Continue improving IDKMesh’s introductory documentation for visitors without specialist AI or software-engineering experience. Follow the completed beginner clarity work with an interactive experience, while saving work in this repository.
Add an optional browser-only page under docs/labs/reviewer-lab.html, using a short first-contact walkthrough: inspect the failed p01 known-bad probe, see overlapping mistakes on c10, select a subset of reviewers, and change the strict approval threshold. No model account, Python installation, credential, server, provider call, external script, analytics or user-data upload is required. Keep the other static documentation pages unchanged in execution behavior; the lab is an explicit exception to the no-JavaScript presentational site default.
The lab embeds the complete, unchanged examples/gate-audit/panel-votes.example.json synthetic fixture; tests compare both documents structurally and check the actual browser calculator for every non-empty reviewer subset (31) against three quorum values (93 cases). The strict threshold mirrors idkmesh/gate_audit.py: accept when accept_votes > quorum * number_of_selected_reviewers, ties reject. The page intentionally counts ordinary-candidate false approvals and false rejections separately from breached probes. It displays pairwise overlapping mistakes, not correlation, statistical independence or effective votes; those belong to the real Python gate-audit implementation.
Default (all five; quorum 0.5): 1 false approval (c10), 1 false rejection (c03), 2/3 probe breaches (p01 and p03). At 0.67: 0 false approvals, 2 false rejections, 0 probes breached. At 0.8: 0 false approvals, 4 false rejections, 0 probes breached. All are properties of the synthetic sample and do not support real-world reviewer quality claims.
Promote the lab from the README, beginner guide, published Start page and docs navigation. Include the new HTML URL in the GitHub Pages sitemap while preserving its generated structure; keep canonical/OG structured metadata and published-page constraints aligned. A dedicated Python test checks the exact fixture, UI accessibility markers, offline code boundary and Node parity when Node is available.
synthetic; changing the voting rule does not change recorded reviewer decisions or ground truth.idkmesh gate-audit-ui service and the read-only Human Control Tower.A curious visitor can experience a meaningful, falsifiable behavior before cloning the repository. No traffic, accessibility study, or external-user comprehension improvement is claimed without measurement. The intended next evaluation is to ask first-time visitors to distinguish recorded verdicts, known correct answers and acceptance authority.
AI/tool provenance: authored with ChatGPT GPT-6 through the connected GitHub interface, source-matched to the committed fixture and Python audit rule. No independent human review is asserted.
PR #1089 merged as
a0cd473de57059ca376e8fb88fb693bdbe0d1a23. The initial Python gates
failed because this conversation record had not been added to the conversation
index and sitemap. Subsequent changes restored both references, and the
required Python 3.11 and 3.13 PR checks, CodeQL and observatory checks passed
for the exact reviewed PR head.
The GitHub Pages build and deploy
reported success for the same merge SHA, with the public environment URL
https://mskazemi.com/idkmesh/. This establishes a successful Pages deployment
job, not an independently observed browser session at the new route.
A deterministic DOM-state simulation of the actual browser script checked the five-reviewer default, fifteen candidate options and table rows, the 67%-rule result, candidate inspection, reset and last-reviewer safeguard. The dedicated Node test added in a subsequent PR makes the interaction checks part of CI. The real-user comprehension study described in PAGES_SETUP remains unperformed.