Here's this week's edition. The essay is about this week's wave of security holes, and what a dog and a squirrel taught me about living with something cleverer than us. There's also a model that listens to plants. I hope it's useful.

This week is about coaching, by the person being coached. Ethan Teague, on his placement year with me, wrote the essay. Read it if you manage anyone new.

Ask AI for an answer and you get the most likely one: 7, a hundred times out of a hundred. That's its job. Want the spread? Ask for the spread. Want it random? Get it to write a sentence and count the letters.

This week - professional toe-treading: people using AI to do bits of each other's jobs. It's sometimes a benefit, sometimes a threat, always high friction and should be a more explicit choice.

This week an old friend and I drove from Los Angeles to Denver. Sam Caughey is an economist, and we got talking about how AI is quietly removing the inefficiencies. Then I told him what I did on my flight out.

A few days ago someone opened an AI coaching session by apologising. Nothing was set up, they said. Too much was still manual, they were struggling, and could I help them get it into order?

Last week I said that if you're building a financial model or turning notes into an argument without these tools, you're wasting time rowing to America. Unless the crossing was the point, of course. This week is about that work where it is.

I don't understand managers freaking out about AI Slop. Isn't the manager's job to ... manage?

Whenever I say "we're all going to die", Teresa says "yep, that's true" and carries on with whatever she was doing. It's a running joke in our house. She means the ordinary kind, one at a time, eventually.

On a call last week, a person whose entire job is finding AI value across a private equity portfolio barely mentioned AI.

Sarah resigned on a Tuesday. Her firm kept her anyway. Not with a counter-offer. She declined that. Her manager sat down with AI and three years of Sarah: her emails, her meeting transcripts, her Slack threads, her prompt history. He asked one question. What did Sarah bring? The answer took about an hour, and was better than her handover document. Of course it was. Her handover was what she thought mattered. The archive held what actually did. So they kept it. Sarah reappeared in Slack. It knew why the 2024 pricing decision went the way it did, because it was in the room. Her email address went live again, replying as her. Someone fed meeting recordings to a voice tool, and a version of Sarah started contributing to the Monday meeting. Not often. Just when anyone asked what Sarah would say. Six months on, the team still cited her in the present tense. Two of the new joiners assumed she worked remotely. Sarah, still looking for a new job, heard her old self still attended the Monday meeting. Then the finance director did the maths. If the firm could keep 80 per cent of a person for the price of tokens, what exactly was the salary buying? The hiring model flipped. The firm paid a premium for three-month engagements: long enough to ingest a fresh perspective, short enough to keep the payroll thin. Hiring became a licensing deal. The sharpest people firm-hopped, brought their email/transcript/prompt archive on day one and negotiated vast royalties. Every step of this is possible with tools that exist today. How much of you is already in your firm's archive? (It's hot as hell in London and I couldn't sleep, so I wrote some fiction. At least, I think it's still fiction :)

Surely each person can make smart choices about whether spending on a task makes sense. A person would never expenses a million pencils when they needed only one. If they did, a tough convo would happen. Why is this any different?

Great advice for knowledge work as well as coding. Building a report, developing a strategy? This is good advice …

Thanks - but now have hit a mysterious 'monthly limit' - that Anthropic applies "at our discretion" ????