design iterations

Internal review hub. Each round of the improvement loop is preserved here. Same content in every version; only the presentation changes.

v7 Round 6: final pass (kept — final)
Cross-page consistency, honest bring-up timescale line, chip moderation, voice sweep, all 42 contract items verified. Final v1-vs-v7 judgment: firmware lead 34 vs 25, skimmer 33 vs 23 — v7 wins on both cards. Shared remaining flaw: the "calendly event: pending" footer. home →
v7/onboarding.html v6 Round 5: evidence artifact (kept)
The real June 18 bench report published at v6/runs/; hero run ID and run card now link to it. Marketing adjectives flattened. home →
v6/runs/bldc-hall-fault-demo.html v5 Round 4: hero run-replay (kept, judge 23 vs 12)
Empty video placeholder replaced with an animated replay of stored run 20260618-080733 (typed log, FAIL then guarded PASS), honestly labeled. Betaflight caption kept as an explicit filming-next promise. home →
v5/onboarding.html v4 Round 3: onboarding visual pass (kept, judge 34 vs 26)
Process-flow strip (hardware → documents → bring-up → verdicts), mono data chips, 3-column boundary cards. Homepage unchanged from v3. home →
v4/onboarding.html v3 Round 2: judge fixes
Both judges picked v2 over v1 (24-17, 33-29). Their shared complaint fixed: AI-free checker sentence and the DShot punchline promoted out of expanders. Onboarding link added to nav. home →
v3/onboarding.html v2 Round 1: homepage de-densified
What-is fact grid, mini-rows everywhere, evidence as stat cards with expanders. Judges: firmware lead 33 vs 29, skimmer 24 vs 17, both over v1. home →
v2/onboarding.html v1 Baseline
Content locked 2026-07-01. Full spine, step-3 mini-rows on onboarding. home →
v1/onboarding.html

New iterations appear here as the loop runs: restructure → fresh-eyes judges → content-contract check → deploy.