Adam Dorr: agents looping scripts is easy; solving real problems remains marginal

adam_dorr · x · 2026-09-26

Roboticist Adam Dorr argues that looping a simple script is one thing, but building or solving anything of real significance remains marginal even with top models on max effort. Peter Steinberger counters that many capable models already work well, noting that when OpenCode got popular no model beat Luna — now priced at $0.1 per million tokens.

Related event: Debate: Is Cost or Capability the Real Bottleneck for AI Agents?(3 posts)→

Original post →

More from coding & agent

coding & agent channel →