Actually put Gemma 4 26B to work on something real: extract trading signals from 2,400 earnings calls

Actually put Gemma 4 26B to work on something real: had it extract trading signals from 2,400 earnings calls. One worked. One almost fooled me.

A Reddit user fine-tuned Gemma 4 26B on 800 labeled earnings-call transcripts and ran inference on 2,400 transcripts over 3 years on one RTX 4090 in about 14 hours. On 600 out-of-sample transcripts, one signal linked vaguer CFO guidance to about 1.8% sector-relative underperformance over 5 days with IC 0.04. A stronger signal showed 0.85 correlation with sector returns after checks and was discarded as a ghost factor; the key point is factor sanity checks, not the profit claim.

Why it matters: Strong HKR-H/K/R: this is a named first-person experiment with concrete setup, metrics, and a useful negative result. It stays at featured, not P1, because it is one Reddit test rather than a product release or industry-wide event.