OpenAI resets ChatGPT Work and Codex usage limits for all paid tiers
Lands today at 10am PST after GPT-6 speed fixes — plus Ant's free 560B Ling-3.1-flash, an open tool giving agents the whole web, and a token-cutter for agents.
Copy markdown
Your Codex and ChatGPT Work caps reset today
OpenAI is rolling a global usage reset for every paid ChatGPT account, landing October 3 at 10am PST (9pm in Dubai), that zeroes the usage counters for ChatGPT Work and Codex. It follows this week's GPT-6 latency fixes — if you've been rationing agent runs or coding sessions, your limits clear today.
Ant's 560B Ling-3.1-flash is free to try now
inclusionAI opened a two-week free trial of Ling-3.1-flash, a 560B-parameter mixture-of-experts model with about 25B active per token, tuned for agents, search and office tasks. The trial caps context at 256K; inclusionAI says the full 1M-token window and open weights land once the trial ends.
Agent-Reach gives your agent the whole web, free
The MIT-licensed Agent-Reach (75k+ stars) routes an agent's requests to web pages, YouTube, RSS, GitHub, Reddit, X and more through free and open endpoints instead of paid platform APIs. It's a drop-in way to feed live web and social data to an agent without per-call costs.
Caveman cuts coding-agent tokens by a third
Caveman (100k+ stars) pairs a 'speak tersely' skill with a local proxy that compresses payloads before they hit the API — roughly 65% fewer output tokens and 33% fewer input tokens in its benchmarks, though whole-session savings run lower. It plugs into Claude Code, Cursor, Copilot and 30+ other agents.