OpenAI's math results spark "superintelligence" debate on TBPN

Adobe's software moat cracks as AI decompiles its entire suite

The hosts discuss a site called Artcraft (getartcraft.com), which offers free, open-source clones of Adobe's entire product line, including PhotoCraft (Photoshop), VectorCraft (Illustrator), FilmCraft (Premiere Pro), LightCraft (Lightroom), and others. One host tested PhotoCraft and found it "pretty remarkable how many tools did make it over," though it lacks Adobe's server-side AI features like generative fill, which rely on models such as Nano Banana.

The hosts argue this creates a split: purely virtual, prompt-based tasks are being rapidly commoditized, while workflows requiring heavy, sustained software use or server-side AI retain more of a moat.

"the modes are are are decreasing, but there still is something where I where if I were doing heavy duty Photoshop use, Photocraft would not be a perfect substitute for my workflow" — the host

They note Adobe's stock is down roughly 3-5% over five days, 10% over the past month, and 33% over the past year, though they say there's no sign of a mass cancellation of Creative Cloud subscriptions yet.

OpenAI's math results reignite the superintelligence debate

The hosts describe OpenAI's new math results as covering "a broad range" of problems across what one host says is "722 manuscripts," calling the framing deliberately vague. They cite an analysis from Francois Chollet, creator of ARC-AGI, who frames the jump as possibly being concentrated in "math plus code," which can be pushed arbitrarily far with reinforcement learning with verifiable rewards (RLVR), while other domains remain bottlenecked by human-generated data.

One host argues this feels less like an AGI moment and more like a "superintelligence" moment, since "no human intelligence has ever been able to solve this many math problems this quickly at this level of mathematical ability." He also references Ilya Sutskever's 2023 comments cautioning against valuing intelligence above all other qualities, tying this to the show's running "golden retriever mindset" bit, and notes that despite the math gains, day-to-day startup work still looks like "high stakes poker" rather than problem-solving.

Separately, Ben Thompson's essay "Apple and the Hacker's Future" is discussed: he argues AI now allows almost any software to be decompiled, and he wants agentic coding tools to inherit the same permissions as the AI agent itself rather than face macOS's per-app approval prompts. The hosts speculate this could eventually extend to decompiling macOS itself, though they note Apple would likely resist, and compare it to iPhone jailbreaking, which most users avoid because it sacrifices features.

SpaceX seeks $40 billion to fund its Nvidia chip order

The hosts summarize reporting that SpaceX is seeking to raise about $10 billion in bank loans and $30 billion in investment-grade debt to fund a large Nvidia chip order, with private capital group Apollo expected to lead the financing and PIMCO among lenders in talks. SpaceX's BBB credit rating (the second-lowest investment-grade tier) would let pension and insurance funds participate.

They highlight a detail from the reporting that investors approached about the financing received only "a short two-page deal memo with pictures of outer space and an arrow pointing out that the company was going to build data centers somewhere in the universe." One host says that despite the thin memo, the deal is conceptually simple to underwrite:

"it really is pretty simple. Like it's $40 billion. Buy some chips. What is the value of the chips?" — the host

They note Elon Musk's quoted comment that SpaceX has "decided to build exclusively on Nvidia" using the Vera Rubin architecture, while observing he is simultaneously pursuing his own chips via Tesla and a deal with Intel.

Wall Street's hiring shift and the "beautiful data center" meme

The hosts read from an AQR Capital letter arguing that quant firms are increasingly hiring "idea guys" — liberal arts grads and generalists — rather than pure quants, betting that creativity and emotional intelligence will be the durable differentiators as AI surpasses many analytical and technical capabilities. One host says he knows hedge fund people with philosophy and history backgrounds who brought differentiated perspectives to investing.

The segment also covers a viral image purporting to show a "beautiful" European data center, which the hosts later learn was actually a French train station, posted by Mistral's co-founder and subsequently community-noted; the hosts joke it still shows appetite for aesthetically pleasing data center design. They also mention a field interview with a construction worker who, asked about data center jobs being "temporary," replied

"every job in my field is temporary" — a construction worker

and close with a product called Hollow Glass, a DIY system that turns car windows into digital ad displays starting at "about $300."