How I work with my AI assistant

These are the rules I hold myself to on this site, whichever piece you are reading.

Every word on the English side of this site is mine, except where a piece tells you otherwise, on its face.

The assistant measures, writes code, checks my memory against the record, and keeps the files. It also helps me keep a list of candidate topics I plan to write about. It does not write the pieces, and if one ever starts from a machine draft, the piece says so. Where a piece is machine prose, the piece says so on its face.

Every number is either measured on hardware I own or cited to something you can open and read yourself. Numbers I cannot stand behind do not go in.

When I am not sure of something, the piece says so. An uncertain figure is labelled uncertain, and an experiment I did not run is labelled not run.

Nothing goes in that I have not checked. If the assistant found it, fine. If I cannot verify it, it is not published.

The assistant does not decide what gets published. I do.

I name the tools I use. There is no reason for that to be a secret.

The disagreements get published too. When the record contradicts my memory, the correction goes in the piece, not in a drawer.

How AI assistant translates from English to Turkish

Turkish is my native language. I write the pieces on this website in English. Translation to Turkish comes afterwards, has no cadence.

It is a simple process:
My AI assistant creates a draft translation.
I edit it, usually 2-3 paragraphs at a time. The agent passes over the edits and saves rules about my "voice" and specific word translations. This is the foundation for future translations, a style sheet of sorts.
When I am finished, it performs a mechanics pass for typos.

That is all. For reference: Statistics for the translation of "Seni bekleyen bir şey yok".

While this was happening, 48 style rules were recorded (the number is high because this was the first translation) and 30 glossary rows added.

Last updated 2026-09-25.