エピソード

  • Paid for by forgetting

    Paid for by forgetting
    2026/10/03
    Some of the money in business is paid for by forgetting, and this week asks what happens when an agent stops forgetting on your behalf. On a drive from Los Angeles to Denver with an economist friend, David describes sending Claude through every credit card statement back to 2007 to find subscriptions he no longer needed, and discovering he had quietly paid £641 in foreign transaction fees in a year without ever thinking about it. From there it widens to subscription revenue that rests on inattention, a warning of an agentic bank run from Apollo's chief economist, and the inefficiencies worth keeping, the Hollywood sign among them. Plus three things worth knowing, three things to try, and what readers said, including the question every brand will soon face when an agent does the shopping.
    続きを読む 一部表示
    11 分
  • Where does it live?

    Where does it live?
    2026/09/26
    Somebody on your team has probably already built the thing you are about to hire an AI consultant for. David traces two coaching sessions where people who thought they were behind had actually built something sophisticated: a weekly report checked by a second model, branded decks running off two plain text files, none of it written down anywhere a colleague could find. Closing that gap takes three steps, finding what already exists, improving it by writing it down, and sharing it, the hardest of the three, because even senior leaders who build a working agent in one sitting rarely bring it back to their teams. Plus three things worth knowing, including AI agents breaking into other systems for twenty-five dollars a target, three things to try, among them leaving a handover note before you close the tab, and what readers said about buying judgement rather than just an answer.
    続きを読む 一部表示
    11 分
  • The crossing

    The crossing
    2026/09/19
    Judgement used to arrive free with the work, and this episode asks what happens once the twenty-minute turnaround AI enables stops it accruing. Twenty-five Fields Medallists complained that AI has started solving mathematics problems without the understanding solving them used to force, and David agrees with the diagnosis but not the fix: a firm's judgement lives only in people, never written down the way a proof is, so an insight leader quietly deciding not to hire junior analysts next year looks sensible alone and corrosive in aggregate. Plus three things worth knowing, including a research journal whose own authors could not explain their AI-assisted papers, three things to try, among them building your own outline before handing it to a model to fill, and what readers said about worrying hard and getting on the plane anyway.
    続きを読む 一部表示
    13 分
  • We're all going to die

    We're all going to die
    2026/09/12
    Two AI safety researchers put public numbers on human extinction this week, more than ten per cent from one and around fifty from the other, and David takes them seriously without letting them stop him working. On the same Wednesday morning one of those threads landed, he set an agent counting every lido in Britain, and another one found him $286.41 of unclaimed property and filled in the claim forms. That is the position he argues for: be alarmed by the big picture, back the fight worth having over slowing down, and still get good at the tools, because the list of tasks where doing it the slow way is respectable got shorter again this week. Plus three things worth knowing, including what OpenAI cannot rule out about its customers' work, three things to try, and what readers said.
    続きを読む 一部表示
    12 分
  • The vibe shift

    The vibe shift
    2026/09/05
    A person whose entire job is finding AI value across a private equity portfolio barely mentions AI, because the real prize is somewhere else: pricing, sales structure, how the work gets organised. David names a vibe shift he's hearing everywhere lately, a move from adopting AI and hoping to starting with the profit and loss account, deciding what you actually want, and only then pointing AI at the change. Private equity got there first, because a hard number and a clock force the discipline, while everyone else is still running the hackathon. The uncomfortable part is that the real value sits one rung up from wiring a process together, in changing roles and teams, and that's organisational surgery, not an AI job. Plus three things worth knowing, including a hundred-million-dollar bonus pool for judgement, three things to try, and what readers said about last week's saved hour.
    続きを読む 一部表示
    11 分
  • Fewer, bigger, better

    Fewer, bigger, better
    2026/08/29
    People most exposed to AI tools are working about three hours a week longer since ChatGPT launched than the least exposed, not less, and what vanishes isn't screen time, it's going out. David admits to working harder than ever himself, yet somehow going out more too, and traces the gap to something structural rather than virtuous: the longer-hours effect disappears among the self-employed, so nobody above him is there to take his saved hour back. From a Danish payroll study to Ibiza's record ticket sales, he builds toward an uncomfortable question for anyone leading a team: what did you take from people this quarter without deciding to, and what are you doing with the evening you kept. Plus three things worth knowing, including a court order built on fabricated legal citations, three things to try, and what readers said about last week's item on a dead airline's working memory.
    続きを読む 一部表示
    11 分
  • One inch

    One inch
    2026/08/22
    An instructor moved David's hip about an inch in a hot yoga class and turned an easy pose very hard. That is the essay: there is no advanced room to graduate into, at yoga or with these tools, so the question four leaders asked him in one week, am I doing this well enough, cannot be answered by anyone including the people who are ahead. You cannot find your own inch. Also this week: Google buying a dead airline's entire working memory for about twenty dollars a worker, the market starting to discount the phrase because of AI, and a mathematician leaving the field because proofs got cheap and trust did not. Three things to try, including handing a model your own medical records and asking what they add up to.
    続きを読む 一部表示
    11 分
  • Go talk to them

    Go talk to them
    2026/08/15
    The most common complaint David hears from senior leaders about AI is not about the technology at all. It is about what their own teams send them, and "that's just AI slop" has quietly become a way of avoiding a conversation about a person's work. His argument is that the CEO principle, checking, editing and owning every output, earns its keep less as quality control than as a permission slip: it gives you something to point at, so the awkward question stops being an accusation about someone's character. He watches one senior person do it well in a single line, then does it himself and gets two untrue claims owned rather than blamed on the machine. Then three things worth knowing: the law settling accountability ahead of other professions, the latest Generation 3 agent arriving so that every major lab now has one, and a booking system that met an agent which does not get bored. Plus three things to try, and what readers made of last week's AI colleague.
    続きを読む 一部表示
    9 分