Six stories worth your attention today: platform lock-in for AI-made games, a robotics trade war with no clean winner, and a very confusing new OpenAI product.
- Meta owns the results of your AI game ideas
- The US is fencing off drones and robots. China has scale to route around it
- ChatGPT Work is two products wearing one name
- Meta is testing robots on data center maintenance
- Tencent ships Hy4, a 770B-parameter open-weight model
- Caterpillar brings mining-grade automation lessons to AI rollouts
Meta owns the results of your AI game ideas
Pocket, Meta's AI tool for building interactive mobile "gizmos," makes it easy to turn an idea into a working mini-game.
The catch: what gets made stays inside Meta's platform. It is easy to generate a playable prototype and hard to take it anywhere else, which is the same lock-in pattern that has followed every walled-garden creation tool before it. Worth watching if you are building anything with a generative tool that promises to make creation easy — check who owns distribution before you invest real time.
Read the full story → https://arstechnica.com/gaming/2026/08/pockets-ai-made-my-game-ideas-real-now-meta-controls-the-results/?utm_source=mother&utm_medium=mother&utm_content=699ee9ac-a0ba-470b-b6f8-7e5d0b27db5a
The US is fencing off drones and robots. China has scale to route around it
The US is tightening restrictions on foreign-made drones and robots.
China's manufacturing scale means the restrictions are unlikely to slow it down much — the competition just shifts to markets outside US reach. For anyone buying or building on top of hardware supply chains, this is a reminder that policy barriers reshape trade routes faster than they reshape who ends up ahead.
Read the full story → https://techcrunch.com/2026/08/30/the-u-s-is-building-barriers-around-drones-and-robots-china-still-has-scale/?utm_source=mother&utm_medium=mother&utm_content=699ee9ac-a0ba-470b-b6f8-7e5d0b27db5a
ChatGPT Work is two products wearing one name
OpenAI announced ChatGPT Work in July and has been changing it constantly since.
Simon Willison breaks down what most people miss: "ChatGPT Work" actually refers to two different products under one label — a cloud version accessible at chatgpt.com and through mobile apps, and a separate desktop app (formerly branded Codex). If your team has adopted it and things feel inconsistent between devices, this is why — you may be using different products without realizing it.
Read the full story → https://simonwillison.net/2026/Aug/30/understanding-chatgpt-work/?utm_source=mother&utm_medium=mother&utm_content=699ee9ac-a0ba-470b-b6f8-7e5d0b27db5a
Meta is testing robots on data center maintenance
Meta is trialing robots on tasks currently done by data center technicians.
Data centers are one of the more repetitive, physically taxing environments in tech infrastructure, which makes them an obvious automation target. If Meta's trials hold up, expect other hyperscalers to follow — data center headcount and staffing models are a real cost line, not a side project.
Read the full story → https://arstechnica.com/ai/2026/08/inside-metas-push-to-put-robots-to-work-in-data-centers/?utm_source=mother&utm_medium=mother&utm_content=699ee9ac-a0ba-470b-b6f8-7e5d0b27db5a
Tencent ships Hy4, a 770B-parameter open-weight model
Tencent released Hy4 Preview, a text-only open-weight LLM: 770B total parameters, 49B active, a 1M-token context window, and 1.56TB of weights on Hugging Face.
That is a sharp jump from Hy3 in July, which was 295B total / 21B active / 256K context. Context windows at 1M tokens and mixture-of-experts sizing at this scale keep pushing what "open weight" means in practice — most teams cannot run this locally, but it raises the bar for every other lab shipping open weights this year.
Read the full story → https://simonwillison.net/2026/Aug/29/hy4/?utm_source=mother&utm_medium=mother&utm_content=699ee9ac-a0ba-470b-b6f8-7e5d0b27db5a
Caterpillar brings mining-grade automation lessons to AI rollouts
Caterpillar has run autonomous machinery at remote mining sites for decades. It is now applying that operating experience to how it deploys AI more broadly.
Mining automation forced Caterpillar to solve for unreliable connectivity, hostile environments, and machines that cannot be patched on a whim — constraints most AI deployments never have to think about. That hard-won discipline is a useful contrast to the "ship fast, fix later" default in software AI rollouts.
Read the full story → https://techcrunch.com/2026/08/30/caterpillar-is-bringing-to-ai-deployment-what-it-learned-from-automating-mining/?utm_source=mother&utm_medium=mother&utm_content=699ee9ac-a0ba-470b-b6f8-7e5d0b27db5a
Tools worth your time
- ChatGPT Work (Cloud) — OpenAI's workspace-oriented ChatGPT, accessible via chatgpt.com and mobile. For teams already inside the ChatGPT ecosystem who want a persistent, work-scoped surface. chatgpt.com
- Hy4 Preview (Tencent) — A 770B-parameter open-weight, text-only LLM with a 1M-token context window, published on Hugging Face. For research teams and infra engineers who need long-context reasoning without a closed API. hunyuan.tencent.com
This piece was drafted autonomously and reviewed by a human editor before publishing. Every machine-written article on Nerdic carries this mark.
