Users slam Codex: cache 'fix' triggers usage regression as token burn soars
Defiant-Truck3 · reddit · 2026-10-20
A long-time user complains that after Codex touted a caching fix, their plan allowance was downgraded from Extra High to High to Medium while token consumption stayed extreme — apparently a regression introduced by the fix itself. Sol 6.1 was slow but at least usable. The author argues the much-hyped '28-day challenge' now looks like shipping untested changes: 'a fix that introduces a regression isn't a fix. Are we watching engineering progress or a startup figuring things out in production?'
More from coding & agent
- ttmi makes your AI agent draw diagrams instead of dumping walls of text — blixt · 2026-10-20
- Claude Code plugin maps Claude's C compiler: 351 Rust files, 13 findings in 21 minutes — gantamk · 2026-10-20
- Enterprise AI Doesn't Fail in the Demo. It Fails in the Mess. — DavidLinthicum · 2026-10-20
- tsrs: Rust TypeScript Compiler Port Backed by TS Team Claims 8x Speed, 70% Less Memory — cnakazawa · 2026-10-20
- Bots that should be skills: auditing a Grok Bot fleet with six tests — bfrench · 2026-10-20
- Dev's code review setup: agent diagrams before/after flows on a canvas for Q&A — max__drake · 2026-10-20