Google Engineer Breaks Down Agents: LLMs in a Loop and the Reality Gap
bigdata · x · 2026-08-20
Google DeepMind's Maarten Grootendorst joins a podcast to demystify AI fundamentals. He explains the importance of tokens, embeddings, and attention, arguing that classic techniques like BERTopic remain valuable. He characterizes agents as "LLMs in a for-loop," discussing the hype versus reality, where coding agents fail, and the limitations of relying solely on prompting. The episode also touches on reinforcement learning, evolving scaling laws, and personal agent security.
More from coding & agent
- Dev shares multi-agent setup: Infra team, salon groups, and voice manager — fofrAI · 2026-08-20
- NVIDIA Tutorial: Post-Train Cosmos 3 Edge for On-Device Robot Control on Jetson Thor — MonaJalal_ · 2026-08-20
- PR count or LOC aren't great metrics: Agents can farm bug fix numbers — Daniel_Farinax · 2026-08-20
- New agent prompting method: Feed intuition instead of independent scaffolding — Sauers_ · 2026-08-20
- Sharing layers across workers via a "prelude" Effect in alchemy — samgoodwin89 · 2026-08-20
- Hands-on with Instinct, Grok Bots and ChatGPT Work: personal agents compared — illscience · 2026-08-20