エピソード

  • Futures and fears

    Futures and fears
    2026/07/18
    Fear of AI and excitement about it aren't competing conclusions; they're both smart readings of the same facts, and the same people often hold both at once. David gathered the fears readers sent in after last week's Sarah story and set them beside three futures: a chief executive who has given up reconciling the two sides, an investor enjoying a brief window of arbitrage, and an L.E.K. panel with Rob Wild and Scott Breitenother. His own answer is uncomfortable for an optimist: no shiny, democratised future is coming, because the gains keep compounding in the few firms willing and able to do the work of reorganising around the technology. The fears that wouldn't sort into a 'machine' pile or an 'us' pile were all about power: who owns what you depend on, and who's left outside when the window closes. Plus three things worth knowing, three things to try, and what readers said.
    続きを読む 一部表示
    12 分
  • Sarah stays

    Sarah stays
    2026/07/11
    A story arrived nearly whole on a hot, sleepless London night, so David got up and wrote it: Sarah resigns on a Tuesday, and her firm keeps her anyway, rebuilt from three years of her emails, transcripts and Slack, because the archive holds what her handover never could. Every step is possible with tools firms already own. The reader comments chilled him more than the writing did, and they shifted something: after years spent on what AI makes possible, he wants to spend more time on its downsides, and asks what would keep you up at night. Alongside, Gartner's finding that cutting jobs for AI did not pay off, a Nobel economist's doubt that AI will repeat the PC-era boom, and the Brown class that scored perfect hundreds at home then collapsed in person. Plus three things worth knowing, three things to try, and what readers said.
    続きを読む 一部表示
    13 分
  • Look at the grass

    Look at the grass
    2026/07/04
    Two photographs of the same Wimbledon court forty-two years apart tell the whole story: in 1982 the grass is worn where players served and rushed the net, and by 2024 the bare patch has moved to the baseline because the game became a rally. David uses that shift to argue knowledge work has made the same move, only in four years instead of forty-two. Producing got cheap, so the effort moved to framing the shot and reviewing what comes back, and yet most firms still run review cycles and approval chains built for the old game. He tests it against a Cannes leader drowning in good-enough drafts from a team of 250. Plus three things worth knowing, three things to try, and what one reader said about quality versus handwriting.
    続きを読む 一部表示
    13 分
  • A thousand small bargains

    A thousand small bargains
    2026/06/27
    David sent an important email last week that a machine wrote, read it, changed nothing, and sent it — one of three handovers that look, on their face, like exactly the surrenders the worriers warn about. None of them were. Taking Rahim Hirji's new book SuperSkills as a generous foil, he argues most AI handovers are good bargains, not a thousand small surrenders dressed as convenience: the machine takes the middle while the parts that decide the outcome climb a level to you — what to ask, whether it's good enough, whether you'll own it. The floor isn't to check every word but to own it to your own standard, and to save your deepest effort for the few tasks where being great beats being good. Never be careless, always be good, sometimes be great. Plus three things worth knowing, three things to try, and what readers said.
    続きを読む 一部表示
    14 分
  • Average by default

    Average by default
    2026/06/20
    Almost nobody who sets up their AI ever tells it who they are. Not just the job title, but how they think, what they notice, how they decide. David Boyle argues that the empty personalisation box is the most useful question the whole product asks, because if you don't tell the model what makes your judgement specific, it has one assumption left: you're average. He builds the case on fresh research showing leading models land on the same argument while people diverge, and on his own work writing down an identity layer a CV can't hold. The edge is the judgement and taste you bring, and it travels with you between systems. Plus three things worth knowing, three things to try, and what readers said about spending, budgets and trust.
    続きを読む 一部表示
    12 分
  • Ride the bike

    Ride the bike
    2026/06/13
    Anthropic's newest model costs exactly double its predecessor, one GitHub Copilot bill jumped from a flat $50 a month towards $3,000, and suddenly the invoice, not the model, is the story. David argues most organisations manage these bills exactly backwards: they celebrate the biggest token burners or cap everyone, and both approaches manage the number instead of the judgement. His maths says the gap between the cheapest sensible model and the dearest buys about 40 seconds of a manager's day. What he would do instead: a floor of five prompts a day for everyone, then a delegation budget run on trust, because a price with a budget behind it sharpens judgement while a price with a cap replaces it. Eddy Merckx supplies the moral: ride the bike. Plus three things worth knowing, three things to try, and what readers said about graduate hiring.
    続きを読む 一部表示
    13 分
  • The open door

    The open door
    2026/06/06
    David finally does the thing he has spent years telling other people to do: he hires a graduate. Ethan joins for a placement year, taken on as an experiment, because the awkward truth is that a graduate today is more capable than ever and less needed than ever. David sets out four hypotheses for why a young person is still worth it: someone has to check and own what the machine makes, managing a machine is real work, it builds judgement fast, and an open door lets useful things in. He borrows Richard Hamming's open-door idea and makes the moral case for keeping the door open, while the front door into work has rarely been harder to push. Plus three things worth knowing, three things to try, and what readers said.
    続きを読む 一部表示
    12 分
  • How We Got Here

    How We Got Here
    2026/05/30
    David walks through how AI got practically useful, grounded in Dwarkesh Patel and Gavin Leech's The Scaling Era. The through-line: the future of AI has shown up long before it shipped, every single time. Seeing wasn't the hard part; believing it enough to bet on it was. He revisits the early scaling-law moments that should have been obvious to outsiders, asks why they weren't, and what today's people who use language models in their work should be willing to bet on now. Pair with next week's Edition 16, written with Rob Wild at L.E.K., on where things go from here. Plus three things worth knowing, three things to try, and what readers said.
    続きを読む 一部表示
    14 分