Misha Laskin
@MishaLaskin
The next generation of breakthroughs in agentic models will come from unlocking long context reasoning capabilities during pre and mid training.
Really enjoyed listening to @achowdhery sharing her thoughts on this next frontier of intelligence
Really enjoyed listening to @achowdhery sharing her thoughts on this next frontier of intelligence
Reflection@reflection_ai · Dec 18, 2025Most approaches to “agentic AI” focus on post-training fixes.
In this conversation, member of our technical staff, @achowdhery argues the bottleneck is pre-training itself. Drawing on her work on PaLM and early Gemini, she explains why next-token prediction breaks down for
Open quoted post →In this conversation, member of our technical staff, @achowdhery argues the bottleneck is pre-training itself. Drawing on her work on PaLM and early Gemini, she explains why next-token prediction breaks down for
6 95