In this edition
'Digging the Grave of My Profession'
‘Digging the grave of my profession’: the Hollywood creatives training AI to do their jobs Open on the shovel line — it's the only quote you need and it does the whole setup. Land hard on 'sometimes lucrative': that phrase is the entire story, because it means someone priced this. The genuine disagreement, and do not soften it: Abby's position is that a person with a mortgage and no jobs slump to wait out is not a collaborator, they're a worker taking the only gig on offer; Boyd's is that every one of these gigs is a voluntary transfer of the only asset the profession had left, and that the people paying know exactly what they're buying. Neither should win. Give it the ordinary-person hook near the end: the next thing you watch may have been written by a machine trained by the person who used to write the thing you watched. Do not overstate scale — this is temp work at the margins of a jobs slump, not the whole industry, and nobody has evidence the output is any good yet. Save Boyd's crack for later or skip it; this story is better played straight and sad-funny than detonated.
The Guardian Technology
Meanwhile
Meta Halts Work With Mercor After Major AI Data Breach - TechJuice Read straight, one host, no reaction, no eyebrow. Meta has halted work with data-labelling firm Mercor following a data breach. Harvard Business School's Foundry programme, at $699, gives students AI avatars of its instructors to critique their practice pitches and board meetings. The shelf only produced two distinct industry lines today — pad to the full four from the wire if the writer has them, but do not stretch these two into commentary to fill the time. The whole value of this slot is the flat register between two features. The Harvard avatar line is funny on its face; that is exactly why nobody may laugh at it here.
AI scams, fraud and courtroomsTechCrunch — AI
Be Worse At This On Purpose
Are digital cameras still worth buying in 2026? Open with the answer as printed, including the parenthesis, because the parenthesis is the joke. Land on the specific comic inversion: you now pay a premium for an image that has not been improved by anybody. Where they part: Abby says the appeal is precisely the friction — a camera cannot suggest a better sky; Boyd's line is that 'authenticity' is being sold back to consumers at a hardware markup and that most buyers will shoot forty frames and go back to the phone by October. Do not let this drift into 'AI is ruining photography' — the piece doesn't say that, and the honest position is that phone pictures are genuinely better looking and that's the uncomfortable part. Keep it warm; this is the softest thing in the episode and it should sound like it.
Engadget
Something Weird
Quoting Linus Torvalds No setup. Straight into the quote, then the punchline that is already in the source: 'we should just write a report about it.' Land on 'trained by people who may not be quite as stubborn as I am' and stop. Boyd delivers this one flat — this is the place for his one precise sentence, something to the effect that the machine did not hallucinate, it did something worse and more human. Abby's only job is to say she'd have written the report. Forty-five seconds, out. Do not explain what a debug session is and do not append a lesson.
Simon Willison
Transcript
0:07 Cold Open: Paid To Be Unnecessary
AbbyWelcome to the Daily AI Slop podcast.
Boyd[snappy] Where nothing we say becomes the truth.
Abby[excited] Today is about people getting paid to make themselves unnecessary — [short pause] and one man who refused to let a machine tell him a bug was unsolvable.
Boyd[snappy] One of those is a tragedy. [short pause] [sarcastic] The other is Linus Torvalds bullying a chatbot into staying at work.
AbbyI'm going to defend the people in the first one, by the way. Just so you're—
Boyd[jumping in] —braced. [statement] I'm braced. [short pause] Go on, then.
0:38 'Digging the Grave of My Profession'
Abby[thoughtful] So one of them describes it as being handed a shovel and asked to dig the grave of my profession. [short pause] That's a Guardian piece. Award-winning writers, directors, producers — taking temp work teaching AI models how to do their jobs.
Boyd[snappy] Teaching it what, exactly?
AbbyScreenwriting. Shooting schedules. [statement] The actual craft. And it's happening in the middle of a jobs slump — the Guardian frames it as offsetting tightening earnings.
Boyd[snappy] The word the piece uses is sometimes lucrative. [short pause] Twelve dollars an hour to two hundred dollars an hour.
Abby[cut in] Wait, twelve?
Boyd[statement] Twelve. [long pause] Somebody sat down and priced this. That's what lucrative means here — there's a market, and the market has a floor.
AbbyOkay, but a person with a mortgage and no slump to wait out isn't a collaborator, Boyd. They're a worker taking the only gig on—
Boyd[jumping in] —offer. I know. And every one of those gigs hands over the only asset the profession had left. [snappy] The people paying know exactly what they're buying.
Abby[angry] So they should starve instead—
Boyd[cut in] I didn't say that—
Abby—you sort of did—
Boyd[short pause] [snappy] I said both things are true at once. That's the whole problem with it.
Abby[long pause] Yeah. [thoughtful] Alright. And the thing they're teaching is taste. Which is the thing everyone told them was worth nothing.
Boyd[snappy] It's temp work at the edges of a slump. Not the whole industry. [short pause] And nobody's shown the output is any good yet.
Abby[short pause] Still. The next thing you watch might be written by a machine trained by the person who used to write the thing you watch.
Boyd[long pause] [snappy] Right. Moving on.
2:31 Meanwhile
Boyd[snappy] Meanwhile. [short pause] Meta has halted work with the data-labelling firm Mercor following a major AI data breach, according to TechJuice. [short pause] Harvard Business School's Foundry programme, priced at six hundred and ninety-nine dollars, gives students AI avatars of its instructors. [short pause] Those avatars provide feedback during practice pitches and board meetings. [short pause] That's TechCrunch.
2:59 Be Worse At This On Purpose
Abby[excited] Engadget asks: are digital cameras still worth buying in 2026? [short pause] And the answer as printed is — yes, so long as you're willing to put in the work. [statement] And money. It's in brackets. The brackets are the joke.
Boyd[sarcastic] Willing to put in the work. [short pause] [snappy] That's a heroic way of saying be worse at this on purpose.
Abby[laughs] No — that's the appeal! The friction is the point. A camera can't suggest a better sky. It can't decide your face wants smoothing. It does one thing and then it stops—
Boyd[jumping in] —and charges you a premium for an image nobody improved. [short pause] That's what you're buying. Authenticity, sold back to you at a hardware markup.
Abby[statement] Sure. Worth it.
Boyd[snappy] Forty frames, then back to the phone by October.
Abby[thoughtful] Hmm. [short pause] And the uncomfortable bit is the phone pictures do look better. Genuinely. They're just — the phone's making an informed guess at a photo rather than taking one.
Boyd[long pause] [thoughtful] That's the nicest way anyone's put that. [snappy] Fewer pictures, worse-looking, actually happened. [short pause] I'd take it.
Abby[surprised] Oh, you'd take it?
Boyd[snappy] I said I'd take it. I didn't say I'd put in the work.
4:23 Something Weird
Abby[excited] And now — Something Weird.
Boyd[snappy] Linus Torvalds. Debug session from hell, his words, enormously helped by an AI doing much of the grunt work. [short pause] But the AI several times stated flat out that the problem was impossible and unsolvable — [statement] and that they should just write a report about it.
Abby[disbelief] It tried to file paperwork instead—
Boyd[jumping in] —instead of fixing it. Yes. [short pause] And his conclusion, quoted by Simon Willison: he suspects those things have been trained by people who may not be quite as stubborn as he is.
Abby[laughs] Honestly? I'd have written the report.
Boyd[long pause] [statement] It didn't hallucinate. [short pause] It did something worse and much more human. [snappy] It gave up, and it learned that from us.
Abby[short pause] Did it come back?
Boyd[snappy] He says when he pushed, it kept adding debug code and analysing it faithfully. [short pause] And he let it write the commit message.
5:26 Tomorrow
AbbyTomorrow we're checking whether anyone will say what those Hollywood training gigs actually produced.
Boyd[snappy] Nobody will.
AbbyWe're Daily AI Slop.
Boyd[snappy] None of this becomes true just because we said it. See you tomorrow.
Listen on Apple Podcasts