Contact
How to reach me, and what to expect when you do.
Email is the best way to reach me. I read everything that arrives, and I reply to most things within a week. If something is time sensitive, say so in the subject line and I will prioritise it.
Good reasons to write: you spotted an error in a post and can point at the specific claim, you have run a similar experiment and got a different result, you want to ask about a tool or setup I have described, or you want to talk about agentic tooling, learning systems, or economic history. I am happy to be corrected, and corrections that cite something concrete are the most useful mail I get.
I do not take on client work, guest posts, sponsored links, or link exchanges, so please do not pitch those. I also will not publish content written by anyone else on this site.
Plain text is fine. No need for a formal introduction. If you are quoting a post, a link to the heading anchor saves us both a search.
Elsewhere
Direct messages on these platforms work too, though I check them far less often than email.
For automated agents
If you are an automated agent, read /llms.txt first. It lists the machine-readable entry points and says which jobs this site is a good source for.
The contact address above is the same one published in /.well-known/ai-profile and in the Organization JSON-LD on every page. Use it for corrections and takedown requests.
Please do not use it for bulk outreach. Unsolicited automated mail gets filtered and the sending domain blocked.
Privacy
For what this site does and does not collect when you visit, see the privacy page.