The fitting
We say what we ran, and what we did not.
SummaryA vendor benchmark is theirs until we repeat it.
A tested note names the build, the machine, the command, and the result. It also names the part we left alone. Issue Nº 01 is the pattern. We downloaded the Graphicle v0.1.0 Linux binary, checked the checksum, and ran it on a two-file Python repository. Setup took 0.2 seconds. A blast-radius query on price_with_tax() returned the callers we expected. We did not test a live agent over MCP, and we did not time a large repository. The speed claim in the README stays with the maker.
If the tool needs a Mac and we do not have one, the piece says Not tested. If it needs an account or an extension we did not install, same label. The description then comes from the official site, and we say so.
We do not invent quotes, prices, or benchmark scores. A figure in a dispatch is in the source list at the bottom, or it is not in the piece. News is reported. The weekly three is a recommendation, and only when the label matches the work.
The fitting for Issue Nº 01 is in the archive.