model - twice

A misconfigured data store left nearly 3,000 unpublished Anthropic documents publicly accessible, including a draft blog post describing a new model called Claude Mythos (also codenamed Capybara). Anthropic confirmed the model exists, calling it a "step change" in capability - larger and more powerful than anything in its current lineup. Days later, the full source code for Claude Code was accidentally published to npm. Cybersecurity stocks dipped. Anthropic is now testing Mythos with a small group of enterprise customers focused on security applications.

Why it matters: Two things for business leaders. First, the next generation of AI models is coming and it will be meaningfully more capable, particularly in coding and cybersecurity. Second, even the companies building frontier AI make basic security errors. If Anthropic can misconfigure a data store, so can your team. Check your own house.

👉 Fortune broke the original story

3️⃣ "The blank page is not deep work" - a Chief AI Officer builds his own AI memory

Eric Porres, Chief AI Officer at Logitech, published a sharp response to recent essays by Ezra Klein and Cal Newport arguing that AI degrades our ability to think. Porres' counter: the bottleneck in knowledge work isn't generating new ideas - it's finding the ideas you already had. He built a semantic memory layer across 4,485 AI conversations on 12 platforms, searchable by meaning in under a second, for less than £20 a month. His argument is simple: the blank page isn't sacred. It's just all we had before.

Why it matters: This is what "own your AI" looks like in practice. Porres isn't waiting for a platform to solve cross-tool memory. He built it himself, on his own infrastructure, under his own control. Even if you never build a vector database, the principle matters: your AI conversations are an asset. Treat them like one.

👉 Read the full piece on Substack

🎨 Weekend Playground

This Easter, try NotebookLM - Google's free AI research tool that turns your documents into genuinely useful learning formats.

Upload something you've been meaning to properly get your head around - a strategy document, a white paper, an industry report gathering dust in your downloads folder. Then ask NotebookLM to explain it back to you in different ways: an Audio Overview (a surprisingly good AI podcast), a set of flashcards, a mind map, a quiz, or even a short video explainer. The newest formats include Debate and Critique modes, where two AI hosts argue the merits of your source material from opposing sides.

Why this matters: We all learn differently, but most of us default to reading because that's how information arrives. NotebookLM lets you discover which format actually makes things stick for you. That's useful self-knowledge - and the long weekend is the perfect time to find it.

👉 Easter mission:

• Upload one document you've been meaning to digest

• Generate an Audio Overview and listen on a walk

• Try the Debate format and see if argument sharpens your understanding

• Create flashcards on the key points and test yourself on Easter Monday

• Ask yourself: which format helped me learn fastest?

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift. If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

🎓 Want to help your team respond to AI with confidence? Hugo runs hands-on AI strategy workshops for leadership teams - no jargon, no hype, just clarity on what to do next. Book a conversation to explore what that looks like for your team.

📰 This was the week that was...

This was the week AI stopped trying to do everything - and started doing more of what matters.

OpenAI shut down Sora, its video generation app, just six months after launch. The reason? Compute costs. The company is cutting expensive side projects ahead of an IPO and refocusing on business tools and coding. Disney's billion-dollar partnership deal collapsed with it. That's a loud signal: even the biggest players are learning that "build everything" is not a strategy.

Meanwhile, Anthropic moved in the opposite direction. Claude can now control your computer - opening apps, navigating browsers, filling in spreadsheets - and you can assign those tasks from your phone using a new feature called Dispatch. And their latest Economic Index report showed something striking: experienced AI users get a 10% higher success rate than newcomers. The gap is growing. AI fluency is becoming a competitive advantage, not a nice-to-have.

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week

✅ The competitive landscape just shifted - the winners are narrowing their focus, not widening it

✅ Time to invest in AI fluency across your team, not just your most technical people

This isn't a week for panic. It's a week for asking: "Are our people getting better at this, or just busier?"

🎯 Strategic Insight

Tension: The most powerful AI tools are now arriving faster than large organisations can adopt them. Claude can control your computer from your phone. It can read your emails, build your reports, manage your tasks - all while you're on the train. But most large organisations can't touch any of this. Data residency policies, cybersecurity constraints, AI governance frameworks, device management rules - they all say no.

Optimistic insight: Here's where it gets interesting. Those same organisations are cutting headcount to fund AI transformation. But that headcount carries tacit knowledge - the "acceptable vagueness" that made processes work. Once those people leave, they take that knowledge with them. And many of them are becoming founders. As startups, they can adopt tools like Claude Dispatch immediately. They can codify what they know, build AI-native alternatives, and compete at a fraction of the cost. The very organisations that let them go are now competing against them - while locked into enterprise platforms that deliver a fraction of the capability.

What's shifting: The AI capability gap is no longer between "companies that use AI" and "companies that don't." It's between organisations that can adopt fast and those constrained by their own architecture. Small, well-led teams with the right tools are pulling ahead. Anthropic's latest Economic Index confirms this: longer-tenured users don't just use AI more - they use it better, on harder tasks, with higher success rates. AI fluency compounds.

Why this matters now: If your organisation is shedding talent while restricting what AI tools your remaining team can use, you're creating competitors and handicapping yourself at the same time. The smart question is no longer "How do we reduce headcount with AI?" but "How do we create the conditions where our best people - inside and outside the organisation - can do their best work with the best tools?"

👉 Takeaway: Between now and Q3 2026, do three things:

• Audit which AI tools your teams can actually access today versus what's available on the market

• Design a partnership or incubation strategy for the startups being created by your former people

• Invest in AI fluency training - not one-off workshops, but ongoing capability building. The Anthropic data is clear: the more your people use AI well, the more value they create

If you'd like help mapping this out for your leadership team, reply and we'll set up a call.

🤓 Geek-Out Stories

1️⃣ Deep Memory: Search every AI conversation you've ever had

If you've been using ChatGPT, Claude, or Cowork for months, you've probably had this experience: you know you discussed something brilliant three months ago, but you can't find it. Deep Memory is a single Python file - no dependencies, no database, no AI required - that indexes and searches across all your AI conversation exports in under a second.

Why it matters: This is a perfect example of frugal design. The problem isn't that you need a fancier AI to find your old conversations. You need a simple search tool. One Python file, zero cost, problem solved. That's the mindset every team should be applying.

👉 Action: Export your ChatGPT and Claude conversation history this week and try it. Then ask your team: "Where else are we reaching for AI when a simpler solution would do the job?"

👉 github.com/ericporres/deep-memory

2️⃣ The AI Vampire: Why your best AI users might be burning out

Eric Porres, Chief AI Officer at Logitech, shares his Whoop sleep data showing a drop from 78% to 70% sleep performance in eight weeks - directly correlated with getting good at Claude Cowork. His morning had eight automated briefings firing before breakfast. His CLAUDE.md configuration file ran to 500+ lines. He calls it "Jevons Paradox applied to cognition" - AI efficiency didn't reduce his workload, it expanded his ambition until it became unsustainable.

Why it matters: This is the conversation I'm having most often right now. The people using AI best are the ones most at risk of overwhelm. And the people earlier in the journey are feeling FOMO and pressure. Neither state is productive. Designing your AI working world purposefully - with boundaries and clear intent - is becoming a leadership skill.

👉 Action: Ask your AI-forward team members this week: "What have you automated that you haven't actually stopped doing manually?" The overlap is where burnout hides.

👉 promptedbyeric.substack.com

3️⃣ Cloud vs self-hosted AI: A practical guide for 2026

A clear-headed comparison of when to use cloud AI APIs versus hosting your own models. The crossover point lands between 100,000 and 300,000 monthly requests. Below that, APIs win on cost. Above it, self-hosting flattens out while API costs keep climbing. For regulated industries - finance, healthcare, government - data residency requirements often make the choice for you.

Why it matters: As AI moves from experiment to critical business system, these design choices matter enormously. Digital self-sovereignty - owning your AI infrastructure and the data that flows through it - is something we teach at AI Night School because it's a strategic decision, not just a technical one. Getting this right early saves you from painful migrations later.

👉 Action: If you're running more than 50,000 AI requests per month, put "cloud vs self-hosted review" on your Q2 agenda. Start with one high-volume use case and run the numbers.

👉 dev.to - Cloud vs Self-Hosted AI

🎨 Weekend Playground

This weekend, try Claude Dispatch in Cowork - a new feature that lets you assign tasks to Claude from your phone while it works on your computer.

You'll need a Claude Pro (£16/month) or Max plan, the Claude desktop app running on your Mac, and the Claude mobile app on your phone. Setup takes about two minutes - open Cowork, click Dispatch, scan a QR code, and you're paired.

Why this matters: This is what "AI as a co-worker" actually feels like in practice. You text an instruction from your phone. Claude does the work on your computer. You come back to finished output. It's early - it won't always nail complex tasks - but it's a genuine glimpse of how AI-assisted work is going to feel in 2027.

👉 Mission:

• Set up Dispatch and pair your phone with your desktop (2 minutes)

• Try a simple task from your phone: "Summarise the last 10 files I downloaded"

• Try something useful: "Check my calendar for Monday and draft a one-paragraph briefing for each meeting"

• Try something fun: "Find a recipe for what's in season this week and create a shopping list"

• Notice how it feels to walk away and come back to finished work

👉 Claude Dispatch setup guide

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift. If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

🎓 Want to give your leadership team the headspace to think clearly about AI?

Our AI strategy workshops use a "superpower framing" - your team solves real business problems by imagining superpowers, not wrestling with technology. It works. Teams leave with a clear, prioritised action plan they actually own.

👉 Book a free discovery call and let's find your starting point.

📰 This was the week that was...

This was the week that AI stopped being about adding a tool and started being about redesigning the whole company.

European AI lab Mistral launched Forge, a system for enterprises to train frontier-grade AI models on their own proprietary data - partners already include ASML, Ericsson, and the European Space Agency. And Manus (now owned by Meta) released My Computer, a desktop app that brings AI agents out of the cloud and onto your laptop - reading files, controlling apps, running multi-step tasks locally.

The direction is clear. AI is moving from something you log into, to something that lives inside your business - in your data, on your machines, embedded in your workflows. The question for leaders is not "should we use AI?" but "if we were designing this company today, would we design it the way it currently exists?"

Let's get into it.

🔥 Urgent Priorities

✅ No fires to fight this week

✅ The shift from "AI as tool" to "AI as operating model" is accelerating - this is strategic, not tactical

✅ Time to move from Job 1 (understanding AI) to Job 3 (reinventing the business)

This isn't a week for panic. It's a week for sketching what your company looks like if you started it tomorrow.

🎯 Strategic Insight

Tension: There are three jobs every business needs to do with AI. Job 1: create the headspace to understand it. Job 2: make room in the P&L by using it. Job 3: reinvent the business itself. Most organisations are still on Job 1 or Job 2 - and that's fine. But this week's news makes it clear that the leaders pulling ahead are already on Job 3. If you were designing your company today, would you design it the way it currently exists?

Optimistic insight: Stuart Winter-Tear's Unhyped AI newsletter introduced a phrase worth borrowing: "tolerated vagueness". Most firms survive on it - ownership is supposedly clear, escalation paths are supposedly known, exceptions are supposedly handled. People absorb what the formal system never resolved. AI exposes all of it. That's uncomfortable. But it's also an extraordinary opportunity. Because fixing those gaps doesn't just make your AI work better - it makes your whole organisation work better.

What's shifting: The smart question is no longer "How do we bolt AI onto our existing structure?" but "What would this company look like if we designed it today?" Fast Company argues that AI doesn't just automate tasks - it exposes how much of an organisation's structure existed to compensate for inefficiency, fragmentation, and internal inertia. That's a mirror most boardrooms haven't looked into yet.

Why this matters now: If you only plan for "AI as a productivity layer", you'll likely optimise a structure that's already outdated. If you instead plan for "AI as redesign catalyst", you get three benefits: a leaner cost base, faster decision-making, and a business that's genuinely hard to compete with.

👉 Takeaway: Block 90 minutes with your leadership team this month and ask three questions:

• If we were founding this company today with AI available from day one, what would we build differently?

• Where in our business does "tolerated vagueness" slow us down most?

• Which of our processes exist because coordination was expensive - and what happens when it becomes cheap?

If you'd like help facilitating that conversation, reply and we'll set up a workshop.

🤓 Geek-Out Stories

1️⃣ LumberChunker: every frustration you have with AI is being worked on, every day

Researchers at Carnegie Mellon published LumberChunker, a new method for splitting long documents into meaningful chunks for AI retrieval. Instead of chopping text at arbitrary points, it uses a language model to find where the narrative actually shifts - producing chunks that preserve context and meaning. In benchmarks, it significantly outperformed every other approach tested.

Why it matters: If you've ever asked an AI a question about a long document and got a vague or wrong answer, this is why. The way AI systems break up your documents before searching them has been a quiet bottleneck. LumberChunker is one example of a much bigger pattern: every frustration we have with AI today is being actively worked on. The tools are getting better, fast. And the organisations that are already building their AI capability are compounding their advantage with every improvement.

👉 Action: Next time your team says "AI can't handle that" - pause and check. The limitation they hit three months ago may already have a solution. Build a habit of revisiting what's possible every quarter.

👉 Read the CMU research

2️⃣ MiroFish: predicting the future just got more human

MiroFish is an open-source AI prediction engine that takes a radically different approach to forecasting. Instead of crunching numbers in a spreadsheet, it spawns thousands of AI agents - each with unique personalities, memories, and perspectives - and lets them interact in a simulated world. Opinions shift. Coalitions form and break apart. What emerges is a prediction that accounts for the messy, social reality of how humans actually behave.

Why it matters: Traditional forecasting treats the world like a maths equation. But business decisions play out among people - customers, regulators, competitors, staff - who react to each other in unpredictable ways. Tools like MiroFish are early building blocks for much, much better predictions. In a world of increasing uncertainty, the ability to model scenarios that include human messiness is becoming a genuine competitive edge.

👉 Action: Pick one strategic decision you're facing this quarter and ask: "What would change if we could simulate how our customers, competitors, and regulators would actually react to this?" That's the direction forecasting is heading.

👉 Explore MiroFish on GitHub

3️⃣ Comprehension debt: the hidden cost of AI-generated code

Addy Osmani (Google engineer) published a sharp piece on comprehension debt - what happens when AI generates code faster than humans can understand it. An Anthropic study found that engineers using AI for code generation scored 17% lower on comprehension tests than those who used AI to ask questions and explore trade-offs. The tests pass. The velocity looks great. But nobody truly understands what was shipped.

Why it matters: This isn't just a developer problem. It's a leadership problem. As AI generates more of the work - reports, analyses, strategies, code - the role of the leader shifts from "producer" to "comprehender". The checks and balances matter more, not less. Are you building the governance to match?

👉 Action: Ask your technology and operations leads: "What percentage of our AI-generated output do we genuinely understand? And what's our plan when something breaks that nobody can explain?"

👉 Read the full piece

🎨 Weekend Playground

This weekend, try Gamma, the AI-powered design tool that turns prompts into polished presentations, documents, and landing pages.

Gamma has recently been upgraded with new templates and now connects directly with Claude via its MCP integration - meaning you can research and create a presentation in one seamless flow. Give it a messy set of notes and watch it produce something you'd actually be happy to present on Monday.

Why this matters: The cost of creating a polished visual output is collapsing. What used to take a designer half a day now takes five minutes and a decent prompt. That's the "cost of doing approaching zero" principle in action.

👉 Mission:

• Create a 5-slide deck summarising your team's top priorities for Q2

• Try at least three different themes and see how the framing changes

• Ask it to generate a one-page document version of the same content

• Share the best version with a colleague and see if they can tell it was AI-made

• Bonus: connect Gamma to Claude and try creating a deck from a voice note or messy brief

📢 Share the Optimism

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift. If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

Your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

📰 This was the week that was...

This was the week when enterprise AI stopped being a nice-to-have and started looking like plumbing.

Against a backdrop of surging oil prices and geopolitical uncertainty that makes last month look calm, two quiet moves may matter more to your business than any headline. Microsoft announced Copilot Cowork - built in collaboration with Anthropic, bringing Claude's agentic capabilities into the Microsoft 365 suite your teams already use. If your organisation runs on Microsoft, Copilot just became a much more interesting default.

At the same time, Anthropic launched Claude Marketplace - an enterprise store where businesses can buy Claude-powered tools from partners like GitLab, Snowflake, and Harvey, all consolidated under one invoice. No commission taken. For leaders wrestling with AI procurement headaches, this simplifies the buying decision. Together, these two moves signal something bigger: AI is embedding itself into the systems organisations already rely on. The era of separate AI tools is ending. The era of AI-inside-everything is beginning.

Let's get into it.

🔥 URGENT PRIORITIES

✅ No fires to fight this week

✅ AI is embedding into your existing stack faster than most teams realise

✅ Time to plan for upskilling, not just adoption

This isn't a week for panic. It's a week for redesigning how your people work.

🎯 STRATEGIC INSIGHT

Tension: AI in the enterprise has been slow. Lots of pilots. Not many production deployments. Leaders have been waiting for clarity. This week, clarity arrived - and it came fast. Microsoft integrating Claude into Copilot, Anthropic launching a marketplace, and the steady drumbeat of agentic tools all point in one direction: AI is about to accelerate inside organisations, not around them.

Optimistic insight: The adoption bottleneck was never the technology. It was procurement, governance, and integration. Copilot Cowork solves the integration problem by bringing agentic AI into the apps your teams already use - Word, Excel, Outlook, PowerPoint. Claude Marketplace solves the procurement problem by letting enterprises consolidate AI spend under one roof. When the friction disappears, adoption accelerates. That means every employee needs upskilling and the organisation needs redesigning for new ways of working.

What's shifting: The smart question is no longer "should we adopt AI?" but "how do we redesign work for a world where AI is already in the stack?" That means building prompt libraries and context libraries (like Open Knowledge Graphs) that work across teams. It means redesigning processes. And ultimately, it means reinventing the business model itself.

Why this matters now: There are three jobs in AI adoption. First, make headroom and learn. Second, make room in the P&L by realising the benefits of AI in your current business. Third, reinvent the business for the new AI economy. Most organisations are still on job one. Copilot Cowork and Claude Marketplace just made job two dramatically easier. Between now and Q3, the gap between organisations that redesign and those that wait will widen fast.

👉 Takeaway: Audit your team's readiness for AI-inside-everything. Ask three questions:

• Do we have shared prompt and context libraries across teams?

• Which processes would change if AI could act, not just answer?

• What business cases were previously unviable that should be reconsidered now the cost of technology has collapsed?

🤓 GEEK OUT

1️⃣ Claude Code Review is the new accelerator for dev teams

Anthropic launched Claude Code Review - an AI-powered code review tool that integrates directly into development workflows. It analyses pull requests, suggests improvements, catches bugs, and explains its reasoning. For teams that have just started coding with AI, this is the next step: AI that reviews the AI-generated code.

Why it matters: This is part of a rapidly developing ecosystem that is compressing the cost of technology. For leaders, that has two implications. First, business cases that were previously unviable - too expensive to build, too small to justify the team - can be reconsidered. Second, perceived technology moats need reassessing. If anyone can build, the advantage shifts from what you build to how well you understand your customers.

👉 Action: Ask your CTO or head of engineering: what would we build if development costs halved? That list is your innovation pipeline.

2️⃣ MIT researchers improve AI's ability to explain its predictions

MIT computer scientists developed a new technique that transforms any computer vision model into one that can explain its predictions using concepts a human can understand. The method extracts knowledge the model has already learned and translates it into plain language - so a clinician, for example, could see exactly which visual patterns influenced a diagnosis.

Why it matters: Understanding AI is hard, and for robust systems in regulated industries, it's essential. But here's the optimistic read: the problems with AI are being solved, quickly. Explainability, hallucination reduction, bias detection - none of these are permanent blockers. They are engineering challenges with active, well-funded research behind them. For leaders, this is a reason to lean in, not hold back.

👉 Action: If "we can't explain the AI's decisions" is blocking a project, revisit it. The explainability toolkit is improving faster than most governance teams realise.

3️⃣ The Anthropic Hive Mind - "improv at scale"

Steve Yegge's essay describes how Anthropic operates internally with what he calls "improv at scale" - a new organisational structure where small teams of humans and AI agents collaborate fluidly, picking up and building on each other's work without rigid handoffs or hierarchies. The analogy is improv comedy: "yes, and" as an operating principle, amplified by AI.

Why it matters: AI isn't a tech change. It's an everything change - people, process, culture, business model, ways of working. This essay is a window into what organisations might look like when AI is truly embedded. For leaders, the takeaway is that organisational design - not model selection - is where the real transformation happens. The companies that redesign how they work will outperform those that simply add AI to existing structures.

👉 Action: Read this essay and ask: what would "improv at scale" look like in our organisation? Where are rigid handoffs slowing us down?

🎨 WEEKEND PLAYGROUND

This weekend, try MsgVault - an AI-powered email search tool that runs locally on your machine.

Let's be honest: email search is terrible. You know that message exists. You know roughly when it was sent. And yet your inbox search returns 400 irrelevant results. MsgVault uses AI to actually understand what you're looking for and find it.

Why this matters: It's a small, tangible demonstration of what "AI that runs locally" feels like. No data leaves your machine. And once you've experienced good search, you'll never tolerate bad search again.

👉 Mission:

• Install MsgVault and connect your email

• Search for something you've been meaning to find for months

• Try a natural language search like "that invoice from the supplier in Leeds last autumn"

• Compare the results to your normal email search

• Report back: was it better?

📢 SHARING IS A GIFT. IF YOU VALUE WHAT YOU READ PLEASE SHARE THE OPTIMISM

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift.

My mission is to create a community of business leaders, properly briefed with Ai Optimism so that together we can create the tomorrow we all want to live in.

Your support sharing this newsletter, referring friends, family and neighbours to AI Night School and Sherpas AI all helps the mission. So please do give the gift of AI optimism.

Of course, If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben

Friends,

Your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

A quick note from Hugo:

Next week I am in the French Alps, detoxing digitally and simply being (on skis). In the back of my mind will be how to further support this community and empower everyone to build the AI Optimist community into the force for good it needs to be.

I’ve set a goal of bringing together 10,000 AI Optimists in this community by the end of 2026 and we are 1/3 of the way there already.

My ask of you is to share this newsletter with other leaders who will find this useful, and who should be a part of the community we are building - Sharing is a gift.

Thank you.

📰 This was the week that was...

AI showed us two things this week: how powerful it's become, and why resilience matters.

On Monday, Claude went offline - overwhelmed by demand after Anthropic's very public dispute with the Pentagon over autonomous weapons and mass surveillance. Anthropic refused to allow unrestricted military use. The US government moved to cut ties. Meanwhile, GPT-5.4 launched with a million-token context window and native computer-use capabilities. Perplexity rolled 19 AI models into a single orchestration platform. And a wave of AI-native startup tools proved you can now take an idea to a working business for $50 a month.

The geopolitics were heavy. The conflict in Iran cast a long shadow. But the practical signal for business leaders is clear: AI capability keeps accelerating, the tools keep multiplying, and building resilience into your AI strategy is no longer optional - it's just good practice.

Let's get into it.

🔥 URGENT PRIORITIES

⚠️ Review your AI dependency risk - within 90 days

This week demonstrated that a frontier AI tool can become unavailable through a simple outage, a contractual dispute, or geopolitical fallout. If your organisation relies on a single AI provider for critical workflows, you have a single point of failure.

Action: Audit which AI tools your teams depend on. Identify which workflows would break if access was revoked. Start mapping fallback options - whether that's a second provider, an open-source model, or a manual process.

👉 Book a call to help you respond to this.

I considered whether this warrants being classed as urgent, as my goal is to consistency reinforce that almost none of the noise is actually urgent. I decided that the outage of Claude on Monday, due to geopolitical as much as technical changes, combined with the outbreak of the Iran war, highlighted something critical enough to raise. If you think I was wrong, then get in touch as I would love to know what you think.

🎯 STRATEGIC INSIGHT

Self-sovereignty is the missing piece in your AI strategy

Tension: Right now, frontier models from closed providers deliver extraordinary power at low cost. For most organisations, the practical move is to use them. But this week showed the fragility of that position. Claude - arguably the best AI tool available - went down. Then it was caught in the crossfire between an AI company's ethics and a government's military ambitions. Defence contractors are already switching providers. The lesson is clear: any tool you depend on can become unavailable - through outage, policy change, or forces entirely outside your control.

Optimistic insight: This doesn't mean you should stop using frontier models. It means you need to plan for when you can't. There are three jobs in AI adoption: (1) make headroom and learn, (2) make room in the P&L by realising AI's benefits in your current business, and (3) reinvent the business for the new AI economy. Short-term, frontier AI plus training equals making room in the P&L. That's the right move today. But simultaneously, you need to be building towards self-sovereignty - treating it the same way you'd treat any core infrastructure dependency.

What's really shifting: The model landscape is fragmenting in your favour. GPT-5.4 launched today with a million-token context window, native computer use, and dramatically better efficiency. Perplexity's new Computer platform orchestrates 19 different models through a single interface - Claude, GPT, Gemini, Grok - routing each task to whichever model handles it best. Open-source models keep improving. The tools for building sovereign AI stacks are becoming accessible. The cost of switching is falling.

Why this matters now: We are living through the week that made AI geopolitical risk real for businesses. Whether your concern is outages, vendor lock-in, regulatory shifts, or the kind of government intervention we saw this week, the answer is the same: diversify your AI infrastructure and start building towards sovereignty. Not because frontier models aren't valuable. Because depending entirely on someone else's infrastructure for your core business operations is a risk you need to actively manage.

Actionable takeaway: Map your AI dependency. For each critical workflow powered by AI, ask: what happens if this provider is unavailable for a week? If the answer is "we're stuck," start building a fallback now.

🤓 GEEK OUT

1) AI startup platforms are rewriting the business case for innovation

Three platforms emerged this week that demonstrate something profound: the cost of bringing an idea to life as a business is collapsing. Polsia autonomously runs companies - planning, coding, marketing, operations - and has crossed $1M ARR with zero employees. Audos and LeanOS are building similar AI operating systems on top of Claude Code, reducing the cost of innovation to a fraction of what it was.

👉 Try Polsia | Explore Audos | See LeanOS

Why it matters: Most organisations have been designed around the status quo. Innovation has traditionally been expensive and risky, which is why most businesses avoid it. These tools change the maths. If you can take an idea from concept to working business for the cost of a monthly subscription, the barrier to reinventing your business just disappeared. As you increase your ambition, these tools give you an innovation capability at a fraction of the previous cost - reducing the risk of doing the work needed to reinvent.

2) AI is helping save the bees - and the rest of nature too

The BBC highlighted how AI is being used to protect nature - from tracking pollinator populations using acoustic monitoring to AI-powered beehives that automatically detect disease, control temperature, and protect colonies from pesticides. Researchers at the University of Edinburgh are training AI to distinguish between the buzzes of different bee species, including the critically endangered great yellow bumblebee. Meanwhile, AI-driven environmental monitoring platforms are helping farmers understand biodiversity across entire landscapes.

👉 Read the BBC piece

Why it matters: This is AI for good at its most tangible. Almost 90% of flowering plants and over three-quarters of the world's staple crops depend on pollinators. AI is giving conservationists tools they never had - remote monitoring, early disease detection, and data at a scale that was previously impossible. A good reminder that AI's most important applications aren't always in the boardroom.

3) The organisational rails of AI - why the standards being set now matter most

Tom Wynne-Morgan's brilliant essay draws a parallel between railways standardising time and AI standardising organisational interpretation. Before railways, every town set its own clock. AI is doing the same thing to how organisations define risk, quality, and "good enough." The standards being encoded into AI systems now will become invisible defaults within a generation.

👉 Read the essay

Why it matters: This is the kind of thinking that separates leaders from followers. If AI is laying the organisational track, the question is: are you paying attention to which standards are being embedded while there's still something to negotiate?

🎨 WEEKEND PLAYGROUND

💐 Give ChatGPT 5.4 a Mother's Day mission

Next weekend is Mother's Day. GPT-5.4 just launched today with dramatically improved capabilities - better writing, native computer use, and a million-token context window.

Give it a proper test: ask it to plan the perfect Mother's Day. Menu, schedule, gift ideas, a personalised card, even a playlist. See how far you can push it with a single, detailed prompt.

Bonus points if you get it to write a poem your mum would actually like.

📢 SHARING IS A GIFT. IF YOU VALUE WHAT YOU READ PLEASE SHARE THE OPTIMISM

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift.

My mission is to create a community of business leaders, properly briefed with Ai Optimism so that together we can create the tomorrow we all want to live in.

Your support sharing this newsletter, referring friends, family and neighbours to AI Night School and Sherpas AI all helps the mission. So please do give the gift of AI optimism.

Of course, If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben