How I Use Claude to Help Run This Blog (and Why You Should Try It)

30 May 2026 AI

If you’ve been following this blog for a while, you might have noticed it’s fairly active for a one-person operation. New posts, new guide sections, new tooling write-ups — and all while I’m working a full-time job in networking. The honest answer to “how do you keep up with it?” is: I don’t do it alone.

The Problem With Blogging as a Technical Person

Writing about technical subjects is genuinely hard. Not because the material is difficult to understand — it’s usually not, once you’ve done the work — but because translating deep operational knowledge into clear prose takes a different kind of effort. After spending a day debugging a FortiGate policy or building a Python tool, the last thing I want to do is switch modes and write 1,500 words about it.

The result is usually one of two things: the post either gets written quickly and feels rushed, or it sits in a mental queue for weeks and never gets published.

How Claude Fits Into My Workflow

I use Claude (Anthropic’s AI assistant) as a writing collaborator. The workflow looks roughly like this:

  1. I do the actual work — build the tool, run the lab, debug the issue.
  2. I describe what I did to Claude: the problem, the approach, the interesting bits, the gotchas.
  3. Claude drafts the post from that description, structuring it, filling in context, and matching the tone I’ve established across the site.
  4. I review and edit — cutting anything that doesn’t sound right, adding specific detail only I’d know, and making sure the technical accuracy is solid.

This isn’t “AI writes my blog.” It’s more like having a technical writer sitting next to me who can turn a brain-dump into a readable first draft in two minutes. I still own every post; I just skip the blank-page problem.

The MCP Server

The part I’m probably most proud of is the infrastructure behind this. I run a self-hosted MCP (Model Context Protocol) server on my home Linux box — the same machine that hosts this site. The MCP server exposes a set of tools Claude can call directly:

  • write_draft — creates a new blog post draft with frontmatter and body
  • deploy_post — flips the draft live and rebuilds the Astro site
  • write_site_file — edits any source file under src/ or public/
  • read_site_file — inspects existing files before making changes
  • rebuild_site — triggers a full rebuild without any content changes

What this means in practice: Claude can write a post, and with my confirmation, deploy it — all within the same conversation. No copy-pasting into files, no manual SSH, no “oh, I forgot to update the series registry.” The whole pipeline runs through the chat interface.

The server is a Node.js/ESM process sitting behind Caddy with TLS, authenticated via OAuth. It’s been running reliably for months and is genuinely one of the more useful things I’ve built.

This blog itself has become the ongoing demo for the AI series — each new capability I wire up tends to turn into a post.

What AI Is Good At (and Not Good At) Here

To be straight with you:

Good at:

  • Drafting narrative structure from bullet-point notes
  • Explaining concepts clearly when I give it the right context
  • Generating consistent formatting and frontmatter
  • Refactoring site code when I describe what needs to change
  • Catching things I’ve forgotten (missing tags, broken internal links, etc.)

Not good at:

  • Technical accuracy without my input — it’ll hallucinate details if I don’t verify
  • Knowing what I actually think about something
  • Replacing the hands-on work (no amount of AI assistance writes a working Python tool for you)

The sweet spot is: I do the engineering, Claude handles the translation into prose, and I verify both.

Try It Yourself

If you’re curious about Claude and want to give it a go, I have a referral link below. When you sign up through it, you get a credit bonus — and I get a small referral benefit too. I want to be upfront about that: this is a referral link and I benefit if you use it. But it’s also genuinely what I use every day, and I wouldn’t point you at it if I didn’t think it was worth your time.

Try Claude with my referral link →

Whether you use the referral link or not, the tool is worth a look if you’re doing any kind of technical writing, scripting, or lab documentation. It’s not magic, but it’s a genuine productivity multiplier when you use it for the right things.

Disclosure: The link above is an affiliate/referral link. If you sign up through it, both you and I receive a benefit (typically account credit). This doesn’t affect my opinions — I’ve been using Claude as a core part of my workflow for months and that’s what prompted this post.