Codex limits draining 5x faster with 18% less usage, dev's tracking claims
StewartalsopIII · x · 2026-09-21
A developer's token tracking (via OpenClaw) suggests his Codex account, running exclusively GPT-5.6 Sol, exhausted weekly limits 4.8–5.9x faster than two months ago while consumption was 18% slower — same work, same reasoning effort. He warns a "token rug pull" is coming; an investor notes cheaper competing models could pressure the platform if true. Unverified personal measurement.
More from coding & agent
- One token is all you need: hacking LLMs into classifiers with logit_bias and max_tokens=1 — AAAzzam · 2026-09-21
- Expanso Builds a Jev-Based Agent That Reads Pod Logs and Auto-Fixes Stale Kubernetes Labels — aronchick · 2026-09-21
- Turning GPT-6 Astra Into a Motion Graphics Studio: 5 Ad Renders in 18 Minutes — PrajwalTomar_ · 2026-09-21
- Sentry CEO: if your product isn't headless, you're investing in the wrong things — zeeg · 2026-09-21
- Clairvoyance marries your AI models to your data via rich metadata layer — draginol · 2026-09-21
- Dev wires Google Astra vs Fable to play RollerCoaster Tycoon over MCP to test computer use — scaling01 · 2026-09-21