How I Run 5 Blogs Solo Using AI and a Second Brain

I run five blogs by myself. Different niches, one person, no team.

For a long time that felt impossible. Writing one blog consistently is already hard, so five should be five times the work. The reason it isn’t comes down to two things: a “second brain” that keeps everything in one place, and AI tools that handle the parts I used to do by hand.

This post is the overview of that system. It isn’t a finished, perfect setup — it’s the one I actually use right now, including the parts that still break.

Quick conclusion

A solo creator can run several blogs at once when two things are true:

One source of truth. Everything — rules, notes, drafts, logs — lives in a single Obsidian vault, so I can pick up any blog from a cold start.
AI does the repeatable work. Research, drafting, fact-checking, polishing, and formatting are split across AI tools by what each one is good at.

The vault keeps me consistent. The AI keeps me fast. Neither one alone is enough.

What this article covers

The problem with running multiple blogs solo
The four parts of my system
What actually works, and what still doesn’t
The smallest version you can copy today

I’ll save the specific configs and prompts for a later, more detailed write-up. This post is the map, not the manual.

The problem: solo blogging doesn’t scale

When you run a blog alone, three things quietly cap your output.

Time. You are the writer, editor, SEO person, and publisher.
Consistency. Each blog needs its own voice and rules, and they drift when they only live in your head.
Memory. Come back to a project after two weeks and half the context is gone.

Add a second blog and these problems don’t add up — they multiply. So the goal was never “work harder.” It was to move the rules and the context out of my head, and to move the repeatable work onto AI.

The system, in four parts

1. A single Obsidian vault as the second brain

Everything lives in one Obsidian vault that syncs across my devices. Not five vaults — one, with a folder per blog.

Each blog folder holds the same structure:

AI context (the instructions I hand to AI before any work)
Knowledge (article rules, SEO and style guides)
Session logs (what happened, what I learned)
A workspace (where articles get built)
Briefings (the weekly plan)

The point of one vault is simple: I never start from zero. When I open any blog, the rules, the past decisions, and the templates are already there. The vault remembers so I don’t have to.

2. AI division of labor

I don’t use one AI for everything. I split the work by what each tool is good at.

Research and spot checks — fast information gathering and image work.
Drafting — turning an outline into a real first draft.
Fact-checking and voice — catching wrong numbers and keeping the writing honest and human.
Formatting and mechanical steps — converting the final text into clean blocks for WordPress.

The rule of thumb: anything that needs judgment or voice goes to the AI that’s better at nuance. Anything structural or repetitive goes to the one that’s cheaper and faster for that job. I choose per article, partly to manage cost.

3. Reusable skills and templates

This is the part that makes five blogs feel like one workflow.

Every article is one folder with the same five stages:

1_research — notes and sources
2_drafts — the first draft and post settings
3_reviews — the fact-checked, cleaned-up version
4_exports — the final file I paste into WordPress
5_check — the post-publish review

On top of that, I built a small set of reusable “skills” — basically saved procedures the AI follows: set up the workspace, write the draft, fact-check and refine, prepare the export, sync after publishing, and run a retrospective. Same names, same steps, every blog.

One more file ties it all together: a handover note. At the boundary of each stage, whoever did the work updates a short HANDOVER.md — what’s done, what’s next, who’s next. Because of that note, any AI (or me, a week later) can resume the article cold without re-explaining anything.

4. The WordPress publishing flow

The last stage is publishing. The final file goes to WordPress as a draft through the REST API, and I publish manually after a last look. Keeping a human at the publish button is deliberate — it’s the cheapest insurance against an automated mistake going live.

What works

I never lose context. The vault means a two-week-old project is still fully there.
Five blogs, one voice each. Rules and style guides live in files, not memory, so nothing drifts.
Speed where it counts. Templates and skills clear away the busywork, so my time goes into the writing that actually matters.
Learning compounds. Each session ends in a short log, and a regular retrospective turns those logs into updated rules. The system slowly teaches itself.

What did not work

I’d rather be honest than impressive, so here are the parts that bit me.

I can’t fetch my own sites. Server-side blocking means my AI can’t read some of my live pages, so post-publish checks take a detour through another tool. Annoying, and still unsolved.
A category mismatch slipped through. A draft’s metadata said one category, the publish settings said another. Now I lock the category before the export step.
Auto-publishing broke once. Calling the WordPress API directly produced an empty body and garbled tags. The fix was to stop improvising and always go through one tested script.
My native language leaks into English. Writing an English blog as a non-native speaker, the translated tone creeps in. That problem got its own style guide — and it’s the subject of my next post.

None of these are reasons to stop. They’re just the real cost of building in public.

What you can copy today

You don’t need five blogs or custom tooling to get most of the benefit. Start with the smallest version:

Pick one app as your single source of truth. Obsidian, or whatever you’ll actually open.
Define one repeatable shape for an article. Even just research → draft → review → publish folders.
Write one instruction file per blog. Voice, rules, things to never do. Hand it to your AI every time.
Keep a one-page handover note. Update it whenever you stop, so future-you starts warm.

That’s the 20% that delivers most of the result.

FAQ

Do I use a different AI for each blog?
No — the same tools across all five. What changes per blog is the instruction file and the style guide, not the tools.

Isn’t one vault for five blogs messy?
The folder-per-blog structure keeps them separate, and shared rules live in common files. One vault is exactly what makes cross-blog consistency possible.

Why publish manually if everything else is automated?
Because a human glance is the cheapest way to stop an automated error from going public. The draft is automated; the decision to publish isn’t.

Can a beginner copy this?
Yes. Skip the custom skills at first. One notes app, one article shape, and one instruction file will already change how much you can handle.

Next steps

This was the map. Over the coming posts I’ll zoom into each part — the vault structure, the AI division of labor, the reusable skills, and the publishing flow — with the actual setup. Some of the deepest, step-by-step configurations may turn into a more detailed write-up later.

The next post digs into the part that gave me the most trouble: writing an English blog as a non-native speaker, with AI.