Skip to main content

The Journey

Most posts here are ad hoc: what I learned building with AI that week, some philosophy, a writing lesson, something I came across.

Subscribe via RSS
Social card. Headline: Told to prove an AI pilot works? Banks already have the method. Below: a challenger, a parallel run against real outcomes, monitoring after go-live, a named owner, an independent checker.
Essay

Banks already have a method for checking AI works. Four steps to steal it from their model-risk teams

20 October 2026
11 min read

Told to show an AI deployment works? Bank model-risk teams already run the test, written down by the regulator: a challenger, a parallel run against real outcomes, monitoring after go-live, a named owner, and an independent checker. Four steps to steal it, whatever business you're in.

Read More
Social card. Headline: An AI company wants to buy your firm. Its job adverts show what it would do to your firm. Below: move the data, consolidate the accounts, end the leases, sit with the staff. The AI is the system they move you onto.
Essay

An AI company wants to buy your firm. Its own job adverts show what it would do, and most of it you can do yourself

17 October 2026
11 min read

AI-backed buyers are approaching small accountancy, lettings and IT firms. Their own job adverts show the first months after a sale: moving data, consolidating accounts, ending leases, months on site with your staff. Which of those steps you can take without selling, and the question to ask before you sign.

Read More
Social card. Headline: AI customers say yes too often. Below it: pretend shoppers predicted 52% would buy, 33% did; asked plainly, with no persona, the model predicts better.
Essay

Telling AI to act as your customers makes its answer worse: 52% of shoppers would buy, it said. Real sales: 33%.

17 October 2026
8 min read

Pre-registered tests show AI customers say yes far more often than real ones, and that asking a model to role-play predicts worse than asking it straight. I built a product on the opposite bet. What a fake customer is still good for.

Read More
Social card. Research paper. How to implement AI in your organisation and check it's working. Who does the work, how to find the real process, how to test it, what it does to people. Every claim graded.
Paper

How to implement AI in your organisation and check it's working

16 October 2026
47 min read

What the evidence supports for putting AI to work in an organisation, graded claim by claim, and a six-step method you can run on one process.

Read More
Social card. Research paper, October 2026. AI law is splitting apart. The duties are converging. What binds now across the EU, UK, US, China and Asia, what lands by 2028, and what every organisation needs.
Paper

The world's AI laws disagree. Four duties keep turning up anyway

14 October 2026
25 min read

Where AI regulation stands in October 2026 across the EU, the UK, the US, China and the rest of Asia: what binds now, what lands by 2028, where the rules contradict each other, and the four duties that keep recurring, with different triggers and deadlines in each regime. The key dates checked at source on 14 and 24 October.

Read More
Social card. Is your AI working? Headline: The usual signs that AI is working can mislead. Below: hospitals, airports and banks already have better tests, as six steps for one process.
Essay

The usual signs that AI is working can mislead

13 October 2026
19 min read

What staff say, what bosses see and the person checking the AI's work can all tell you AI is working when it isn't. Hospitals, airports and banks already have better tests. Here they are, as six steps for one process.

Read More
Dark card reading: at $1 million it looks like an income. Until the bad year. Beneath it: if my bot passes its test, plus 9% in a normal year and minus 30% in a bad one. So far it hasn't passed.
Essay

Can a crypto trading bot pay your bills? Mine would need at least $1 million, and a bad year would take $300,000

13 October 2026
5 min read

A crypto bot making $450 a year on $5,000 looks like it only needs more money to become an income. If mine passes its test, $1 million would expect $90,000 in a normal year and plan for losing $300,000 in a bad one. So far it hasn't passed. The three-step test to run before you scale any bot.

Read More
Social card. Forward deployed engineers, explained. Headline: The engineers AI firms like Anthropic send inside companies, and how they make AI pay off. Below: their four moves, as a plan you can run on one process of your own.
Essay

How to make AI pay off at work: the job of the engineers Anthropic sends into companies

12 October 2026
9 min read

Forward deployed engineers are the engineers AI firms like Anthropic send to work inside companies. What they do, their four moves as a plan you can copy, and why mapping the work gets four weeks, not forever.

Read More
Social card. Headline: AI companies are spending billions buying accountants, IT firms and lettings agents. Below: which of their numbers anyone has checked, sorted into known, claimed, reported and not measured. Last checked 26 October 2026.
Essay

AI companies are spending billions buying firms. Here's which of their numbers anyone has checked

12 October 2026
15 min read

AI companies are buying accountants, IT firms and lettings agents, and billions are riding on it. Every number I could check, sorted into what the filings show, what the buyers say and what nobody has measured, dated 26 October 2026.

Read More
Jev, the AI at a fiftieth of OpenAI's price: is it any good?
Video

Jev, the AI at a fiftieth of OpenAI's price: is it any good?

10 October 2026
Watch

Jev is a new AI model that costs a fiftieth of the price of OpenAI's GPT-5.4. On my first test it beat GPT-5.4 by 11 points, and I told the internet. Two days later I ran the same test again, and the win was gone.

Watch on YouTube (opens on YouTube in a new tab)
Dark card reading: a year of AI crypto trades made $722, all of it in one crash. Three lines beneath: its confidence ran backwards; US fees alone would sink it; where it can and cannot be run.
Build Log

Thinking of buying a crypto trading bot? One year, 268 trades and $722 of profit say read this first

10 October 2026
11 min read

A year of an AI picking crypto trades, 268 of them on paper, made $722 and all of it in one crash. Fees, the exchange you can legally use and the country you live in decided more than the strategy. The five numbers to get before you build or buy a bot, and where it can't be run.

Read More
A card reading: Jev, the AI that can't write. It picks from answers you give it, with a probability. Cheap at volume, pointless at small scale.
Essay

Jev, the AI that can't write: where it saves money and where it's hype

8 October 2026
14 min read

What Jev is in plain English, whether it's secretly an LLM, what TypeSafe still hasn't published, and where it genuinely saves money. One independent paper sent Jev 346,009 requests for under $10. Plus five questions to decide whether it would pay for you.

Read More
Social card. Headline: Workers expect to put the time AI saves into other work. Their pay hasn't pulled ahead of non-users' by more than 2% on average. Below: who decides what saved time becomes. Source: Humlum and Vestergaard, March 2026, a study of 25,000 workers in Denmark.
Essay

If AI saves you time, why aren't you being paid more for it?

7 October 2026
12 min read

In a study of 25,000 Danish workers, people saving time with AI say they expect to spend it on other work, and their pay hasn't pulled ahead of non-users' by more than 2% on average. Who decides what saved time becomes, and a log to track your own.

Read More
Social card. Headline: Companies worth billions are buying accountants to run them with AI. Below: what would prove it works, and who has shown it so far.
Essay

An AI company wants to buy your firm? Ask it to prove the AI works first

6 October 2026
10 min read

AI-backed companies worth billions are buying accountants and IT firms. If one calls you, here is what proof its AI works would look like, who has shown any of it, and what to ask before you sell.

Read More
Social card. Headline: Amazon hasn't said AI did the jobs it cut. Below: its chief executive expects AI to reduce the corporate workforce in the next few years.
Essay

Amazon's 30,000 job cuts weren't proof AI is taking jobs. What its boss says is coming is closer to it

5 October 2026
11 min read

Amazon said AI was why it had to move faster, not who did the work. Its boss expects AI to shrink its corporate workforce later. Here's the record, in date order.

Read More
Social card: They checked the computer's debts by hand. 29% were wrong. The volume went up anyway. Cut for speed: the officer who checked whether a mismatch was real. The pace: 20,000 a year became 20,000 a week, by the minister's count. The check: officers redid 919 debts by hand and marked 29% wrong. Robodebt, Australia, 2016, rules-based, not AI.
Essay

Automating with AI? Australia cut one human check from its welfare debt system and unlawfully billed 433,000 people

4 October 2026
9 min read

Robodebt cut a slow human check to go faster. Officers then redid 919 of its debts by hand and marked 29 per cent wrong, and the volume went up anyway. Five things to do before AI takes over a process.

Read More
Card headed 'Built to barely use water: one 200 megawatt block of the Stargate campus in Abilene, Texas.' Three rows. The builder's website: 910,000 gallons a year. The builder's own report: about 2 million gallons a year, about 20,700 litres a day, three Kansas restaurants. The power plants making its electricity, on the Texas grid average: about 6.2 million litres a day, and a water-cooled site pays this too. Footer: Crusoe 2025 impact report and web page, VanSchenkhof 2011, Li and colleagues 2025 using a World Resources Institute 2020 factor.
Essay

OpenAI's Texas data centre is built to barely use water. The power plants it runs on use 300 times more water than it does

2 October 2026
10 min read

The company building OpenAI's Stargate campus in Abilene says a 200 megawatt block uses 910,000 gallons of water a year on its website, and about 2 million in…

Read More
Card headed 'The most quoted AI water number was withdrawn by its own authors in October 2023', with two dated quotes from the same paper: 6 April 2023, 'ChatGPT needs to drink a 500ml bottle of water for a simple conversation of roughly 20-50 questions and answers', and 25 October 2023, 'GPT-3 needs to drink (i.e., consume) a 500ml bottle of water for roughly 10-50 responses'. Footer: the first sentence is the one still quoted in 2026; source arXiv 2304.03271, versions 1 and 2.
Essay

AI water use: the bottle of water per 20 ChatGPT questions was withdrawn by its own authors in 2023. It's still quoted in 2026

29 September 2026
9 min read

The most quoted AI water number, a bottle per 20 ChatGPT questions, was withdrawn by its own authors in October 2023 and is still being repeated. Every AI…

Read More
A card quoting Greg Isenberg: $10 million businesses, $100 million businesses. Beside it, his own guest saying to use it for routing only. The expensive queue the episode claims to unlock costs $1.29 a month on the old model for a thousand enquiries, so nothing was rationed. A narrow tool with four clear uses.
Essay

Jev is a narrow tool with four clear uses. Greg Isenberg's $10m framing is not one of them.

24 September 2026
10 min read

Greg Isenberg's episode on Jev promised $10m businesses. His own guest says use it for routing and keep it away from your money. Here is each claim against the data, and the four conditions under which the tool earns its place.

Read More
A card reading: that chart where one AI beats another? Run it again and the winner can vanish. Mine did. Ask before you pick the tool: was it run more than once, who marked the answers, did the two really disagree?
Paper

Test an AI model before you rely on it: six mistakes I made testing Jev

23 September 2026
22 min read

One run said Jev, TypeSafe's new model, beat GPT-5.4 by 11.1 points. Five runs couldn't tell them apart. The six mistakes behind that headline, and a checklist to run before you rely on any AI comparison, mine included.

Read More