Cowboy. Architect. Debugger. Tinkerer. Prompter-First. Your sessions already know — and Crixin tells you which one, with the data to back it up.
Crixin scores each one on signals from your real history — session length distribution, user-vs-assistant ratio, mean turns per session, depth of long-form runs. Highest-scoring label wins.
Many short, decisive sessions. Fast turnaround, low message count per ride. You hop in, fire one prompt, ship.
Triggers when: ≥100 sessions and <12 mean messages each.
Fewer sessions; each goes long. You stay in the flow, building one big thing at a time. Long-form planning, multiple revisions.
Triggers when: ≥30 mean messages per session and ≥10 sessions over an hour long.
Concentrated, deep work. High message count clustered into mid-length sessions. Your AI partner is mostly answering "why is this broken."
Triggers when: ≥5,000 total messages with 15–30 messages per session.
Many sessions, each short, lots of source diversity. You spin up small experiments instead of one big bet.
Triggers when: ≥60 sessions with <15 mean messages each.
You talk more than you listen. User-to-assistant message ratio above 1:1 — your prompts do the heavy lifting; the model fills in the gaps.
Triggers when: user/assistant ratio > 1.05.
Ties break in this order: Architect → Debugger → Prompter-First → Tinkerer → Cowboy. (Long-form work scores highest because the data signals are strongest.)
Two ways. Both run on your machine, both take seconds.
# In the dashboard — open the Wrapped strip on first launch: $ npx -y crixin # As a static HTML report (shareable as a screenshot): $ npx -y crixin wrapped --year 2026 --anonymize → writes ./crixin-wrapped-2026.html with your archetype + data project paths anonymized as Project A/B/C unless you drop --anonymize.
The heuristic is v0.2 — calibrated against the author's own corpus and a small private beta. v0.4 will tune the thresholds against real-customer data. If your label feels wrong, send us a screenshot of the Wrapped report (PII-stripped) and we'll fold it in.
Free for the dashboard. $5/mo unlocks the pro packs and an unlimited MCP server. 3-day trial, cancel anytime.