I’ve started and abandoned a personal site several times. The pattern is always the same: I pick a stack, spend a weekend on the design, publish one post, and then never touch it again because publishing takes more effort than the thought is worth.
So this version is built around one constraint: the cost of publishing has to be lower than the cost of not publishing.
Practically, that means:
- Posts are markdown files in a git repo. No CMS, no admin panel, no login.
- There’s an Ideas section with a deliberately lower bar. Half-formed is fine there. If something grows up, it becomes a post.
- No analytics. I don’t want to know how many people read this, because knowing would change what I write.
- Static HTML. Nothing to keep running, nothing to patch, nothing that breaks in eighteen months because a service shut down.
What goes here
Mostly the things I currently only write into scratch files and then lose:
- ClickHouse — schema design, query patterns, the failure modes you only meet in production.
- Python and Linux tooling, when I build something worth reusing.
- Agent harnesses. I’ve spent a lot of the past year on the question of how you get an LLM to do sustained engineering work rather than impressive one-shot demos. There’s more to say about that than fits in a README.
The honest part
I don’t know how often I’ll post. The archive page is grouped by year, which will look either respectable or embarrassing depending on how this goes. That seems like the right amount of pressure.