Get independent Run 003 eval from an external user

[ardamoustafa1]

Goal

Fill the proof gap: a third-party run using evals/RUN_TEMPLATE.md, raw outputs, and blind judge scores.

Acceptance criteria

• External runner, not the maintainer
• Same model/settings in control and treatment arms
• Raw generated artifacts attached or linked
• Validator logs attached
• Three judge passes or human judge notes included
• Results added to evals/RESULTS.md, positive or negative

Starter path

Open an issue using the Independent eval template or PR a completed evals/runs/run-003.md.

[ardamoustafa1]

Repo-side prep completed in 4ece44c and ca6adfb: added evals/RUN_TEMPLATE.md, an Independent eval issue template, and launch/outreach copy asking for Run 003. This stays open until a real external runner submits raw outputs; filling it ourselves would weaken the honesty doctrine.

[ardamoustafa1]

Opened a public Discussions call for external runners: #8

Keeping this issue open until a real external Run 003 PR lands.