A humanoid robot just broke the human half-marathon world record

On Sunday in Beijing, a humanoid robot built by smartphone maker Honor finished a half marathon in 50 minutes 26 seconds. The human world record is 57 minutes. Last year’s winning robot took 2 hours and 40 minutes.

Three times faster in one year of progress.

The humanoid robotics market is projected to hit 1 billion machines by 2050. Sunday’s race is what the early stages of that look like.

The NSA got caught using the AI the Pentagon banned

Anthropic released Claude Mythos earlier this month, a model powerful enough to find zero-day vulnerabilities across every major OS and browser. They refused to release it publicly. Instead they locked it down to around 40 organisations under Project Glasswing.

The Pentagon responded by labelling Anthropic a “supply chain risk” and banning its tools across the Department of Defense.

Then Axios reported that the NSA, which sits inside the Pentagon, is one of the 40 organisations quietly using Mythos to scan its own networks for vulnerabilities.

The Department of Defense is suing Anthropic in court while its own intelligence agency uses the model. The DoD found out from a news article.

Anthropic CEO Dario Amodei met with the White House chief of staff and the Treasury Secretary last Friday to discuss government use of Mythos. Both sides called it productive.

Allbirds sold its shoe brand for $39M and pivoted to GPU cloud

Allbirds was once a $4 billion company. The Silicon Valley crowd loved the wool trainers. Then the hype faded, the stock collapsed, and last week the company sold the entire Allbirds brand and its shoe assets to American Exchange Group for $39 million.

What’s left of the company is rebranding as NewBird AI, a GPU-as-a-service and AI compute cloud provider. They raised $50 million from an undisclosed investor to buy high-performance GPUs and lease them to customers who need compute.

Stock up 600% on the news.

TechCrunch compared it to 2017, when a company called Long Island Iced Tea rebranded as a blockchain company and the stock jumped 275%. That one got delisted within a year.

Blue Origin reused a booster for the first time, then lost the payload

New Glenn launched on its third flight on Saturday, successfully recovering and reusing its first-stage booster for the first time. A real milestone for Jeff Bezos’s rocket company, which has been chasing SpaceX’s reusability record for years.

The satellite it was carrying didn’t make it to the intended orbit.

Half a win.

The AI data centre buildout has a PR problem

Activist groups in 24 US states are now actively organising against AI data centers. The complaints are utility bill spikes, water consumption, noise pollution, and land use. There have been ballot measures, council ousters, and isolated vandalism.

65% of Americans now say they oppose new data center facilities being built near them. That number is bipartisan.

Big Tech is committing hundreds of billions to AI infrastructure while communities are voting against the buildings. That tension is going to get harder to ignore.

All 11 original xAI co-founders have left

The last of the original team departed this week. Musk posted on X: “xAI was not built right first time around, so is being rebuilt from the foundations up.”

He then hired Andrew Milich and Jason Ginsberg, the two engineers who scaled Cursor to $2 billion in revenue, to lead the rebuild.

xAI was valued at $250 billion as part of its merger with SpaceX in February. The company now has a new leadership team, a new technical direction, and an IPO on the horizon. A lot riding on the rebuild going right.

Revolut wants a $200B valuation

Revolut told investors this week it is targeting a valuation of $150 to $200 billion in its eventual IPO. The company was valued at $75 billion in a secondary sale last year. The CEO says the listing is at least two years away and will be in the US.

The numbers back it up. $6 billion in revenue in 2025. $1.7 billion net profit. 68 million customers. $9 billion in revenue projected for 2026.

It will be one of the biggest fintech IPOs in history. And it won’t be in London, which has been trying to keep Revolut as a UK listing for years. Another one the UK loses.

How I can help

👨‍💻 Software

Need a top tech talent for your next product? Are you looking for your next gig? I’ve got you sorted.

📢 Advertising

Advertise in my newsletter to get in front of 13,893 founders.

🤿 Get more deep dives

Upgrade to a premium subscription to get an extra deep dive a week.

— Oliver

Anthropic built an AI too dangerous to release

Every week there’s an Anthropic news. Sorry but it is what it is.

Now they developed a new model called Claude Mythos. It’s powerful enough to find zero-day vulnerabilities across every major operating system and browser in the world. All of them.

They’re not releasing it publicly.

Instead they launched Project Glasswing, a closed partnership with Apple, Google, Microsoft, Amazon and Nvidia to use Mythos defensively.

The idea is to find and patch critical vulnerabilities before attackers build something similar and go on the offensive.

US Treasury Secretary Bessent and Fed Chair Powell held an emergency meeting with major banks this week specifically about Mythos. They're trying to lock down financial infrastructure ahead of any potential leak.

OpenAI isn’t far behind. They confirmed this week they’re releasing a dedicated cybersecurity AI product to a limited set of partners in the coming months. Both frontier labs are now explicitly in the arms race between offensive and defensive AI capabilities.

A paralysed veteran raiding World of Warcraft dungeons with his thoughts

Jon Noble is a British Army veteran paralysed from the shoulders down after a car accident in 2004. In December 2025 he became the 18th person to receive a Neuralink N1 brain-computer interface, implanted directly into his motor cortex.

• This week he hit 100 days with the chip.

• By week three, controlling a cursor felt natural

• By day 80 he loaded up World of Warcraft. The first raid was clunky. Then his brain synced with the interface and he describes it as “pure magic.”

• He’s now raiding WoW dungeons at full speed, no keyboard, no mouse, only thoughts.

“The N1 didn’t just give me a new way to use a computer. It gave me a new way to live,” he says.

Neuralink is still a clinical trial with a tiny number of participants. It is not a consumer product. But the rate of progress is remarkable.

Scientists just worked out why some people get ALS and dementia and others with the same gene mutation don’t

The most common genetic cause of both is the C9orf72 mutation. Loads of people carry it and never develop either condition. Nobody could explain why.

Case Western Reserve just found the answer: gut bacteria.

Certain microbes produce a toxic sugar that triggers an immune response, which crosses into the brain and starts killing cells. 70% of ALS patients had elevated levels. Only a third of healthy people did.

When researchers reduced the harmful sugars, brain health improved and lifespan extended.

India built a 1,000km quantum network six years early

India’s National Quantum Mission demonstrated a 1,000km quantum communication network this week. One of the longest in the world. Built with indigenous technology by a startup called QNu Labs.

The original timeline was 2,000km in eight years. They hit half that distance in under two years.

Quantum key distribution means the network is physically impossible to intercept. If anyone tries, the quantum state of the signal changes and the receiver knows immediately.

Also, they are inaugurating their first indigenous, open-access quantum computer tomorrow.

Compute is the only thing that matters now

Meta signed a $21B additional compute deal with CoreWeave this week, on top of a prior $14.2B commitment. Anthropic signed a major deal with them too. CoreWeave stock jumped 11% on the news.

Meta’s total capex for 2026 is projected at $115-135 billion. Nearly double last year. All of it AI infrastructure.

Amazon confirmed their cloud AI revenue hit a $15B run rate in Q1. Their custom chips business (Graviton and Trainium) is now doing $20B annually, double the number they cited earlier this year.

Software stocks are getting cooked

• The SaaS ETF (IGV) is down 27% year to date.

• ServiceNow dropped 8% in a single session this week.

• Salesforce led the Dow down.

• Datadog off 5%.

The market is repricing every software company that built its moat on complexity.

If Claude can automate the workflows that justify your $50k/year enterprise contract, your revenue multiple collapses. Investors are figuring this out and the sell-off is accelerating.

The companies most at risk are the ones where the core product is a wrapper around human labour doing repetitive knowledge work. So.. most of SaaS.

Scientists watched Alzheimer’s form in real time

For decades we knew copper ions triggered the protein clumping that destroys brain cells. We could see the before and after, just couldn't see the process.

Oregon State built a technique that tracks it second by second as it occurs. They also found a molecule that can grab the copper and reverse the clumping after it starts.

The lead researcher said "with the correct targeting, some of the brain damage might be reversible".

How I can help

👨‍💻 Software

Need a top tech talent for your next product? Are you looking for your next gig? I’ve got you sorted.

📢 Advertising

Advertise in my newsletter to get in front of 13,782 founders.

🤿 Get more deep dives

Upgrade to a premium subscription to get an extra deep dive a week.

— Oliver

Some of these are niche, you’d need to dig to find them organically. So whether you’re unsure what to do with your life or just need something to read on the train home - I’ve got you.

Some essays you read. Some essays read you.

These are the second kind.

1. How to be remarkable by Seth Godin

Remarkable doesn’t mean remarkable to you. It means remarkable to me. Am I going to make a remark about it? If not, then you’re average, and average is for losers.

Being noticed is not the same as being remarkable. Running down the street naked will get you noticed, but it won’t accomplish much. It’s easy to pull off a stunt, but not useful.

2. When to do what you love by Paul Graham

When you can't decide which path to take, it's almost always due to ignorance. In fact you're usually suffering from three kinds simultaneously: you don't know what makes you happy, what the various kinds of work are really like, or how well you could do them.

What if you're not sure what you want? What if you're attracted to the idea of making money and more attracted to some kinds of work than others, but neither attraction predominates? How do you break ties?

3. How to fix your entire life in 1 day by Dan Koe

Most people set a surface-level goal, hype themselves up to remain disciplined for the first few weeks, then go back to their old ways without much struggle, because they were trying to build a great life on a rotting foundation.

The bodybuilder has to grind to eat unhealthily. The CEO has to force themself to lie in bed past their alarm clock, and they hate every second of it.

4. Tail end by Tim Urban

It turns out that when I graduated from high school, I had already used up 93% of my in-person parent time. I’m now enjoying the last 5% of that time. We’re in the tail end.

When you look at that reality, you realize that despite not being at the end of your life, you may very well be nearing the end of your time with some of the most important people in your life.

5. 1000 true fans by Kevin Kelly

If you lived in any of the 2 million small towns on Earth you might be the only one in your town to crave death metal music, or get turned on by whispering, or want a left-handed fishing reel. Before the web you’d never be able to satisfy that desire.

Whatever your interests as a creator are, your 1,000 true fans are one click from you. There is nothing — no product, no idea, no desire — without a fan base on the internet.

6. No yes. Either hell yeah! or no by Derek Sivers

Every event you get invited to. Every request to start a new project. If you’re not saying ‘HELL YEAH!’ about it, say no.

We’re all busy. We’ve all taken on too much. Saying yes to less is the way out.

7. This is water by David Foster Wallace

Learning how to think really means learning how to exercise some control over how and what you think. It means being conscious and aware enough to choose what you pay attention to and to choose how you construct meaning from experience.

Because if you cannot exercise this kind of choice in adult life, you will be totally hosed.

8. No one can teach you to have conviction by Ben Kuhn

If you join a company where someone else is already making those decisions well, you’ll never get the type of practice that you need in order to build your own models and heuristics.

You’ll end up with a good, but not perfect, model of ‘what would my boss do?’ — a model that can make the 95% of easy decisions, but not the 5% of hard ones that add the most value.

9. How to be successful by Sam Altman

Most people are primarily externally driven; they do what they do because they want to impress other people.

The most successful people I know are primarily internally driven; they do what they do to impress themselves and because they feel compelled to make something happen in the world.

10. You and your research by Richard Hamming

If you do not work on an important problem, it’s unlikely you’ll do important work. It’s perfectly obvious.

The average scientist, so far as I can make out, spends almost all his time working on problems which he believes will not be important and he also doesn’t believe that they will lead to important problems.

11. Solitude and leadership by William Deresiewicz

What we have now are the greatest technocrats the world has ever seen, people who have been trained to be incredibly good at one specific thing, but who have no interest in anything beyond their area of expertise. What we don’t have are leaders.

What we don’t have, in other words, are thinkers. People who can think for themselves. People who can formulate a new direction.

12. The days are long but the decades are short by Sam Altman

Do new things often. Not only does doing new things seem to slow down the perception of time, increase happiness, and keep life interesting, but it seems to prevent people from calcifying in the ways that they think.

Aim to do something big, new, and risky every year in your personal and professional life.

Hope you liked this 💎 [FREE] deep dive 💎

Paid tier drops next week: 1 free deep dive/month, rest paywalled.

Running a pre-sale now: 22% off FOREVER (just 1 kebab/month)

Get 22% off forever

How I can help

👨‍💻 Software

Need a top tech talent for your next product? Are you looking for your next gig? I’ve got you sorted.

📢 Advertising

Advertise in my newsletter to get in front of 13,428 founders.

🤿 Get more deep dives

Upgrade to a premium subscription to get an extra deep dive a week.

Must be something about it. So naturally I went, gave Claude Code a chance and almost build a full startup in 4 days.

Here’s how the first 3 days looked, including setup, niche recommendations and more.

Have a look at Landing Page, or Waitlist, or the Product I’m building. Synthetic data generation is not yet connected so you don’t cost me $10T in api calls, cheeky.

This is where UPDF comes in. It’s what I use to handle specs, reports, research, and PDFs without switching apps or paying Adobe’s overpriced subscription.

This AI is useful for my research agents; it summarises a 100-page investor report in seconds, translate a competitor filing, turn a dense doc into a mindmap. One license covers all your devices. Edit a contract at your desk, sign it on your phone on the way to the gym.

If your document workflow is still Adobe + 3 other tools, you’re carrying weight you don’t need.

Download UPDF and automate your document workflow today!

What’s Claude Code?

Anthropic’s command-line agent. You open your terminal in a project folder, type claude, and talk to it. It reads your entire codebase, edits files, runs commands, writes commits — no copy-pasting snippets into a chat window.

Think of it as pair programming with an AI that already has your repo open.

20 Minutes into coding I realised I need something called tmux (terminal multiplexer). Two features matter here:

1. Split panels. I’m currently running 4 agents in unison - Coder, Tester, Researcher and Designer.

2. You can Detach and walk away: I give Claude Code a big task, press Ctrl+B then D, close your laptop, go hit the gym. Come back, type tmux attach, and Claude has finished the work.

If you don’t have tmux your SSH/Wi-fi drops results in your laptop sleeping.

Quick Setup (5 minutes)

Installation:

1. npm install -g @anthropic-ai/claude-code
   or

   curl -fsSL https://claude.ai/install.sh | bash

2. brew install tmux

Starting a session:

1. tmux new -s project

2. cd your-project

3. claude

Detach anytime with Ctrl+B, D. Reattach with tmux attach -t project.

Split your terminal: Ctrl+B, % for a vertical split, Ctrl+B, “ for horizontal. Switch between panes with Ctrl+B, arrow key.

My current workflow

I open a tmux session with 4 panels. I had claude code write a script to launch all agents in the given way (iterated maybe 10 times over these 3 days to adjust it).

It includes running researcher and designer first to pick up on tasks, communications from previous sessions. 3 minutes later tester and coder start. Their backlog gets populated by the first two agents.

When I want a manual change I type something like: “Add a billing page that pulls the user’s Stripe subscription status and shows their current plan.”

Claude reads my existing auth setup, checks how I’ve structured other pages, creates the component, wires up the API call, updates the router.

20 minutes into the first coding session I realised this is not optimal and to have the guy commit like he does now.

I had to make it all more autonomous.

Current 4-Agent Setup

Let me break down what each agent actually does and why I landed on this split.

- Researcher: Its job is to read documentation, check for breaking changes in dependencies, look up best practices, and dump findings into a shared markdown file. Before I added this agent, I kept running into situations where the Coder would implement something using an outdated API or miss a better approach entirely. The Researcher always runs first.

- Designer: It reviews the existing UI, checks for consistency, and produces component specifications. It writes these specs into task files that the Coder picks up. Early on I made the mistake of letting the Coder make design decisions on the fly. The result was functional but visually inconsistent. Having a dedicated agent that only thinks about design patterns solved that.

- Coder: The workhorse. It reads task files from the Researcher and Designer, picks the highest-priority one, and implements it. It has access to the full repo, runs the dev server, and commits when a task is complete.

- Reviewer: Runs alongside the Coder. Every time a new commit lands, he pulls the changes, runs the test suite, and if something breaks, writes a bug report into the task backlog.

How They Communicate

I landed on something a dead simple shared protocol: a shared directory of markdown files.

Just files on disk that every agent can read and write. Git handles the versioning. If two agents try to write at the same time, one commit wins and the other retries — which in practice almost never happens because they’re operating on different files.

Each task file looks something like this:

id: TASK-042
priority: high
status: pending
created_by: designer
assigned_to: coder

## Add billing page
### Context

User needs to see their current Stripe subscription status.

### Requirements

- Pull subscription data from /api/billing/status
- Show current plan name, renewal date, payment method
- Follow the card layout pattern used in /settings pages
- Handle loading, error, and empty states

### References

- See /coordination/specs/billing-api-research.md
- Follow component structure in src/components/settings/

The Coder picks this up, implements it, marks the status as done, and moves on. If the Tester catches an issue, it creates a new task referencing the original.

The Launch Script

I mentioned I had Claude Code write the launch script. Here’s the logic behind it:

1. Phase 1 (t=0): Researcher and Designer start. They scan for any leftover tasks from the last session, check git history for what changed, and populate the backlog.

2. Phase 2 (t=3min): Coder and Tester start. By now there are tasks waiting. The delay is important — without it, the Coder would start with an empty backlog and either idle or pick up stale work.

Each agent gets launched in its own tmux panel with a system prompt tailored to its role. The system prompt includes:

- Which directories to read from and write to
- What its priorities are
- When to stop and wait vs. when to keep going
- How to format its outputs so other agents can parse them

The script also includes a health check — if an agent crashes or gets stuck in a loop, it restarts with context about what happened.

What Surprised Me

- The agents get better over time: It’s because the codebase grows with consistent patterns. The more code that exists in your style, the more context Claude has for matching it. By day 2, the Coder was producing components that looked like I wrote them.

- The Researcher pays for itself immediately: Originally I was sceptical about dedicating a whole agent to “just reading docs.” But the number of times it caught a deprecation or found a simpler API before the Coder went down the wrong path made it the highest-ROI agent in the setup.

- You still need to architect: The agents are excellent at implementing well-defined tasks. They’re not great at deciding what the product should be. Building redundant features without a higher endgoal in mind is coder’s motto I feel like.

What I Need to Change

- I’ve added a ‘night-agent’. I try to overflow backlog with ~100 features before I go to bed and pass it to night agent to code. Problem is he churns out after an hour. Need to fix for now.

- Researcher and Designer are rarely needed by the day 3. There’s only so much needed to research. I’ll switch to running two coders, frontend & backend + feature designer on launch

- I set up an imsg notification each time coder commits but it’s mental. Need to make it more decluttered

- Sometimes I find the agents just fucking around. Today I had a broken synthetic data generation method (the key functionality) and let the agents fix it on its own. Checked up on them 3 hours later, it was still unfixed + they added ?!64?! items on a navigation bar. HAVE YOU SEEN A PRODUCT WITH 64 ITEMS ON A SIDEBAR?! Got it down to 11.

Hope you liked this 💎 [FREE] deep dive 💎

Paid tier drops next week: 1 free deep dive/month, rest paywalled.

Running a pre-sale now: 22% off FOREVER (just 1 kebab/month)

Get 22% off forever

How I can help

👨‍💻 Software

Need a top tech talent for your next product? Are you looking for your next gig? I’ve got you sorted.

📢 Advertising

Advertise in my newsletter to get in front of 13,219 founders.

🤿 Get more deep dives

Upgrade to a premium subscription to get an extra deep dive a week.

Read more

Read more

Read more

The last couple of months have been... let's say chaotic. Wait — chaotic? I don’t mean disordered and confusing.

The latter definition seems closer to what I’ve been trying to say. The property of a system whose behaviour is so unpredictable, as to appear random.

Emphasis on random. That’s the word.

I'll use chaos interchangeably with randomness, timing, luck, and other hidden constraints. Because that's the problem that made me write this dive.

What would my day today look like if I'd made one less misstep last year? Would anything change? Would the butterfly effect screw me over?

People keep assuming outcomes come from variety of shaping factors: skill, planning, intelligence, grit. This leaves this often overlooked variable: chaos.

It affects absolutely everything.

Where you were born. How your brain functions. What personalities your closest people had growing up. What your early ambitions became and how you chose to execute them.

If you took two random p…

Read more

Read more

Read more

Some of you liked it, so here’s the longer version with justifications why I think these predictions are going to happen.

Read more

Read more

Read more

Read more

Read more

Read more

If you’re not deep into gaming, here’s the quick overview and why a video game market crash might be the most relevant lesson your startup learns this October.

Read more

Read more

Read more

Read more