Podwise Weekly Picks, Week 42: Superintelligence, Superlearners, and Who Pays for It
Ten episodes from the week of October 12: Bostrom and Soares on risk, David Silver on learning without humans, agents that spend your money, and the crack in the AI trade.
Weekly picks6 min read
Two conversations about existential risk, one about a model that learns science from scratch, and a markets episode about why the trade cracked. The week's shelf oscillates between the very long view and the very quarterly one, which makes it a decent study in how the same industry talks about itself depending on the audience.
These are the ten episodes from the Podwise Weekly Picks shelf for the week of October 12. Show names link to the episode in Podwise.
The ten
Nick Bostrom on navigating superintelligence
The a16z Show · 11 Oct · 54 min
Bostrom revisits the arguments he helped make mainstream, now that capable systems exist: reinforcement learning pushing models toward goal-seeking, what agent experiments suggest about instrumental convergence, and what he thinks should be preserved. Less "will it happen" than "what do we do with the fact that it is."
Every company should operate like a lab team
Lenny's Podcast · 11 Oct · 62 min
Josh Woodward runs Google Labs, the Gemini app, and AI Studio, and has spent 17 years inside Google's product culture. On how to generate ideas, kill them, and scale the survivors — with NotebookLM and Flow as the worked examples. If you lead a product team, start here.
Nate Soares on how AI really could kill us all
Odd Lots · 12 Oct · 76 min
The co-author of If Anyone Builds It, Everyone Dies makes the case in a finance podcast, which is a strange and useful venue for it. Worth listening even if you disagree; the disagreement is usually about timelines rather than mechanisms.
A superlearner for the age of experience
Training Data · 13 Oct · 68 min
David Silver built AlphaGo on human data, then removed the human data and got a stronger player in two days. He and Lasse Espeholt explain Ineffable Intelligence's plan to build an agent that starts knowing nothing and discovers science and mathematics for itself. The most technically ambitious episode this week.
GLM 5.3 on AMD, Nvidia, and Chinese chips
SemiAnalysis Weekly · 14 Oct · 44 min
A sparse-attention open-weight model that only attends to the tokens that matter, post-training that closed most of the gap, and the question of whether it was quietly tuned for Chinese silicon. Also the practical bit: when teams fall back to GLM if Claude or GPT is unavailable.
Lab-in-the-loop at compute speed
Latent Space · 14 Oct · 110 min
Leash Bio's Ian Quigley on running chemistry experiments with a model in the loop, and what it takes to make a lab that iterates at the speed of compute. Long, specific, and a good antidote to the more abstract "AI for science" episodes.
Greg Brockman on cyber and drug discovery
No Priors · 15 Oct · 39 min
OpenAI's president on frontier models, the changing physics of cybersecurity and drug development, and why the company thinks science is the place where the returns are largest. Short and promotional in places; useful for the roadmap signals.
When AI agents start spending your money
The a16z Show · 15 Oct · 51 min
Alex Rampell and VGS's Terry Angelos on what happens to payments when an agent moves from recommending a purchase to making it: credential infrastructure, wallet competition, and the liability question nobody has settled. The most concrete "agents in the real economy" episode of the week.
CoreWeave crash, AI trade cracks, Anthropic IPO red flags
All-In · 17 Oct · 94 min
The debt selloff, AI's deflationary pressure on software, a mixed investor day, and a segment on self-replicating models. The best parts are the arguments between the hosts; the worst is the running time.
Horizontal vs vertical agents, with LangChain's Harrison Chase
The Twenty Minute VC · 17 Oct · 62 min
LangChain's CEO on whether model routing becomes a commodity, where the inference market is heading, and why he thinks founders should take discounts from tier-one firms. A founder-facing episode that pairs well with the Bostrom and Soares conversations above.
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.
