Podwise Weekly Picks, Week 41: Assistants, Walkouts, and Who Funds the Buildout
Ten episodes from the week of October 5: the assistant war, an OpenAI safety lead who quit, the cloud built for agents, and the debt behind the compute.
Weekly picks6 min read
The week the assistant wars got personal and the buildout got a balance sheet. There is a resignation letter, an IPO filing, a database professor explaining what agents do to storage, and two episodes about who is actually lending the money for all those GPUs.
These are the ten episodes from the Podwise Weekly Picks shelf for the week of October 5. Show names link to the episode in Podwise.
The ten
Who wins the AI assistant wars?
Big Technology Podcast · 5 Oct · 64 min
M.G. Siegler on why Meta's Muse is stronger than expected, Apple's dark-horse position, Anthropic's absence from the consumer race, and Google's repeated product missteps. A useful snapshot of a market that is still refusing to consolidate.
The Anthropic S-1 numbers nobody expected
Big Technology · 6 Oct
$4.6 billion of revenue last year against a plan to spend $518 billion on compute over the next decade, 80% of it binding. Alex Kantrowitz and Ranjan Roy read the leaked figures and ask whether this is a genius bet or a bankruptcy risk. The numbers in this episode frame the rest of the week.
An OpenAI insider explains why he quit
The Ezra Klein Show · 7 Oct · 71 min
David Robinson wrote the safety reports for new models at OpenAI and left because he no longer believed the safety culture was adequate. The interview is less about one company than about what internal accountability looks like when the product cycle is faster than the review cycle.
Building the cloud for an agentic world
The a16z Show · 8 Oct · 56 min
AWS CEO Matt Garman on reserving scarce GPU capacity for startups, custom chips like Trainium and Graviton, and how the cloud changes when agents write code and manage infrastructure. The closest thing this week to an incumbent's answer to the agent question.
When billions of agents hit your database
The MAD Podcast · 8 Oct · 77 min
Andy Pavlo argues that the AI world's mental model of databases stopped at vector search and RAG, while agents are becoming the biggest users of the data layer — creating, querying, and occasionally deleting. If you build backends, this is the most directly useful episode on the shelf.
Synopsys will take a royalty on Amazon's chips
Semi Doped · 8 Oct
A shift in the semiconductor IP business: royalty-based custom IP, starting with a billion-dollar Amazon deal, plus a partnership with OpenAI and the idea of "agent engineers" in chip design. Niche, and a good signal of where design work is heading.
Synthesis superintelligence, from semiconductors to superconductors
Latent Space · 8 Oct · 84 min
Periodic Labs' Liam Fedus and Ekin Dogus Cubuk on building an autonomous lab one year after launch, and the talent density required to do it. A companion to the Leash Bio episode from week 42; together they are the best picture of "AI scientist" as a real lab rather than a demo.
How AI is upending mathematics
Odd Lots · 9 Oct · 60 min
An AI-generated proof for Navier-Stokes, a field where models used to fail at counting, and the classroom problem that follows. The episode is careful about what a proof claim means; that care is the reason to listen.
Beam, the great American open model
No Priors · 9 Oct · 70 min
ReflectionAI's Misha Laskin on a 500-billion-parameter open-weight reasoning model, the pre-training and RL behind its efficiency, and the shift from renting intelligence to owning it. The case for open weights from someone shipping one.
Who's funding the $11 trillion AI buildout?
SemiAnalysis Weekly · 9 Oct · 83 min
H100 rental prices went up on a GPU that should be depreciating, which breaks the curve lenders underwrite. Daniel Nishball and Jordan Nanos follow the money through Nvidia backstops, credit signals, and whether each token is profitable. The finance counterpart to the Garman interview.
How these are picked
The shelf is curated by Podwise from the week's episodes, and the order follows the collection. All ten are processed, so reading a summary or transcript costs no credits.
Browse the full shelf, including past weeks and themed collections, on Explore.
