How we built Signal Persona Research Corpus (and why we shipped it externally)
The internal-tool phase
Signal Persona Research Corpus didn't start as a product. It started as F1503/F1641 persona registry — an internal tool we built for our own customer apps. We needed it. We built it. It worked.
The "wait, others might want this" phase
After running Signal Persona Research Corpus in production for months across multiple internal apps, the pattern became obvious: every team building similar workloads hits the same wall. The build-it-yourself path is 4-12 weeks of plumbing. The buy-it path costs 5-10x more than it should.
We had the engine. We had the runbook. We had the bug-fixes. The only step left was wrapping it in pricing pages and signup forms.
What changes when you ship externally
Three things changed when we exposed Signal Persona Research Corpus as a product:
- Documentation got real. Internal Slack threads don't count as docs.
- Edge cases got prioritized. Single-tenant code makes assumptions that break the moment a second tenant arrives.
- Pricing got honest. No more "free internally" hand-waving.
What stayed the same
The engine. The code path that handles your request is the exact same code path serving our own apps. We don't ship a stripped-down "public" version. We ship the one we use.
Try it
Free tier: /contact/. We're a small team. Reply directly to any email and you'll get a human.