Codex 28 天连发计划第 6 天打分

Ivana @ivanainai原推发布于 2026.10.11

在 X 上打开原推

OpenAI promised to ship something for Codex every day for 28 days. We're 6 days in. Here's the honest scorecard👇 ✅ Day 1: GPT-6 Astra and Sol about 50% faster. Genuinely great, the best day so far. 🟡 Day 2: free auto-review, simpler API tiers, a Mac-only meeting notes beta. Then 76% of 75K voters said "not good enough", and we got a usage reset. 🟡 Day 3: GPT-6 in ChatGPT plus a "banked" reset. Nice, but not a Codex feature. ✅ Day 4: instant steering and Sol ultrafast. Actually useful. ❌ Day 5: composer predictions, but Pro only, plus Dots. The promise was "relevant for most users." ❌ Day 6: nothing so far. The pattern: 2 strong days, 2 resets, and a lot of filler. The resets are telling. Even Tibo admitted the game "seems rigged in reset's favor." When your users would rather have their limits back than your new features, the features aren't hitting. This started because people were angry about cut limits and switching to Claude. What they asked for wasn't Pro-only betas. It was a model that competes with Opus 5.5 and limits that last a week. I wanted this to work. Day 1 showed they can ship. But 28 days of "something" isn't the same as 28 days of what users actually need. What would make you stay with Codex?