Ollama’s official docs show how to install the app, pull a local model, and call an OpenAI-compatible API at localhost. I didn’t run this setup. Here’s what those docs say, including the bind-address and API-key caveats.
The Eclipse Foundation today announced Eclipse Enclave 1.0, a vendor-neutral sandbox that runs AI coding agents in isolated, policy-controlled containers. I haven’t installed it. This is what the announcement and the current official docs say.
Stack Overflow’s 2026 Developer Survey, fielded June 23 through August 5 among 30,903 valid responses, says 73% of people who answered the frequency question use coding assistants daily. Most still trust the output only when they can verify it.
On November 17, Replit launched Missions. You define the outcome, the company says, and it coordinates a swarm of agents to plan, build, test, and iterate toward it within the budget and boundaries you set.
AGENTS.md is a plain Markdown file at the root of a repo that gives coding agents project-specific instructions. The official docs treat it like a README for agents: build and test commands, conventions, and security notes.
On November 17, at Unite Tokyo, Unity took the Unity CLI and the Pipeline package out of beta as version 1.0. The company says the CLI is how an agent works with Unity: it opens the project, runs commands in the Editor, and reads the result.
On November 16, JetBrains introduced an experimental AI Agents plugin for TeamCity that runs Codex and Claude Code in CI builds and connects them to the LLM provider of your choice. Functionality may change, the company says.
On November 13, OpenAI released Codex CLI 0.169.0. The opt-in terminal workspace now supports multiple side-chat tabs, and codex exec keeps project configuration and MCP servers off in untrusted directories, even with –sandbox workspace-write.
On November 12, Google introduced Gemini 4 Hydrogen and Orchestrator in Antigravity: a larger model plans, Hydrogen does the work, and control returns when something unexpected happens. Those are Google’s claims. Here’s what the docs say.
On November 12, Anthropic opened a private beta so you can host Claude Managed Agents yourself — the harness, sandboxes, credentials, and session history. Docs say each model call still sends the conversation to a provider. Here’s what I’d check.