Models still can't orchestrate themselves, and that's what makes this researcher bearish
iruletheworldmo · x · 2026-11-11
The author argues models don't know how to use themselves well enough: an expensive orchestration model should be picking which models to call, probing user intent, and setting thinking levels — e.g. answering "I grabbed this with Haiku since it's beneath me, but Sonnet will double-check it." Yet every attempt at this kind of orchestration has been catastrophic, one of the few things making him bearish on progress.
More from coding & agent
- Fable 5.2 lands and users already built 5 playable 3D games with it — FinanceYF5 · 2026-11-11
- Dev vows never to run coding agents locally, sparking workflow debate — verrsane · 2026-11-11
- Using Codex with Capwise CLI to Auto-Generate a Product Intro Video — vista8 · 2026-11-11
- Amazon blocks Scale AI's Muse browser agent, citing Conditions of Use — mohamedmansour · 2026-11-11
- Asking how much code agents wrote is like asking how much assembly the compiler generated — hargup13 · 2026-11-11
- PostHog takes Self-driving GA, admits pricing mistake: reports were where the value was — ycombinator · 2026-11-11