User slams OpenAI's GPT-6.1 Sol: 20 hours lost, ignores human feedback
RexDouglass · x · 2026-10-23
A user details a frustrating experience running strategy research on OpenAI's 6.1 Sol model: the agent executes for days without stopping, loops repeatedly, burned 20 hours on an API/RPC problem without progress, and 180-degree pivots when given feedback, forgetting its original goal. Verdict: this head-down, feedback-resistant execution mode makes it poorly suited for long-running tasks that need human steering.
More from coding & agent
- GitHub moved 99.99999% of repos to Azure; per-reference quorums unlock agent-scale Git concurrency — marlene_zw · 2026-10-23
- Protocol-style prompt makes Codex self-review pages to Awwwards-tier standards — huangyun_122 · 2026-10-23
- Token Harness: Open-Source Local Dashboard to Cut Claude Code and Codex Token Spend — giuliastro · 2026-10-23
- Microsoft's FrontierTour: AI governance becomes everyone's problem as agents take action — donasarkar · 2026-10-23
- An agent edits its host app's own source code and rebuilds it on-device, with preview and rollback — osbornep · 2026-10-23
- Grady Booch revives Weinberg's 1971 quip: software so buggy one woodpecker could destroy civilization — Grady_Booch · 2026-10-23