Blog

Research May 2026
We stress-tested our factor claim. Here's what we found.

The most-stable-language quintile earns 93.2 bps/month (t=9.16). We tried to break it three ways — factor adjustment, removing the normalization, and the Lazy Prices benchmark. It held.

Case Study May 2026
What's going on with Chevron?

Chevron's FY2025 10-K flagged at 1.4× the distress ceiling — its risk-factors section quintupled after the $53B Hess acquisition closed. A clean structural false positive: a merger, not a meltdown. The stock is near all-time highs.

Research May 2026
We analyzed language drift in 4905 SEC 10-K filings over 10 years. Here's what the data shows.

The backtest across 7069 flag events. How the signal is constructed, what it predicted, and where it missed.

Case Study May 2026
What 10-K language drift actually looks like — four annotated examples

SVB, Bed Bath & Beyond, Rite Aid — plus Party City, a documented miss. What the filings said, why the scores were what they were, and what happened next.

Case Study May 2026
SVB's final 10-K: what the language showed before the bank run started

The filing date was February 24. The FDIC arrived March 10. What happened in between, and what that window means.

Research May 2026
Why we used embedding models instead of LLMs — and when that's the right call

Determinism, cross-sectional comparison, no context window — five reasons the tool most people reach for was the wrong one here.

This site uses a session cookie for authentication. We also use Plausible Analytics, a privacy-friendly, cookieless tool that collects no personal data and requires no consent under GDPR. See our Privacy Policy.