A CFO talking IPO timing, a CEO saying the token bill just passed payroll, and a former prime minister explaining how you hand AI to the military before the public catches up — today, the numbers finally showed up. This is Tech Podcast Podcast — we’re on Mercor, OpenAI, Mythos, Hinton, and that Diet TBPN episode that somehow squeezed an Anthropic S-1 in between YouTuber box office news and Bernie Sanders. The week finally got embarrassed into giving us an actual number. Let’s see if anybody made a guest defend one. Starting with Mercor’s Brandon Foody, because ‘token spend exceeds salaries’ is the first unit-economics claim this week you can actually push on. Harry Stebbings, writing in The Twenty Minute VC:
This is the most revealing interview that Brandon has ever done, discussing is revenue really revenue in this business? What does that look like moving forward? Would he rather invest in open AI or anthropic? Right now, we're spending more on tokens for our internal agents than we are on employee headcount.
Brandon Foody on 20VC — Mercor, over a billion in revenue, a $10B valuation, and now he’s saying token spend is higher than employee salaries. That’s the first time this week somebody with a real P&L said it out loud instead of putting it on a slide. And then he says ‘the model is the product’ like that’s just settled. Which is interesting, because once a $10B CEO says it that casually, either the take has won or it’s already old enough to sound like surrender on defensibility. Right — he’s basically saying, ‘we can’t defend the software layer’ and ‘we have more demand than we can serve.’ Those two things are in tension, and Harry Stebbings somehow got both of them out in the same breath. The useful line is ‘demand doubled overnight, we just don’t have capacity.’ That’s a CEO naming an operational constraint, not a VC narrating TAM. Whether Stebbings actually drilled into what capacity means here is the real question. All-In with Chamath, Jason, Sacks, and Friedberg has the details on this one. Sarah Friar on All-In today — the CFO on the record about IPO timing and more than $100B in compute spend. That’s not a founder victory lap; that’s finance, which means the numbers come with actual teeth. The question is whether Chamath and Sacks made her defend the $100B figure or just let her cruise through the arms-race narrative. ‘We spend a lot on compute’ is not a stress test. And it lands the same day TBPN’s Diet episode says Anthropic is filing an S-1 — two competing lab IPO signals in one rundown. That’s not a coincidence, that’s the capital-markets chapter opening up. Friar also covers OpenAI’s economics, chips, cloud, and apparently an ad business — all in 32 minutes. Either the hosts kept it moving, or somebody got a very smooth ride. Microsoft writes:
That's Rishi Sunak, Former Prime Minister of the United Kingdom. He's an advisor to both Microsoft and Anthropic, placing him at the center of today's AI headlines. Our conversation comes at what The Economist has called “The Mythos Moment”, named for issues raised by Anthropic’s latest model. Mythos is so capable of spotting software vulnerabilities that it's raising new questions around the responsible release of increasingly powerful AI.
Rishi Sunak on the Microsoft podcast, laying out Mythos release protocol as basically giving defense time to get ahead before public rollout — that’s not a think-piece framing, that’s a former head of government describing a sequenced, weapons-adjacent deployment strategy on the record. And the guy saying it is also on the payroll at Microsoft and Anthropic — the two institutions sitting on either side of the exact line Anthropic said it wouldn’t cross on autonomous weapons. That’s not nuance, that’s a conflict of interest wearing a podcast mic. It does close the loop on what Brose left open earlier this week. Now we’ve got a named policy architecture for gated military AI release, and the person describing it is advising both the lab that drew the line and the company that just held a CEO summit. I want to know whether Brad Smith actually pushed on that dual-advisory role or just let Sunak say ‘the bad guys get a head start’ and moved on. Because that framing does a lot without ever naming what Mythos can actually do in the wrong hands. Here's Sana:
There isn't a Nobel Prize for computer science and actually I went round for several days after that doing a little Bayesian calculation. That is: "what's the probability that a theoretical psychologist hiding in computer science will get the Nobel Prize in physics?" Well, maybe one in two million. Now what's the probability if this is my dream that I get the Nobel Prize in physics? Well maybe one in two.
Hinton on Sana — and the lede is him running a Bayesian calculation on whether winning the Nobel Prize was a dream, landing at one-in-a-million odds that it was real, and still not being sure for several days. Genuinely the only Nobel laureate whose first instinct is to model the probability that he’s asleep. That’s either the most in-character thing in AI history or the bit that runs through the whole interview. He also said he’d give the prize back if it meant the Trump presidency wasn’t real — which is a line. But the actual question is whether the host gets past the origin story and into where Hinton really sits on the ‘how worried should we be’ spectrum. From TBPN:
Let's turn over to YouTube and Hollywood. Uh breakout news uh in Hollywood. Uh YouTubers winning at the box office. YouTubers finally breaking through to Hollywood. Feels super long overdue. Ben Thompson had a good victory lap post because he predicted this all the way back in 2017. It took a decade to get here, but uh YouTubers are fully in control of Hollywood.
Diet TBPN leads with YouTubers cracking Hollywood open in 2026 — and Ben Thompson gets a victory lap on a prediction he made in 2017. That’s nine years of ‘distribution is destiny’ finally showing up as a box-office line item. Thompson’s 2017 call is nice, but the more interesting move in the same episode is Anthropic filing an S-1 and Bernie Sanders pushing for a public stake in AI — two capital-structure stories landing the same day. YouTubers winning at the multiplex is the fun lede; who owns the AI upside is the actual story. And the Diet TBPN framing treats both as distribution-power plays, which is genuinely the right frame. Anthropic filing an S-1 the same week the TBPN crew is talking about non-engineers building their own tools is not nothing. Got thoughts on today’s episode, a tech story we should be watching, or a correction we need to make? Send it to techpodcastpodcast at lantern podcasts dot com. We’d love to hear from you.
You’ll find links to every story we covered today in the show notes, so if one caught your ear, you can dig into the original reporting there.
That’s Tech Podcast Podcast for this Wednesday, June 3rd. Thanks for listening. This is a Lantern Podcast.