publicity: ClawBench launch story / data points pitched to LMSYS, Anthropic, OpenAI, Google evaluator teams

[reacher-z]

Problem

Frontier-model providers run their own evaluation suites for marketing claims. They don't currently include ClawBench results. Two reasons:

1. We haven't pitched them.
2. The path to "run our benchmark" wasn't frictionless. (Now it is — pip install clawbench-eval is on the homepage.)

A single mention in an Anthropic / OpenAI / Google model launch blog post would 10x our inbound traffic for that quarter.

Action

1. Identify contacts at provider evaluator teams (model-card maintainers, internal eval leads):
   • Anthropic: model cards maintainers
   • OpenAI: evals team
   • Google DeepMind: Gemini model-card team
   • Alibaba / Z.AI / DeepSeek: open-source evaluator teams
2. Pitch packet: 1-paragraph description + 4-stat snapshot (corpus, platforms, top model, two-stage methodology) + pip install clawbench-eval reproduction line + offer to help integrate
3. Track outreach in a outreach/ directory: who, when, response, follow-up
4. Once accepted, write a blog post ("ClawBench in Anthropic's Claude Sonnet 4.X model card") for compounding press

Acceptance

• Outreach list with ≥10 named recipients across 5 providers
• At least 1 personalized email/DM sent to each by 2026-06-15
• ≥3 responses recorded
• If any provider integrates: a follow-up blog post under /blog (publicity: embeddable leaderboard widget — iframe at /embed/leaderboard for blogs + benchmarks aggregators #186)

Why now

The publicity flywheel needs external validation, not just on-site polish. Provider mentions are the highest-leverage external validation available — and we've never explicitly tried to land one.