$ The best sources for AI coding agents and vibe coding on Linux

Where to follow terminal coding agents such as Claude Code, Codex CLI and Aider on Linux: release notes, hands-on writers and discussion, and how to keep up without drowning in hype.

Why follow coding agents at all

Terminal coding agents read your repository, edit files and run commands in your shell. Their behaviour, defaults and permission models change often, sometimes from one week to the next, so an update can change what an agent is allowed to do on your machine.

Following a few good sources means you know about new features, changed defaults and known issues before you rely on them, instead of finding out in the middle of a session.

Release notes come first

The GitHub release pages of Claude Code, OpenAI Codex CLI and Aider are the primary sources for what changed in each version. Read them before upgrading, and pay particular attention to anything about permissions, sandboxing, configuration files or supported models.

People who test things in practice

Simon Willison's weblog documents hands-on experiments with models and agent tools, including what went wrong. Latent Space covers AI engineering through interviews and essays with people building these tools. Both are good for judging whether a new feature is useful in real work.

Vendors and research

OpenAI News and the Microsoft Research Blog announce new agent capabilities and research. Treat vendor posts as announcements, and check release notes and independent write-ups for how the feature behaves in practice.

Discussion and opinion

Hacker News threads about LLMs and coding tools are a quick way to see early reports and criticism, and DHH's blog adds a practitioner's view on AI-assisted programming. For the kernel's perspective, Linus Torvalds' posts about AI and LLMs on the kernel mailing list are worth reading. Read discussion for signals, then verify in the primary sources.

Using agents safely on Linux

Run agents inside a project directory under version control, review diffs before committing, and keep secrets out of the working tree. For untrusted repositories or long unattended runs, use a container or a VM, and check what each agent's permission settings allow before you enable automatic command execution.

How to combine them

A practical setup is the release notes of the one or two agents you actually use, one hands-on writer, and one discussion source. Check the release notes before upgrading, and keep a note of the version that works for you.

If you would rather not check these pages every day, Linux Doctor sends a free daily email with the new items from these sources and publishes the same list on the latest checkup page.

All sources mentioned

Claude Code Releases

Release notes for Claude Code, the terminal coding agent.

OpenAI Codex CLI Releases

Release notes for Codex CLI, an open-source terminal coding agent.

Aider Releases

Release notes for Aider, AI pair programming in your terminal.

Simon Willison's Weblog

Daily notes on LLMs, local models, tools and practical AI hacking.

Latent Space

Newsletter/podcast for AI engineers: agents, coding tools, infra.

Hacker News: LLMs

Hacker News stories about LLMs and local models with 100+ points.

OpenAI News

Product launches, model releases and research from OpenAI.

Microsoft Research Blog

Research updates incl. agents, small models and AI systems.

Linus Torvalds on AI (LKML)

Linus Torvalds' own Linux kernel mailing-list posts (lore.kernel.org) that mention LLMs, AI tools, AI slop or AI code review such as Sashiko. Keyword search, so the odd post only touches AI in passing.

Hacker News Best

The best-voted stories on Hacker News (all topics, tech-heavy).

DHH (David Heinemeier Hansson)

DHH's personal blog on world.hey.com (not his X account): the creator of Omarchy and Ruby on Rails on Linux, open source, programming and building software with AI.

Related

AI agents newsVibe coding & dev tooling newslatest checkupall sources