Notizen aus dem Bau.
Architekturentscheide, die Überlegungen dahinter und die Teile, die sich als falsch herausgestellt haben. Geschrieben für Leute, die wissen wollen, wie etwas funktioniert, bevor sie ihm vertrauen.
Die Beiträge erscheinen nur auf Englisch.
- Architecture6 Min. Lesezeit
Why a model has to be able to stay quiet
Fluency and correctness are separate abilities. Models got good at the first one much faster, and the gap is where the damage happens.
Lesen - Models5 Min. Lesezeit
What a small model can actually do
The honest version: the line runs between what the context brings and what the model must recall unaided. Everything else follows from that.
Lesen - Data4 Min. Lesezeit
Where the servers are, precisely
Most pages answer this with a flag. Here is the actual split, including the part that is not in Switzerland.
Lesen - Tools5 Min. Lesezeit
An agent that works it out instead of guessing
Fifteen tools now, and every one of them exists because a language model sounds confident about something it cannot actually do.
Lesen - Pricing4 Min. Lesezeit
Limits that have actually been worked out
Why our usage limits are not published as numbers, and how they are derived behind the scenes.
Lesen