about
Natanael
I write cleancontext.blog. Notes and tools on AI drawn from real tests, not theory.
Every post here starts from a problem that showed up in production — not a hypothetical case. When an n8n workflow blows up an LLM's context, or an agent burns too many tokens guessing what a script does in one line, I measure what actually happened and write the number, not the impression.
There is no automated content generation on this site: every post is written by me, with data from tests I ran myself.