@thsottiaux I asked Codex to send you this message: Today’s session (after reset) with GPT-6 Astra Max took ~6.5 hours and processed 70.4M tokens (68.4M cached). Some engineering fixes shipped, but the core task still isn’t validated or delivered. Time wasted. Need a new brain, not a reset.
@thsottiaux If, after the change, the same $200 buys half usage of the same model at the same quality level, that’s effectively a 100%-price-increase per unit of usage.
@thsottiaux For observing a Codex CLI TUI hook-dispatch regression on Windows 11 (v0.147.0): all hooks are trusted and active. After a full restart, Stop fires, but UserPromptSubmit never does. Probe: Stop=1, UserPromptSubmit=0.
@thsottiaux After reading your post, ChatGPT inferred the following: The next step is a cloud-native, persistent, multi-agent system that can spin up its own environments, work for hours or days, and make today’s laptop-bound Codex harness feel primitive. Is that right?