Who to Follow
Written 2026-10-04. Every link below was fetched and resolved on that date.
Claude Code changes weekly, and most of what you need to know first appears in a handful of primary sources. This page lists them, then the people whose writing helps you judge what to do with the changes. It is a reading list, not a ranking. Some posts (The Pragmatic Engineer, Latent Space) are partly or fully subscriber-only. Inclusion means the person or source published something on coding agents that was worth a Claude Code user's time recently. I left out anyone whose recent writing I could not tie to coding agents.
1. Official sources
- Claude Code: What's new is a weekly digest of notable features. Start here to learn what shipped.
- Claude Code changelog is the version-by-version record, including fixes. Use it when a behavior changed and you need to know when. The same history is in the GitHub repository.
- Anthropic newsroom carries model launches and company announcements.
- Anthropic engineering blog is the best primary source on how the harness is built. Useful posts include Effective harnesses for long-running agents, Effective context engineering for AI agents and How we contain Claude across products (May 2026). Its posts are infrequent. The newest dated post on the page when I checked was April 2026, apart from a featured May post.
- Claude blog covers product announcements and how-tos, for example Customize Claude Code with mods in TypeScript.
- Claude Platform release notes list API and model changes. Read them if you build on the API or Agent SDK.
2. Claude Code team and Anthropic people
Anthropic staff mostly speak through the official channels above. These are the individual voices with a verifiable written or recorded trail.
- Boris Cherny, who created Claude Code. He has no blog of his own that I could confirm. Read the Platformer interview (26 May 2026) on how the engineer's job is changing, and John Gruber's note on his YC talk (August 2026), which describes having Claude Code rewrite the Claude app and verify the result itself.
- Cat Wu, head of product for Claude Code. Her post Product management on the AI exponential (March 2026) explains how to plan roadmaps when model capability keeps moving.
- Erik Schluntz and Barry Zhang co-wrote Building effective agents (December 2024). Anthropic has added a note that much of its tooling description has changed since, so read it for the patterns, not the tool list.
3. Independent practitioners and writers
- Simon Willison, simonwillison.net. The most consistent daily record of what models and agents can do. Recent: 2026 in LLMs (so far) and We're going to need default hard budget caps on pretty much everything, which bears directly on agent cost control.
- Addy Osmani, Elevate newsletter. Writes about keeping human judgment in an agent workflow. Recent: The Code Nobody Reads and Mastery Still Comes From Doing the Reps.
- Gergely Orosz, The Pragmatic Engineer. Reports on how real engineering orgs work with agents. Recent: Inside OpenAI's agentic software factory and What is happening with code reviews?.
- Kieran Klaassen, Source Code at Every. Practical "compound engineering" workflow for coding agents, including To Read—Or Not to Read the Code?.
- Geoffrey Huntley, ghuntley.com. Opinionated essays on agent-driven development, such as software doesn't need to be readable anymore. it needs to be explainable.
- Steve Yegge, yegge.ai. Argues for governing many autonomous agents with rules rather than containment alone in Fences, not Sandboxes.
- Armin Ronacher, lucumr.pocoo.org. Long-form reflections on working with agents. Browse the index for current posts.
- Ethan Mollick, One Useful Thing. Useful for the non-programmer's view of what agents can do at work.
4. Tool and harness builders
People who build the tools around or beside Claude Code often explain the design trade-offs best.
- Jesse Vincent, blog.fsck.com. Maintains Superpowers, a skills framework and methodology for coding agents. See Superpowers 6.4.
- Dex Horthy and HumanLayer, humanlayer.dev/blog. Context engineering and agent skills, e.g. show-me.
- Thorsten Ball, thorstenball.com. Co-founder of Amp and author of What I believe about the future of software development. Amp's own chronicle shows a competing harness's decisions.
- Mario Zechner, mariozechner.at. Author of the Pi coding agent, a deliberately minimal harness. Pi 1.0 is covered in Latent Space's AINews.
- LangChain, blog. Harness and agent-framework engineering, written by Harrison Chase and colleagues.
5. Researchers and educators
- Lilian Weng, Lil'Log. Harness Engineering for Self-Improvement (July 2026) is a clear map of what a harness is.
- Hamel Husain, hamel.dev. The practical voice on evals. Recent: Claude's new auto eval tool.
- Eugene Yan, eugeneyan.com. How to Work and Compound with AI treats verification as the way to earn autonomy.
- Andrew Ng, The Batch. Weekly letter that regularly covers coding agents; see issue 372 for evals and feedback in new versus mature projects.
- Swyx, Latent Space. Podcast and daily AI news digest for AI engineers.
Keep up without drowning
You do not need all of this. Pick three to five:
- What's new for Claude Code changes, read weekly. Skim the changelog only when something breaks.
- The Anthropic engineering blog, for the few posts a year that explain why the harness works as it does.
- One daily or weekly generalist, such as Simon Willison or Latent Space, to see what else changed.
- One voice that disagrees with your habits, for example Addy Osmani on keeping judgment, or Steve Yegge on running many agents.
- One source close to your own stack, such as Hamel Husain if you write evals.
Read the primary source before the thread about it. A release note, a changelog entry or the author's own post tells you what changed. Commentary tells you what someone thinks about it, and arrives later and louder. When a post claims a new feature or flag, check it against the docs or claude --help before you change your setup.
Sources
All fetched 2026-10-04.
- Claude Code Docs, "What's new" and "Claude Code changelog", Anthropic. https://code.claude.com/docs/en/whats-new ; https://code.claude.com/docs/en/changelog
- Anthropic, Newsroom and Engineering index. https://www.anthropic.com/news ; https://www.anthropic.com/engineering
- Anthropic, "How we contain Claude across products", 2026-05-25. https://www.anthropic.com/engineering/how-we-contain-claude
- Anthropic, "Building effective agents", 2024-12-19. https://www.anthropic.com/engineering/building-effective-agents
- Anthropic, "Product management on the AI exponential", 2026-03-19. https://claude.com/blog/product-management-on-the-ai-exponential
- Platformer, Boris Cherny interview, 2026-05-26. https://www.platformer.news/boris-cherny-interview-ai-jobs/
- Daring Fireball, "Boris Cherny on Trying to Get Claude Code to Rewrite the Claude App", 2026-08-02. https://daringfireball.net/linked/2026/08/02/cherny-claude-swift
- Simon Willison, 2026-09-27 and 2026-10-03 posts (URLs above).
- Lilian Weng, "Harness Engineering for Self-Improvement", 2026-07-04 (URL above).
- Remaining links are the authors' own sites, each loaded on 2026-10-04.