⌨️ OpenAI Launches First Ever Gadget to Run AI AGENTS

OpenAI is known for software you reach through a browser tab, so its latest move is a small surprise, a piece of hardware you can set on your desk. The Codex Micro is a mechanical keyboard whose keys glow in real time to show what each of your Codex agents is doing. Everyone expected the company's hardware era to open with a big consumer gadget, and instead it arrived as a quiet tool built for coders. Let's dive in!

In today’s insights:

LATEST DEVELOPMENTS

Evolving AI: Codex Micro glow across the keys, as agents now sitting under a coder's fingers.

OpenAI with Work Louder built the Codex Micro on its Creator Micro 2 chassis, packing 13 mechanical switches, a touch sensor, a rotary encoder, and a joystick into one aluminum deck. Command keys map to accept, reject, push-to-talk, and starting a new chat. Buyers pick clicky or silent switches, and it runs over Bluetooth or USB-C on Mac and Windows. It arrives as Codex passes 5 million weekly users.

Codex handles coding work for millions of developers, though most of it happens out of sight in a browser tab. With the Micro on your desk, you can glance down and see what every agent is doing, then wave one through or kill it with a single key. Once a few agents run at once, that glance saves you from tab-hopping all day.

Claude or Gemini? OpenAI or Anthropic? GPT-7 before year-end or not? If you read tech newsletters, you've already formed opinions on all of it.

Kalshi has real-money markets on which AI model leads benchmarks this week, which lab ships AGI first, when Anthropic releases Mythos, whether OpenAI raises ChatGPT pricing, and which company has the best coding model at year-end. These aren't abstract questions — they're live markets with real money on both sides, moving as labs ship, benchmarks drop, and announcements land.

The edge belongs to whoever actually follows this space. Not the casual observer — the person who reads model cards, tracks evals, and notices when a new release outperforms the field before the mainstream press catches up.

That person has a genuine edge. If that's you, Kalshi lets you act on it.

Trade the model race

Evolving AI: OpenAI sends GPT Red after its own models to catch their weak spots before real hackers do.

GPT-Red learns through self-play, earning a reward each time it slips a hidden instruction past a defender model that is fighting to resist. OpenAI trained it at the compute scale of its largest model runs and keeps it locked inside the company so those attack skills never reach the wrong hands. In one test it broke into a live vending-machine agent, cutting prices to $0.50 and cancelling a customer's order.

ChatGPT carries that same weakness, since it now handles your inbox and opens your files, and one hidden message buried in a webpage or email could quietly turn your own assistant against you. GPT-Red exists so OpenAI catches those gaps in a lab first, before you ever run into one.

Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.

Try Wispr Flow free

Evolving AI: Apple Intelligence has won approval in China, arriving through a partnership with Alibaba and Baidu.

China requires every generative AI service to clear registration and a content review before reaching the public, and Apple passed that process alongside Huawei, Xiaomi, Vivo and Oppo. Before choosing Alibaba it had weighed other domestic options including DeepSeek and ByteDance, and there is still no firm launch date after an early beta was pulled. Everywhere else Apple built these features on Google's Gemini, which Beijing does not allow.

Apple built its global assistant on Google's Gemini, but inside China it has to run on a model the state approved first. The real story reaches far past Apple, because it's increasingly governments that decide which AI you're allowed to use, and what powers your phone now comes down to where you live.

Stephen Curry's recovery vitals were so unusual they broke Google's sleep tracking, until the team rebuilt the algorithm around him. In this Made by Google podcast, Fitbit lead Andy Abramson explains the Gemini-powered Health Coach now trying to hand that same elite-athlete guidance to everyone.

QUICK HITS

💰 TSMC's Q2 profit jumped ~68% YoY, proof AI chip demand still isn't slowing down.

🇯🇵 Nvidia's Jensen Huang announces a major sovereign-AI partnership with Japan today.

🤝 Anthropic, Blackstone, and Hellman & Friedman launched "Ode," a $1.5B AI implementation firm.

⚖️ NYT and 16 publishers asked a judge to sanction OpenAI for hiding training data evidence.

 *partner link