Today, the AI ROI question finally gets a named economist on the record — and his answer is the dinosaurs looking up at the meteor. This is Tech Podcast Podcast. Three shows, three completely different answers to whether enterprise AI actually pays — and we're mapping all three. Plus, SpaceX slapping a fixed $135 a share on a $75 billion IPO. First hard number I've seen all week that isn't a projection. And DJ Patil and Tim O'Reilly put a name to a downstream cost nobody else touched. We'll start there. Patil's point is concrete: PhD and med school slots are vanishing because the funding's gone, while the same institutions tell students to go spend more on tokens. Right — where are the dollars for the tokens? That's the question nobody bothered to ask in the DeepMind episode while Kilpatrick was telling everyone to reset their ambition every three months. Second time this week we're hearing the two-to-three-month cadence from Logan Kilpatrick. So is that the durable developer framework, or are we getting a rerun? It's confident vagueness on loop. 'Which skills really matter,' 'somebody should build that company.' Cool — somebody being who, and with what budget? And that vagueness is exactly where Kedrosky walks in on Better Offline. He's the skeptic side of the split — sitting right next to Bradway's 'agentic forms are magic' from Amgen yesterday. The pharma CEO finally gets his denominator, and it's an economist saying there isn't one. Foday said tokens already exceed salaries. Bradway called it magic. Kedrosky says the math doesn't close. Same question, three very different answers from the field. And the cleanest tell might be SpaceX. A founder rejects Wall Street's whole price-discovery game with a fixed $135 the same day economists ask if AI returns anything. One person willing to put a real number down. It's contrarian by design. Put it next to the IPO signals from Anthropic and OpenAI this week, and the capital story has three named data points now instead of just vibes. Then Project Lightwell circles back on IBM Technology — and they finally say the quiet part: some of these projects have zero or one part-time maintainer. It's our second pass at the burnout framing. The clearinghouse model's there, but whether it actually fixes upstream incentives is still open. We're going to trust the model on top of code one guy maintains on weekends. Sleep tight. Here's O'Reilly:
There are plenty of students that I have talked to who are supposed to be going to a doctoral PhD program or a medical school or something like that, the slots aren't there because of the overall budget impacts. And so whether you call it AI impact or economic reframing, the thing is broken. And then we're saying to them, hey, well, you should learn more about AI. You should burn a lot more tokens.
Finally. DJ Patil names an actual cost — PhD slots and med school seats are just gone because the budget's blown up, and the same institutions are telling those students to go burn more tokens. And he asks the thing nobody on the productivity-pitch shows asks: where are the dollars for the tokens? Where are the dollars for the internet connection? What I like is how concrete it is. All week, the AI-and-jobs talk has been abstract displacement — here it's a specific broken pipeline. The slot's not there, the funding's gone, and the advice is spend more. And Patil's the right person to land it — he co-coined the term data scientist and was the first federal chief data scientist. He's not a guy who reflexively bets against the tech. So when he says the thing is broken, that carries. His best line is the trust one. You told these kids college was where the puck was going. Now you're saying skate somewhere else. Why should they believe you the second time? Beyond Coding, with Logan Kilpatrick:
There's so much interesting innovation that's happening in this way that like was not happening three years ago. Every two weeks, every three months, every six months, what's possible changes. You have to reset your level of ambition. Somebody should build that company. Which skills really matter nowadays? The rules of software engineering have changed.
Logan Kilpatrick's back, second time this week, still telling developers to reset their ambition every two to three months. Either that's the most durable framework on this beat or it's a rerun, and I'm leaning rerun. What gets me is the cadence claim never comes with a denominator. Reset your ambition every three months — okay, against what budget? Who's paying for the tokens that ambition runs on? Right, and the host opens with 'I love your keynote yesterday.' That's the whole temperature of the interview. Nobody on Beyond Coding is going to put the ROI question in front of a Google DeepMind product lead. And it sits in the same rundown as Kedrosky calling enterprise AI a meteor-and-dinosaurs moment. Same week, two episodes, and neither one has to face the other. That's the fault line. 'Somebody should build that company.' Confident vagueness — same posture as 'the model is the product.' Easy to say when you're not the one writing the check. Better Offline, with Ed Zitron:
Everyone's uh deeply upset because this week and the last week everyone has been saying, "Huh, does AI have a return on investment and it's I've really been enjoying it because it's like watching the dinosaurs look up and see the meteor." They're just like, "What do you what do you mean? What do you mean this costs money?"
Finally. Ed Zitron books an actual economist and opens with 'watching the dinosaurs look up and see the meteor.' That's the bluntest the ROI question has gotten all week. And Kedrosky doesn't hand-wave it — he triangulates through the GitHub Copilot study, the Peterson Institute piece, and Anthropic's Jack Clark asking where AI shows up in GDP statistics. Which lands the same day Logan Kilpatrick's out there telling developers to reset their ambition every three months. Two episodes, two universes, and neither host has to look the other in the eye. What I want is the actual number behind the GitHub study. Zitron's good at the meteor metaphor — I need Kedrosky to stay on the P&L. That's the gap between a clip and a finding. Remember the Amgen CEO calling agentic form-filling 'magic' yesterday? This is the denominator that pitch never had. Magic for how much, against what return? From Bloomberg:
Elon Musk rejects another Wall Street convention and sets a fixed price for his SpaceX IPO ahead of the marketing phase of the deal. The plan, offer shares at $135 a piece.
SpaceX, $135 a share, fixed. Musk is just walking past the whole book-building circus and slapping a price tag on it. Finally, a number today with someone making a decision instead of floating a projection. A $75 billion target at a fixed price — that cuts against how Wall Street does price discovery, and it's deliberate. The bankers don't set the range; he does. And it lands the same day economists on Better Offline are asking whether AI returns anything at all. One founder pricing himself with total conviction while Kedrosky's out there watching the dinosaurs look up. Stack it up — Anthropic and OpenAI signals earlier this week, now SpaceX at a fixed $135. Three capital-markets data points with names on them, and SpaceX is the one rejecting the convention everyone else plays along with. Meanwhile Alphabet quietly upsizes its raise to $84.75 billion to feed AI infrastructure. So the money's flowing in regardless of whether anyone's nailed down the return. From IBM Technology:
IBM and Red Hat announced Project Lightwell. Now, this is a $5 billion commitment from IBM and Red Hat, and they're aiming to elevate the security posture of the open source ecosystem as a whole by establishing a trusted enterprise clearinghouse and a team of 20,000 AI augmented engineers to streamline the process of reporting and resolving vulnerabilities, deploying validated patches, and coordinating upstream disclosures.
Okay, here's the line that actually lands: 'some projects have zero or one maintainer working part time.' That's the whole open source supply chain in one sentence. And IBM and Red Hat's answer is a five-billion-dollar clearinghouse with twenty thousand AI-augmented engineers. Yesterday, we left open whether Lightwell changes the incentives for those upstream maintainers — now we have the maintainer-burnout fact, but the incentive question's still wide open. Right — you can fund twenty thousand engineers to triage CVEs, but does one cent of that five billion land in the part-time maintainer's pocket? Because that's the person who burns out. And then there's the panelist line — 'trust the model, not the code.' On a security show. That's a remarkable thing to say out loud while announcing AI-augmented patching. Trust the model, not the code, said the cybersecurity podcast. Bold. Got a story we should be watching, a correction, or just a thought on the show? Send it our way at techpodcastpodcast at lantern podcasts dot com. We read every note, and it helps shape what we cover next.
We've put links to every story from today's briefing in the show notes, so if one of them deserves a closer look over the weekend, you know where to find it.
That's Tech Podcast Podcast for Friday, June 5th. Thanks for listening, and have a great weekend. This is a Lantern Podcast.