Scoreboard
Every Refacto Agents story ends with a prediction — a concrete, dated claim about what will or won't happen — and a falsifiable condition that says when we're right or wrong. This page is the public tally. Misses don't get quietly retired. Readers can up- or down-vote each prediction.
Season record · since launch
-
JUL 9 2026 Medium confidence
By the time Ollama reports its next major user milestone (following the 9M reported at its July 2026 raise), at least one open-weight model will be the documented default for a routine, high-volume ad-tech classification task (brand safety or audience tagging) at a named vendor, with frontier models reserved for lower-volume reasoning workloads.
Why Reasoning-model token costs are rising (today's AI-cost-surge brief) while open-weight tooling is getting real funding and 9M users; high-volume/low-value ad-tech calls are exactly the workload where per-decision token cost forces migration to cheap models.
Right if: a named ad-tech vendor publicly documents an open-weight model handling routine classification while frontier models handle harder reasoning. Wrong if: credible reporting shows frontier models still dominating routine, high-volume ad-tech classification with no open-weight migration.
20VC: Open Models vs Frontier Models: Who Actually Wins? | The $100,000 Token Budget Every Engineer Will Need | Why Forward-Deployed Engineers Are the Future of Enterprise AI with Clay Bavor, Co-Founder of Sierra Listen to the episode →
PendingRevisit Dec 31, 2026
Your take?