NoCap public-site staging + documentation

Staging tree for the public NoCap evidence page and stranger-facing markdown docs.

Path	Role
`index.html`	Staging source (edit here; sibling may promote)
`../cf-pages-nocap-agent/`	Production deploy directory (currently `index.html` only)
`DOCS_INDEX.md`	Start here for the documentation set
`data/README.md`	Public CSV catalog + recompute (jury-ready)
`WORKSPACE_MAP.local.md`	Private real profile ↔ code map — never deploy
Live	https://nocap.think2earn.com

Workspaces

Public pages name workspaces by code (Clock-W, Claim-W1, Claim-W2, Research-W*, Archive-W*). Do not copy WORKSPACE_MAP.local.md into cf-pages-nocap-agent/.

Documentation (strangers + agents)

See DOCS_INDEX.html for reading orders and the full catalog.

Doc	Role
CAMPAIGN_OVERVIEW.html	Causal story
GLOSSARY.html	Terms
FAIR_CLOCK.html	Speed rules + non-claims
HOW_TO_READ_EVIDENCE.html	Resolution order + CSV recompute
Agent guide	`../nocap-repo/docs/AGENT_HANDOFF.md`

Hosting `.md` on the live site

Second deploy required. Production currently ships only index.html. Companion markdown is durable in this tree; it is not live on nocap.think2earn.com until an operator deploys selected .md files (or the whole staging folder) excluding WORKSPACE_MAP.local.md.

Preview

xdg-open /home/q/Downloads/software_i_built/modal/cf-pages-nocap-agent-dev/index.html
# or (serves .md too for local review)
python3 -m http.server 8765 --directory /home/q/Downloads/software_i_built/modal/cf-pages-nocap-agent-dev

Requires network only for Google Fonts. Offline, the page still reads.

Claim language

Bounded headline (aligned with nocap-repo/results/claim_stack_v1/MULTI_SEED_CLAIM.md):

On three independent seeds (1337, 2029, 4242), the claim-stack candidate—Keller Jordan’s Muon with B32 plus timing-cheap split-QKV—first reaches val_disjoint ≤ 3.3821 in 3.553 h, 3.567 h, and 3.511 h. That is 8.1%, 7.7%, and 9.2% less first-passage train time than our sealed stock AdamW F1 baseline at 3.866 h (mean ≈3.544 h, mean ΔT% ≈+8.3%).

Also on-page / in docs:

Comparison is against our sealed F1 — not an official NoCap win.
Joint-QKV stock control missed quality on 1337/2029; seed 4242 had no stock arm.
Matched candidate-vs-stock first-cross ΔT% is N/A.
GPU: A100-SXM4-40GB.

Accuracy pack

NUMBERS_CANON.md — canonical numbers (workspace codes)
FACT_CHECK_CHECKLIST.md — claim → source
ACCURACY_NOTES.md — residual conflicts
ACCOUNT_MAP.md — public roles via codes
WORKSPACE_MAP.local.md — real profiles (local only)

Promote / deploy

Default (current prod practice — HTML only):

cp /home/q/Downloads/software_i_built/modal/cf-pages-nocap-agent-dev/index.html \
   /home/q/Downloads/software_i_built/modal/cf-pages-nocap-agent/index.html
cd /home/q/Downloads/software_i_built/modal/cf-pages-nocap-agent
wrangler pages deploy . --project-name=nocap-agent --commit-dirty=true

Confirm deploy directory contains only intended public files (no WORKSPACE_MAP.local.md, no EMAIL_SNIPPET.md). Do not deploy unless an operator asks (sibling may own live deploys).

Changelog

2026-07-17 — jury-ready pass

Author byline on the live report: Carol Calin (ML Engineer and NLP Researcher), contact via mailto.
Verify section (#verify): first-passage chart + same-origin data/*.csv downloads + recompute formula (first row val_disjoint ≤ 3.3821; T_h = train_time_ms/3.6×10^6; ΔT% vs F1_seal 3.866 h).
Challenge-terms mapping strip: official NoCap-Test rules ↔ our measurement (val threshold, non-4090 baseline seal, 1 GPU, fair comparison / documented recipe).
Compressed workspace map: Clock-W / Claim-W* roles; isolated workspaces are volume/provenance/clock separation — not a leaderboard.
Reviewer-facing section replaces generic “Next”: application context, seal dates, engineering practices — still not an official NoCap win.
data/README.md — CSV catalog + recompute notes; seed 1337 summary-only (no CSV in bundle).
Claim language unchanged: under-baseline vs F1_seal — not an official NoCap win; matched candidate-vs-stock ΔT% N/A.