Hello AI Enthusiast,
This week was all about OpenAI's DevDay, with more than 20 announcements. The headline is Dots, agents that keep working on your goals even when you are not around. OpenAI also launched Space, a shared workspace for teams and their agents, and a Decisions API that answers simple questions in milliseconds. Let's get into it.
The Big Picture 🔊
OpenAI launches Dots, always-on AI agents
At DevDay, OpenAI introduced Dots, agents powered by GPT-6 Astra that keep working toward your goals instead of waiting for prompts. Each Dot gets its own cloud computer and browser, connects to more than 4,000 apps, and can be reached in ChatGPT, Slack, Teams or by voice call. Dots are rolling out to Pro and Business Premium in eligible markets, and the first one is included in the plan.
|
||
|
Most AI still waits for you to decide what to do. Dots work more like a good employee: they own a task and check in when needed. That only works if you trust the choices they make on their own, and agents that jump in uninvited get annoying fast. An always-on agent is also very expensive to run, so a free first Dot looks like the hook before the bill. |
ChatGPT Space puts teams and agents in one room
OpenAI also launched ChatGPT Space, a shared workspace where teammates, ChatGPT and their Dots build on the same files and project context. It comes with Pages, which people and agents can edit and comment on together. Collaborative slides and spreadsheets are coming soon. Space is available on Pro, Business and Enterprise plans.
|
||
|
AI has mostly been a one-on-one conversation. Space puts you, your colleagues and AI in the same place, with a mini office suite included, just weeks after Claude launched its own Docs and Slides. We are curious to try it. The strategy is less surprising: once your projects live in Space, leaving gets much harder. |
OpenAI has Dots, Microsoft has Autopilot, and every big platform now wants an agent working inside your company. Knowing which tasks to hand over, and how to check what comes back, is the part the launch videos skip. That is what our corporate training is built for.
OpenAI's Decisions API answers in milliseconds
OpenAI announced the Decisions API, built on a version of its low-cost GPT-6 Luna model. Instead of writing text, it picks one answer from options the developer defines, for tasks like classifying content, routing requests or choosing an agent's next step. OpenAI says it responds in 150 milliseconds, compared to 1.6 seconds with regular Luna. It is in limited preview, and pricing has not been shared yet.
|
||
|
Many automations only need a quick choice. Is this invoice safe? Is this lead a company or spam? Until now we used models built to write essays to reply with one word. A dedicated classifier fits most of the automations we build. The funny part: TypeSafe's Jev, which we mentioned last week, went viral with the same idea two weeks before OpenAI shipped its own version. |
Bits and Bobs 🗞️
OpenAI released GPT-6.1 Sol, which nearly matches GPT-6 Astra on coding and professional work at a fifth of the price. It is available in ChatGPT Work and Codex, not yet in Chat.
Microsoft introduced the new Copilot, with Chat and Cowork together in Home, Code for building small apps, and Autopilot, an always-on agent. It rolls out first through the Frontier program.
Google will replace Gems with skills in Gemini, starting in November for personal accounts and in 2027 for work and school accounts.
Anthropic released Claude Sonnet 5.5, which runs over 30% faster than Sonnet 5 and costs up to 30% less per task.
Claude Tag can now use your personal connectors, like your calendar or drive, when you ask it something in a Slack channel. It is rolling out on Team plans first.
Anthropic shared that Claude discovered a new enzyme system similar to CRISPR, after about 950 agents searched DNA data for 21 hours.
Meta launched Muse for Small Business, new skills and connectors that link Muse to Instagram, Facebook Pages, Meta ads and tools like Canva. Muse is available in the US and Canada.
NVIDIA launched the Open Agent Safety Platform, open source software and a hardware design that monitor AI agents and stop them when they cross their limits.
OpenAI now lets you use your ChatGPT plan in other tools like Notion, Vercel and Devin, with usage counting toward your plan limits.
Codex can now run in the cloud from any device, with reusable environments that teams can share.
That's a wrap on our newsletter!
If you're looking to build AI skills in your team, check out our Corporate Training. Every program is tailored to your team and built around hands-on practice.
Catch you next week! 👋
