Problem
The landing-page hero presents a roadmap statistic as a generic “Formalized N%” headline. It does not identify the denominator as decomposed, declaration-bearing DAG leaves, so readers can reasonably interpret it as progress through the named book or thesis.
The current implementation:
- counts only formalizable leaf nodes;
- treats
stated as settled, although that state explicitly means the theorem proof is absent;
- computes the headline with
round(100 * done / total).
This creates three separate accuracy problems:
- Mapped, deferred, undecomposed, and out-of-scope source material disappears from the headline without the scope being shown.
- A statement-only theorem can count as finished when its dependencies block it (
stated), while the same statement does not count once it becomes ready to prove (can_prove).
- Incomplete work can display as 100%; for example, Python rounds 199/200 to 100%.
Reproduction
Hartshorne (historical deployment)
When this issue was filed on 2026-09-19, the latest successful Pages deployment was run 34066433925, built from commit 3b73cddb…. Its landing page displayed:
100% Formalized
68 of 68 items settled
At that commit, the coverage contract limited scope to Chapter I §§1–3, pages 1–23, while pages 24–420 were untouched.
This is a pinned historical example; later deployments may change the counts without changing the renderer semantics reported here.
Bundled Cabannes example
Autoform own fixture says it develops one representative chapter in detail. Its coverage table has one DECOMPOSED area, five MAPPED areas, and one OUT area.
From the repository root:
cd skills/setup/assets/cabannes-thesis-project
output=$(mktemp -d)
autoform render blueprint --output "$output" --lean-root .
grep -oE '>[0-9]+%<|>[0-9]+ of [0-9]+ items settled<' "$output/README.md"
Output:
>71%<
>5 of 7 items settled<
That is 71% of the seven decomposed targets, not 71% of the thesis.
Expected behavior
The homepage should distinguish source coverage from progress inside the decomposed roadmap. For example:
Decomposed roadmap: 5 of 7 targets complete (71%)
Source coverage: 1 decomposed · 5 mapped · 1 out
Exact wording is flexible, but:
A whole-source percentage is not required; source areas are rarely equal-sized. A clear scoped count plus coverage dispositions is sufficient.
Related groundwork: #8 and #19. The future aggregate progress work in #24 should consume the same clarified semantics.
This issue concerns progress semantics only. Deployment freshness and Markdown-link rewriting are separate concerns.
Problem
The landing-page hero presents a roadmap statistic as a generic “Formalized N%” headline. It does not identify the denominator as decomposed, declaration-bearing DAG leaves, so readers can reasonably interpret it as progress through the named book or thesis.
The current implementation:
statedas settled, although that state explicitly means the theorem proof is absent;round(100 * done / total).This creates three separate accuracy problems:
stated), while the same statement does not count once it becomes ready to prove (can_prove).Reproduction
Hartshorne (historical deployment)
When this issue was filed on 2026-09-19, the latest successful Pages deployment was run 34066433925, built from commit
3b73cddb…. Its landing page displayed:At that commit, the coverage contract limited scope to Chapter I §§1–3, pages 1–23, while pages 24–420 were untouched.
This is a pinned historical example; later deployments may change the counts without changing the renderer semantics reported here.
Bundled Cabannes example
Autoform own fixture says it develops one representative chapter in detail. Its coverage table has one
DECOMPOSEDarea, fiveMAPPEDareas, and oneOUTarea.From the repository root:
Output:
That is 71% of the seven decomposed targets, not 71% of the thesis.
Expected behavior
The homepage should distinguish source coverage from progress inside the decomposed roadmap. For example:
Exact wording is flexible, but:
A whole-source percentage is not required; source areas are rarely equal-sized. A clear scoped count plus coverage dispositions is sufficient.
Related groundwork: #8 and #19. The future aggregate progress work in #24 should consume the same clarified semantics.
This issue concerns progress semantics only. Deployment freshness and Markdown-link rewriting are separate concerns.