Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Back to Podcast Digest
More OpenAI C-Suite Drama, Is Siri Seriously Broken?, Meta’s Elusive Next Hit
TL;DR
OpenAI’s CFO fight looks normal until you add the numbers — Alex Kantrowitz and MG Siegler say CEO/CFO tension is common, but it hits differently when Sam Altman is reportedly committing OpenAI to $600 billion in spending over five years, eyeing a Q4 IPO, and the company may burn more than $200 billion before generating cash.
The quiet red flag is revenue growth, not just executive drama — both hosts zero in on The Information’s note that OpenAI’s revenue growth has been slowing, because that weakens the exponential-growth story just as Anthropic closes the annualized revenue gap from about $10 billion vs. $20 billion late last year to roughly $19 billion vs. $25 billion now.
OpenAI and Anthropic are converging on the same “super app” thesis — after taking different paths, both now seem to be building all-in-one products that combine chat, coding, and agentic computer use, with MG calling it a sprint for OpenAI to fuse ChatGPT, Codex, and possibly its Atlas browser before Anthropic’s momentum hardens.
Apple finally seems ready to admit AI needs a real app — Mark Gurman’s reporting about a standalone Siri app, Dynamic Island placement, and Action Button access is framed as a meaningful shift, because Apple is abandoning the old “AI should stay invisibly in the background” idea and conceding users now expect a ChatGPT-style destination.
Apple’s real AI risk isn’t just product delay — it’s cultural drift — MG argues that outsourcing Siri’s core intelligence to Google Gemini may help Apple ship something, but it could leave the company without the internal DNA, urgency, and technical muscle required to compete in a field moving faster than any prior platform shift.
Meta still looks like a company with one great business and no clear next act — despite the metaverse push, open-source Llama strategy, Scale AI deal, and “superintelligence” recruiting blitz, MG says Meta keeps failing to create a true second hit beyond ads, raising the uncomfortable possibility that Zuckerberg’s nightmare scenario is becoming real.
The Breakdown
OpenAI’s latest boardroom mess has a very expensive backdrop
The episode opens with fresh reporting that Sam Altman and CFO Sarah Friar are clashing over OpenAI’s spending and IPO timing, with Altman reportedly targeting a public offering as soon as Q4 while committing to $600 billion in spend over five years. MG’s first reaction is basically: of course a CEO and CFO fight — that’s the job — but at OpenAI, even a normal disagreement gets dialed to 11 because the spending and ambition are so extreme.
Why Sarah Friar’s skepticism may be the most rational voice in the room
What catches both hosts’ attention isn’t just the feud, but the line that OpenAI’s revenue growth has been slowing. MG says that may simply be the law of large numbers, but if you’re trying to IPO, you need a locked narrative on future growth, and that’s hard when the business is still pivoting and the burn is enormous. Alex adds that any CFO in Friar’s seat would be having a heart attack trying to reconcile exponential-tech dreams with a very non-exponential bankruptcy risk.
The awkward part: financial meetings without the CFO
Things get dicier when Alex reads reporting that Altman excluded Friar from some key server-spending conversations and that she no longer reports directly to him, instead reporting to Fidji Simo. MG says maybe Altman wants “outside the box” brainstorming that a CFO would naturally shoot down, but he doesn’t sugarcoat it: if the CFO is missing from major financing meetings, the optics are terrible. He also notes the organizational chart increasingly made Simo look like a potential day-to-day CEO before her medical leave complicated that story.
Anthropic is suddenly the company forcing OpenAI to focus
From there, the conversation shifts to the competitive pressure underneath all this. MG says OpenAI spent years on “a million things” while Anthropic stayed disciplined around its core market, and now Anthropic’s rise — especially through Claude Code and broader enterprise traction — appears to be forcing OpenAI to kill side quests like Sora and race toward a more unified product. The subtext is sharp: OpenAI may still be bigger, but Anthropic is the one dictating the tempo.
The AI “super app” race is here, and both companies picked the same battlefield
Alex asks whether OpenAI and Anthropic are both right to chase the same destination: an AI app that combines chat, coding, browser/computer use, and agentic task execution for regular people, not just developers. MG says yes, probably — at least directionally — though he warns the risk is building a bloated “Microsoft Office” monster full of dropdowns and feature sprawl. Alex’s own vision is memorable: instead of just answering your question, the AI starts the Adobe Premiere edit for you because it understands what you’re actually trying to do.
Siri may finally become an app because Apple has run out of ways to deny reality
In the Apple segment, Alex tees up the absurdity that Siri is still bad nearly four years after ChatGPT launched. MG is actually more optimistic here: if Mark Gurman’s reporting is right and Apple is testing a dedicated Siri app with saved threads, search, pinned chats, Dynamic Island access, and Action Button integration, then Apple is finally accepting that users expect AI to live in a visible, app-like interface. He calls this a good sign precisely because Apple had spent so long acting like chatbots were beneath its grander vision.
Apple’s deeper problem: outsourcing intelligence may leave it permanently behind
But MG says the larger danger is not just product lag — it’s failing to build the AI mindset internally. If Siri is powered by Gemini and Apple mostly orchestrates others’ models, it may save on R&D while losing the culture and technical reflexes needed for the next phase, especially as OpenAI and Anthropic move from chatbot to agent. That leads into a thought experiment: maybe Apple should buy Anthropic, not just for the models, but to change the company’s DNA the way buying NeXT once reset Apple.
Meta’s nightmare is becoming easier to describe
The final section is blunt: Meta still hasn’t found its next hit beyond advertising. MG says the company has been world-class at acquisitions and copying, but much shakier when it has to invent the future itself — whether that was the metaverse, crypto, encrypted social, or now AI. Alex closes with a brutal DeepMind anecdote from an upcoming Demis Hassabis book: when Zuckerberg pitched him, he sounded equally excited about AI, VR, AR, and 3D printing, and Hassabis took that as proof Google understood AI would be bigger than the rest.
Share
Keep Reading
Two Ways to Follow
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.
