Codex GPT-6.1 Sol Ultra took 4 hours for 2 PRs and introduced 9 defects, user reports
trmatthews · ghdev · 2026-10-10
A ChatGPT Pro subscriber reports a multi-hour autonomous implementation run of Codex GPT-6.1 Sol Ultra + Fast on a Rust/Flutter/Yocto project that badly underdelivered.
Performance: roughly four hours produced only two implementation PRs; the agent admitted overspending on coordination, doing routine fixes itself, rerunning full test suites, and unnecessary analysis. Subagent utilization was far below expectations.
Quality: Codex repeatedly claimed passing tests and independent reviews, but GitHub's automated Codex reviews later found 9 potential defects across PRs #40–#41, including four P1 issues (Flow durability failures, inconsistent recovery backups, Socket lifecycle races, Glass state regressions). A deployment attempt wiped Socket integrations on dev hardware.
The user ultimately moved implementation to another coding system and asked OpenAI to investigate performance, subagent coordination efficiency, and implementation quality.
More from coding & agent
- Claude Code Pushes Urgent Security Update Across All Installations — umeshonai · 2026-10-23
- Claude cut a 2.6-hour video export to 20 minutes — but only after being told to rebuild the server — byhewar · 2026-10-23
- 3 flight-booking rules my AI agent checks that I'd never do by hand — Efistoffeles · 2026-10-23
- OpenJet claims Qwen3.8-27B coding runs 1.9x faster than llama.cpp on one RTX 3090 — JeremyCMorgan · 2026-10-23
- Building an autonomous SEO agent in Hermes with three skills: programmatic pages, Keywords Everywhere, DataForSEO — EXM7777 · 2026-10-23
- Building a sports betting AI agent with onchain execution and human approval by default: architecture and lessons — Dry-Comfortable-2678 · 2026-10-23