Codex 사용량 정책, Wan 3.0 영상 생성, 에이전트 인프라 확장

What I wanted to say yesterday is that we hit 25M active users and to celebrate we have now reset usage for all paid subscriptions for ChatGPT Work and Codex. See you soon for more news from The Reset Company.

For clarity, while both are called 20X, in Codex they apply specifically to weekly usage limits. And we also don't have 5h limits for both Pro plans. The Pro 20X is quite precisely 20X the usage of the Plus subscription, so it does exactly what it says on the tin.

Peter Steinberger

@steipete

4일 전

Two months ago, we started the mission to “build OpenClaw with OpenClaw,” and bit by bit, we moved everyone from using their local coding harness to using http:// team.openclaw.ai - our shared agent that knows what everyone’s working on and orchestrates it all. Multiplayer coding + infinite compute with nodes and cloud sessions has been a game changer for how we build. Local harnesses feel like relics of the past now.

one thing i wish more folks understood you should be deleting your agents md and skills on the often. especially between model releases take the learnings and revalidate them every couple months. the models are getting better and less malloc() is needed to get outcomes/drive behaviours. it’s still WAY too early for standardisation

glm-5.3 now beats gpt-5.6 sol and claude fable 5 on agentic benchmarks 5.3 flash is right behind it glm-5.3 didn’t even need a new base model to get there @zai_org kept the glm-5.2 base and scaled post-training with more long-horizon environments, more diverse tasks, and more rl compute

Prompting Wan3.0 like a pro — from the Head of Video Product at @Picsart . Narek Hayrapetyan shares the prompting playbook his team actually uses: structuring prompts, layering audio, and getting the most out of Omni-Reference. Plus his honest take on where the model fits into a

Wisnu AI Project shows how Wan 3.0 holds a character together across an entire AI film. Full video ↓

Open weights last week, more open infra this week Introducing CubeSandbox v0.7.0, where sandboxes stop being tied to one machine. Highlights: > Single box: still 60ms cold starts, and batches start faster now > Clusters: pause on one machine, wake it on another. State lives in shared storage, so draining a machine no longer kills what's paused on it. Preview for now > Upgrades: months old templates and snapshots still work, no rebuild, no downtime 239 commits from 57 contributors this release, and the roadmap is public Come argue with us about what lands next. @CubeSandbox repo and changelog ↓

QwenCloud

@qwen_cloud

4일 전

Building with AI shouldn’t mean stitching together models, APIs, and infrastructure. QwenCloud is built for the next AI developer journey: a website for human exploration, Skills for agent reasoning, and CLI for workflow execution. One platform for model access, multimodal APIs, fine-tuning, deployment, monitoring, and predictable cost control. At Qwen Conference, the QwenCloud team walks you through it all live: how to access the Qwen model family, fine-tune for your use case, deploy at scale, and monitor in production. Enterprise-grade scalability, reliability, and cost optimization — built in. If you're an enterprise builder, a developer, or anyone who's ever thought "there has to be a better way to ship AI" — this is your session. Sep 4, Bangkok — Register → https:// qwencloud.com/events/qwen-co nference-thailand-2026?utm_content=g_20000002890 …

Building with AI shouldn’t mean stitching together models, APIs, and infrastructure. QwenCloud is built for the next AI developer journey: a website for human exploration, Skills for agent reasoning, and CLI for workflow execution. One platform for model access, multimodal APIs, fine-tuning, deployment, monitoring, and predictable cost control. At Qwen Conference, the QwenCloud team walks you through it all live: how to access the Qwen model family, fine-tune for your use case, deploy at scale, and monitor in production. Enterprise-grade scalability, reliability, and cost optimization — built in. If you're an enterprise builder, a developer, or anyone who's ever thought "there has to be a better way to ship AI" — this is your session. Sep 4, Bangkok — Register → https:// click.qwencloud.com/m/20000002466/

A 15-second 768p clip with synced audio can now be generated faster than you can watch it. reactorworld just turned that into an infinite Twitch broadcast.

If you care about world models, this is the repo to fork tonight - https:// github.com/reactor-team/i nfinite-livestream … https:// twitch.tv/dereactor