Form FN-0 · The rating scale · read this first

How I rate my work

Every file carries two marks. The evidence stamp says how well its main claim has been tested. The contribution line says what kind of contribution it is. Both change only when there is new evidence, and every stamp links to that evidence.

How to read a file header

E2 · Cold-replicated

Contribution: Reproduction of [paper or result]

Click the stamp to see the run, review or source behind it.

1 · Evidence stamp · on every file

StampMeansWhat puts it thereLinked evidence
E0 · UntestedA conjecture or speculationNothing proven yet. Drawn dashed everywhere, including callouts and links on the map.none, by definition
E1 · Self-checkedI re-ran or re-derived it and it matchedA rerun, recalculation or check against the published value. For known results, checked against the primary source.the run, notebook or source
E2 · Cold-replicatedRebuilt weeks later without notesI rebuilt it from scratch and could justify every step. Pass and fail criteria were fixed before I started.dated rebuild + comparison
E3 · ReviewedA qualified person tried to break itSomeone with the right expertise was asked for the weakest step and replied, or it was presented with the questions logged.who, when, what they found
E4 · ExternalPeer-reviewed, or reproduced by someone elseReviewer reports in hand, or another person reproduced the result independently.the review or their reproduction
✗ · Error foundA gap, error or failed rerunThe claim stays up, the error is shown, and the fix goes in the amendments log. Usually the claim is narrowed to what holds.the error and the amendment

2 · Contribution line · on work files, and essays that argue something

LevelShown asNeeds
L1Learning notesNothing beyond the notes themselves. Understanding existing material, in public.
L2Reproduction of [paper or result]Same result as the original, rebuilt by me (E1 or better).
L3Independent re-derivation of [result]Dated notes or commits showing I got there before reading the known result. Without them it stays L2.
L4New application of [method] to [problem]A literature search showing it isn't already done, a named difference from the closest prior work, and a real open question it answers.
L5Original result, externally checkedEverything L4 needs, plus E3 or E4: a qualified expert tried to break it and it held.

The numbers stay in my validation log. Pages show the words.

3 · Kinds of claim inside a file

  • Conjecture A guess worth writing down. Always E0, always dashed.
  • Observation Something seen in a run or experiment.
  • Derivation Worked out step by step, steps shown.
  • Result A finding with its own E mark beside it.

4 · House rules

  1. 01Only evidence moves a mark. Nothing goes up because it feels finished.
  2. 02Every mark links to what earned it.
  3. 03Errors stay visible. Amendments are dated and never deleted.
  4. 04No file calls itself novel or original unless its contribution line is L4 or L5.
  5. 05My own review, and AI review, don't count as review. E3 needs a qualified person.
  6. 06Only work that passes my sharing check is filed here. Some directions stay private on purpose, so a gap on the map is a choice, not an absence.

5 · Where this scale comes from

The numbered ladder borrows its shape from NASA's Technology Readiness Levels and medicine's GRADE certainty ratings. The habit of fixing what each word means comes from the IPCC's calibrated uncertainty language. Reproduced and replicated follow the ACM's artifact badges. The claim kinds are the ones maths and physics papers already use.

On file today

E0 · 1E1 · 23 files · ✗ none on file