Rating
Every artifact carries a score, including the ones that score badly.
Each artifact on this site is scored 1–10 on five axes and published with the result. The scale is not a review score — it answers one question: how much does this artifact earn the time it asks for?
On cards and in search results it appears as editor 8.4, not rated 8.4, and the wording is deliberate: a bare “rated” reads like a crowd score, and this is not one.
It is marked [derived] everywhere it appears, which on this site means a judgement
made against a written rubric, not a measurement. One reviewer, no panel, no reader votes.
The weights below were fixed before any artifact was scored, and changing them retroactively
would make the whole set meaningless — so they are published here rather than kept in a config.
| Axis | Weight | Question |
|---|---|---|
| useful | 30% | is it usable tomorrow |
| evidence | 25% | a real incident and a number, or an assertion |
| pull | 20% | would anyone open it and finish it |
| original | 15% | is the same thing in fifty other blogs |
| form | 10% | tl;dr, scannability, length against content |
The weights follow what this site claims to be. It says it is an operating record and that the files can be used tonight, so useful and evidence carry more than half the score between them. A well-written artifact with nothing behind it lands around 6.
Distribution
The low end is not an accident of grading. Five development logs from the first attempt at publishing sit at the bottom of the table; they describe a newsletter this site replaced and they are kept because removing them would be tidier than it is honest. A score is the cheapest way to say this is here for the record, not for you without hiding it.
By section
| Section | Artifacts | Mean |
|---|---|---|
| Receipts | 1 | 8.25 |
| Architecture | 8 | 7.94 |
| Failures | 11 | 7.68 |
| Playbooks | 12 | 7.66 |
| Patterns | 23 | 7.62 |
| Downloads | 5 | 7.07 |
| Start | 6 | 6.01 |
| Log | 7 | 5.39 |
What the score does not say
It is one reader's judgement, applied to the author's own work, which is the weakest possible review arrangement and is stated here rather than glossed over. It says nothing about whether a claim is true — that is what the incident, the number and the date beside it are for. And it is a snapshot: an artifact that gets a real revision should be rescored, and the date of the scoring pass belongs beside the number the moment there is more than one pass.
Machine-readable: rating, ratingAxes and ratingKind
travel in index.json and in every .md mirror.