PERSON DIRECTORY
Sarah Guo
Host of No Priors. Sarah Guo appears in 71 indexed conversations across No Priors, Invest Like the Best, Latent Space. This directory brings every appearance, source, TL;DR, digest, and transcript into one searchable feed.
Really Big Test-Time Compute in AI Changes Benchmarks, Safety and Research with OpenAI's Noam Brown
Noam Brown argues that model quality must be measured as a cost, token, or time curve, because fixed benchmark scores hide gains from test-time compute.Models can keep improving beyond 100 million tokens, while safety policies still lack a clear budget for evaluating cyber, bio, and other dangerous capabilities.Routing and orchestration businesses therefore face a demanding test: outperforming a single stronger model allowed to think longer at the same cost, with gains that transfer beyond benchmarks.
No Priors Ep. 105 | With Director of the Center of AI Safety Dan Hendrycks
AI safety is framed as a geopolitical and economic problem that alignment alone cannot solve, with biological capabilities becoming more consequential and reliable agency more important than closed-ended benchmark scores.Hendrycks proposes mutually assured AI malfunction to deter destabilizing superweapon projects while preserving ordinary competition, but efficient training, model-weight theft, talent interdependence, and uncertain agent reliability complicate compute controls and timelines.
Reshaping America’s Economy for the Superintelligence Century with Jacob Helberg
Jacob HelbergSarah GuoElad Gil
Jacob Helberg argues that AI, deregulation, and abundant energy could shift the U.S. from consumption toward industrial investment, with CapEx above 2% of GDP and potentially doubling by next year.Critical minerals, semiconductors, nuclear permitting, and platform distribution remain decisive bottlenecks as America competes with China for supply-chain control and the Global South’s AI stack.
No Priors Ep. 127 | With SemiAnalysis Founder and CEO Dylan Patel
OpenAI’s expected compact, reasoning-heavy open model could push intelligence pricing below the frontier, while inference differentiation shifts from kernels toward orchestration, networking, reliability, and physical infrastructure.Roughly 200 neoclouds face consolidation as GPU economics diverge, while NVIDIA’s hardware, networking, and software moat remains difficult to replicate; bottlenecks in power, packaging, labor, and export policy are the next constraints to monitor.
America’s Plan to Dominate the Full AI Stack with Sriram Krishnan
Sriram KrishnanElad GilSarah Guo
DeepSeek was the administration’s “starting gun,” exposing how narrow America’s lead may be and making full-stack inference share—from American chips to models and applications—the proposed scoreboard for “America Inc.”Execution now depends on power generation, grids, permits, data centers, domestic labor, open-model policy, allied exports, and robotics within 18-24 months, while regulation, cyber and bio risks, Chinese standards, and an acknowledged AI “event horizon” remain unresolved.
National Security Strategy and AI Evals on the Eve of Superintelligence with Dan Hendrycks
AI safety is fundamentally a statecraft problem: labs are “predetermined to race,” while aligned US and Chinese systems could still intensify military integration, labor automation and strategic risk.Hendrycks proposes mutually assured AI malfunction, using espionage, cyber options and compute tracking to deter destabilizing projects while export controls constrain rogue-actor access.Humanity’s Last Exam may signal superhuman closed-ended STEM performance, but agents remain “near the floor”; reliable digital agency is the catalyst that could change the stakes.