Nathan Lambert on the open-vs-frontier gap: China labs close behind, Trillium opens RL research
a16z Podcast · rss · 2026-10-14
The a16z Podcast hosted AI researcher and Trillium founder Nathan Lambert to discuss how durable the gap between frontier AI labs and open models really is.
Key points:
- OpenAI and Anthropic hold structural advantages in compute, capital, and the ability to pioneer new training paradigms
- Open-model labs, particularly in China, have proven remarkably capable of following close behind
- The conversation unpacks how much comes from distillation versus technical talent, and why compute remains a major constraint
- Data, RL environments, and post-training increasingly shape model progress
Trillium plans: Lambert introduced his new nonprofit research lab, which aims to open up parts of AI research that are hard to study from the outside—releasing RL environments and documenting how AI is actually being used to automate AI research, to demystify recursive self-improvement for researchers and policymakers.
More from AGI Musings
- AGI timeline moves up to 2027 as industry figures say it should be even sooner — charles_irl · 2026-10-14
- McKinsey now makes some candidates debate an AI avatar in 10-minute interviews — paulnovosad · 2026-10-14
- Designer rebuts 'AI rots your brain' fear: AI lets him run 2-3 side projects alongside his day job — AIandDesign · 2026-10-14
- Hassabis says OpenAI's math breakthroughs should have involved the math community — TorturedPoet30 · 2026-10-14
- Accurate Predictions Don't Guarantee Better Decisions: The Question of What AI Optimizes — AryHHAry · 2026-10-14
- Proposed Context Inspector UI Shows What Your AI Remembers vs. Silently Dropped — DanThinker · 2026-10-14