The engine watches each competitor's public surfaces, turns changes into typed events, scores them, then measures whether the market actually reacted. Read-only by architecture. Nothing invented along the way.
Every company you track runs the same five-stage path — deterministic, and traceable back to the exact bytes we fetched.
Changelogs, blogs, careers pages, pricing, GitHub releases, ATS boards, Hacker News, Bluesky, news. No logins, no scraping behind auth, no private data. If it isn't public, it isn't in the engine.
A diff on a changelog becomes a ship. A new role becomes a hire. A tier change becomes a price move. Each event keeps the source artifact, so every claim traces back to bytes with a timestamp.
Scoring is deterministic and de-saturated: most moves are correctly small, a thin tail is genuinely big. Because there's no model in the source-of-truth path, nothing is hallucinated, and every score carries the exact features that produced it.
We couple real reactions — Hacker News, Bluesky, GitHub, news — to each launch by content match and timing. Points, comments, sentiment, and the momentum over the days that follow. A big ship that shipped to silence is a very different signal from one that roared.

What shipped across your set, whether it landed, and what it means for you. Forwardable in one email. Teams get it role-routed — engineering, design and security each read their slice.
We show what we can and can't see per company. We never fake completeness.
Look at anyone, notify no one. Your research stays your side of the firewall.
Can't verify it in ten seconds? It doesn't ship in the brief.
Same inputs, same score. The features are attached to every number.
Track your set and get the landing report — built from live, cited data.