Sharpen positioning from both review panels…

Details

Final pass folding the audit panel + the external-persona panel feedback,
keeping every claim honest:
- hero "bet" line gets a contrarian edge (names the false consensus: scale
  alone won't make AI trustworthy) — reads as conviction, not a destination
- proof bar reordered to lead with the strongest signals (18 awards, top
  venues, national HPC), made louder, and made verifiable (papers->Publications,
  awards->Awards links); affiliations row higher-contrast with the AAAI
  reviewer credit accented
- honesty fix: "9 peer-reviewed papers" -> "9 papers & preprints" (the 9
  include a preprint + a thesis); matching stat relabelled; proof venues
  narrowed to accepted ones (AAAI/ICLR/CIKM), NeurIPS stays "under review"
- projects: capability-first claims; PyDiffGame star claim made exact ("the
  most-starred project I maintain") with a live, falsifiable stars badge that
  hides on load failure; PyEDCR card gets a "Run the demo" link to the explainer
- contact de-duplicated and flattened ("Get in touch."), CV demoted to a quiet
  link, scarcity line removed
Panels: 5/5 audit reviewers SHIP; 5/5 external personas WOULD_PURSUE
(established-feel 8/8/9/8/8). Career/founder intent remains off the site.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HShRpQUmDUttxD5F89ziDu

Files changed (6)