Builders, p. 2
Builders
56 stories · page 2 of 7n8n released agents that work out the steps themselves, while workflows set how far they can reach
On 25 September n8n added a new type of agent: you describe the task, choose a model, tools and workflows, and the agent decides how to reach the result. Workflows can become its tools and so limit what it can touch. The old AI Agent node stays unchanged.
Read →Docker opened a format for what an agent is allowed to do and is handing it to the CNCF
On 24 September Docker released the Sandbox Kit Spec under Apache 2.0: the agent, its tools and a list of everything it wants to reach, in one ordinary container image. The standard goes under the neutral governance of the CNCF. The same day brought Cloud Sandboxes for agents that work for hours.
Read →Google closed Gemini 2.5 to new users
In a note dated 18 September Google limited access to the Gemini 2.5 models in the API to users who have actively used them. They are not deprecated, but for a new app Google points to 3.5 Flash-Lite or 3.8 Flash, citing capacity.
Read →Open models passed half the traffic at Vercel: 56 per cent of tokens, but 14 per cent of the money
The monthly AI Gateway index for August: open models handle most tokens for the first time, the price per token falls for the third month in a row, and Anthropic takes 64 cents of every dollar.
Read →Grok Build now remembers your project between sessions and writes its own notes
After every completed turn the agent records conventions, decisions and project facts as markdown notes. Secrets, tentative conclusions and anything already documented stay out of memory, and an instruction in the conversation beats a note.
Read →TypeSafe released Jev: a model that returns decisions with probabilities instead of text
TypeSafe AI came out of two years in stealth with Jev, a model that takes state and questions and answers with code-ready choices, scores and yes or no, each with a probability. Input costs 4.2 cents ($0.042) per million tokens, output is free.
Read →Devin got its own Mac and now runs iOS apps instead of only compiling them
Cognition launched macOS support for its agent Devin. It works in virtual Mac machines in the cloud and can launch the app, close it, open it again and check how it behaves.
Read →OpenAI gave every developer the Codex harness as an API and charges nothing on top
On 10 September OpenAI launched the Agents API in public beta. With one call you create a cloud agent with a given task, model, tools and environment, and behind it sits the same harness that powers Codex. There is no extra fee: you pay for tokens and tools.
Read →VS Code 1.137 runs agents on a schedule and lets you interrupt them by voice
The 9 September release adds Automations, agent tasks that start every hour, day or week. The experimental Voice Mode lets you talk to the agent and interrupt it while it works. GitHub issues and pull requests open in the Agents window without the repository being open.
Read →