This page as Markdown, byte for byte: /walk-timings.md
The two measured walks#
Two of this product's own production checks are stopwatches, and their budgets come from the operating brief rather than from anybody's judgement: a stranger gets from the published documentation to a first authenticated call in under ten minutes, and from the published documentation to a technical price out of a model they composed themselves in under fifteen.
This page publishes what those two walks have actually measured. No figure on it is typed by hand. Each one is read from checks/walk-timings.log, the append-only record the two harnesses write a line to whenever they complete a timed walk; the page is regenerated by the same call that appends the line, and tests/unit/test_walk_timings.py fails the build if the committed page differs from what the committed record renders. A number here that nobody measured is therefore a red build, not a marketing decision.
Latest measured timings#
| walk | check | budget | latest measured | share of budget | verdict | measured at |
|---|---|---|---|---|---|---|
| signup to a first authenticated call | PROD-01 | 600 s | 14.0 s | 2% | PASS | 2026-08-27T05:10:28Z |
| public docs to a technical price from a composed model | PROD-12 | 900 s | 60.8 s | 7% | PASS | 2026-08-26T01:45:08Z |
signup to a first authenticated call (PROD-01)#
A session that has never seen this product reads the published documentation, signs itself up, and makes one successful authenticated call. The clock starts at its first request to the docs site and stops at the first authenticated 2xx from the API, both read out of a recording shim rather than out of the session's account of itself.
Latest measured run: 14.0 seconds, 2% of the 600-second budget, recorded 2026-08-27T05:10:28Z — PASS.
10 measured runs are in the record, from 2026-08-13T13:51:25Z to 2026-08-27T05:10:28Z: fastest 14.0 s, slowest 61.3 s.
Perform the walk again, and record it:
bash checks/production.sh --only PROD-01public docs to a technical price from a composed model (PROD-12)#
A session that has never seen this product reads the published documentation, composes a simple parametric model in the no-code builder out of the four primitives, takes the automatic validation pass and its forty-year replay, and uses that model to produce a technical price. The clock starts at its first request to the docs site and stops at the first successful result labelled with the model and version it composed.
Latest measured run: 60.8 seconds, 7% of the 900-second budget, recorded 2026-08-26T01:45:08Z — PASS.
16 measured runs are in the record, from 2026-08-13T14:12:14Z to 2026-08-26T01:45:08Z: fastest 27.9 s, slowest 215.0 s.
Perform the walk again, and record it:
bash checks/production.sh --only PROD-12How the record is written#
One line is appended to checks/walk-timings.log for each walk that reached its stop condition:
<ISO8601Z> PROD-NN <elapsed seconds> <budget seconds> PASS|FAILThe record is append-only. A line, once written, is never edited, reordered, re-dated or deleted, and the file is never truncated or regenerated — the same rule this repository applies to its check history and its golden corpus, for the same reason: a published number computed from a file its author may rewrite is not evidence of anything. If a line is wrong, the repair is a later line and a journal entry saying so.
A run that did not reach its stop condition leaves no line here, because it produced no elapsed time to record. Whether the check passed or failed that day is a separate record, checks/history.log, and the two are deliberately not merged: a failure with no measurement must not be able to publish itself as a fast walk.
Regenerate this page from the record with:
python3 checks/prod/walk_timings.py --refresh