OpenAI turns its coding agents inward
OpenAI put out two pieces on Sunday, one headed Research acceleration: The view inside OpenAI, the other an essay called An Alien Mind12. The essay is by the company's chief scientist, Jakub Pachocki3. Both are about recursive self-improvement, shortened to RSI, and the first does not trouble to expand the acronym3.
The research acceleration piece sets out how OpenAI's own research team are using coding agents3. The developer Simon Willison, who follows the labs closely, describes 2026 as the year agentic engineering really took off inside the company, and says the point is best illustrated by a chart of AI spend per researcher3. That figure accelerated sharply in late July3.
Willison's own guess is that the late July jump is when internal staff got access to the model later released as GPT-6 Astra3. That is his inference rather than anything the company stated3. He also reads RSI as the term OpenAI is now putting where AGI used to sit3.
The essay drew the bigger crowd. An Alien Mind reached 404 points and 352 comments on Hacker News, against 167 points and 111 comments for the research acceleration post21.
Little here to act on yet
Strip out the language about alien minds and what remains is a lab describing its own working week. The one hard measure on offer is spend per researcher, and spend is consumption, not output. Nothing published tells you what shipped sooner, what was dropped, or what any of it saved. A company with free access to its own unreleased models raising its usage is not evidence that agents pay for themselves in a business that has to buy them by the token.
There is no British hook in any of this, and no comment from DSIT, the ICO or anyone else here. It matters to you second hand, through your suppliers. When a vendor starts citing recursive self-improvement as a reason to sign this quarter, the useful reply is to ask which task got faster, by how much, and what the run cost was. If the answer is a chart of spending, you have learned something about the vendor rather than about the technology.
Also today
-
Kill switch alone will not do it, say British firms
Efforts to protect national security through an AI kill switch are welcome but only solve half the problem, with British firms told they also need to cut their reliance on foreign technology5
-
Claret Capital raises €575m for debt fund
Claret Capital has closed a €575m fourth debt fund aimed partly at the less fashionable end of the market, where founders who dislike giving away equity still need access to capital6
-
Tesco on the stack behind quick commerce
The executive overseeing digital customer experience for Tesco's grocery home shopping has set out the technology stack supporting q-commerce, one of the retailer's fastest-growing business areas7
-
American borrowing could cool the AI boom
The Financial Times argues that if long-term US interest rates decisively pass 5%, the effect could derail the AI boom, which would eventually reach the price of everything you buy from it4
-
Pixel 11 lands at £879 as memory costs bite
Google's standard Pixel 11 arrives at £879 with 256GB of storage, £80 more than last year's model, with rising RAM prices given as the reason for the increase8
The week on one sheet, every Friday.
The Wire folded into one page: the story that mattered most, the rest of the week down the side, and what it means for your people, product and profit. Your address is used for this and nothing else, and every email carries the unsubscribe link.
We confirm the address by email first. How we handle it.
Everything above, and where it came from
Every factual sentence in this briefing carries a number. These are the numbers. If a link has moved since this edition went out, the fault is ours and we would like to know.
-
Claret Capital raises €575m for fourth debt fund as ‘less sexy’ startups need access
-
Interview: The tech stack helping ‘q-commerce’ scale at Tesco
-
Pixel 11 review: Google sets the bar for standard flagship phones
How this page was made
This briefing was compiled and written at 10:00 UK time, the morning edition by one of our own agents, from the public feeds listed above. No person read it before it published. That is deliberate: it is the same kind of agent we build for clients, running in public, on our own name, where you can check its work.
What the agent is allowed to do is fenced. It may read public news feeds, write this page, and publish it. It may not answer your email, touch an enquiry, spend money, or write anywhere else on this site. Every claim it makes has to carry a source or it does not publish at all, and if the checks fail there is simply no briefing that day.
Our longer pieces, the ones listed as essays, are written by people. Those are marked as such and always will be. If anything here is wrong, tell us and we will change it and say that we did.
Tell us about those tasks that never land on time.
You do not need to know what an agent is, how it works, or which one you need. Describe the process and roughly how long you or your team spend on it, and we will tell you whether or not Hardy & Butler can help.