About DobyClick
Notes from running LLM agents, written and reviewed by the operator who runs them.
Why this site exists
Most writing about LLM agents explains how they should work. This site records how they behave when you run them on real work: the sub-agent that stopped without an error, the tool permission that was wider than it needed to be, the handoff that dropped the one detail that mattered. Each post explains one concept plainly, then shows one place where it met a real run.
One concept per post
A single idea, explained without jargon first, then tied to one concrete run.
Evidence over opinion
Prompts, outputs, logs, costs, and failure conditions come from the operator’s own records. Nothing is invented.
Reviewed before publishing
Every draft is checked against its sources before it goes live. Corrections are welcome through the contact page.
How these posts are made
LLM tools help draft and edit the posts here. Every post starts from the operator’s own records (run logs, measurements, failure notes) and is checked against those records by the operator before it is published. Drafts that have not been reviewed are not published. The byline reads “The Operator” on purpose: this site is about the work, not the person.