Time-in-market.

How we derive this

Each operating_since year combines three independent signals. We never stamp a year from a single source — and when sources disagree we either pick the most-conservative one or surface the conflict in a tenure_note.

1. WHOIS / RDAP — domain registration year
2. archive.org — first 200-OK snapshot of the project page (not just the root domain)
3. Public X record — account creation date + earliest project-mentioning post
4. Operator self-claim — About / FAQ / footer copyright — cross-checked, never trusted alone

Known traps we explicitly guard against (caught publicly by Monerica + Jaun on 2026-05-28 — thanks):

• Premium-domain trap — crypto.com 1993 ≠ Crypto.com project 2016. kagi.com 1993 ≠ Kagi search 2018. Pre-2000 WHOIS years are always cross-verified.
• Shared-host trap — youtube.com, reddit.com, github.com WHOIS year ≠ the channel/subreddit/repo creation year. Our probe refuses to auto-write when the linked domain is a shared platform.

Full methodology + verification cadence: /methodology