A free, open-source, AI-assisted deep personality assessment.
Built as an alternative to closed-source tools that charge for what is, underneath, an aggregation of public-domain psychological instruments (Big Five / IPIP, ECR attachment, SWLS life satisfaction, plus a "zone of genius" preference layer).
This repo runs the same kind of assessment in ~20 minutes instead of 45+, for free, with the AI doing most of the work — reading your own files, scoring on your behalf, flagging where it's uncertain, asking you targeted follow-ups, and accepting your critique before producing a final profile.
- You drop evidence into
./me/— your repos, writing samples, calendar, LinkedIn posts, journal entries, whatever you have. - An AI agent reads it all and answers the 24 assessment items on your behalf.
- Each answer is labeled by confidence — high / medium / low.
- The agent then runs a follow-up Q&A on the low-confidence items, asking you targeted probing questions.
- You critique the synthesis. The agent re-scores based on your pushback.
- You get a final profile — Big Five trait scores, attachment style, zone-of-genius archetype, and a written synthesis with caveats.
The 24 questions are derived from public-domain instruments. The novelty is the AI-assisted loop, not the items themselves. See METHODOLOGY.md.
# 1. Populate ./me/ with your evidence (see me-template/ for the structure)
cp -r me-template me
# 2. Run the assessor (uses Claude via Anthropic API)
export ANTHROPIC_API_KEY=sk-ant-...
python cli/run.py --evidence-dir ./meThe CLI will:
- Read everything in
./me/ - Score all 24 items with confidence labels
- Prompt you with follow-up questions on low-confidence items
- Accept your critique of the synthesis
- Write the final report to
./me/report.md
n=1 so far. One full walkthrough is in examples/john-peslar-walkthrough.md — initial scores, follow-up round, user critique, final re-scoring, and where the model was wrong (it under-weighted broadcast extraversion for a public-facing solo founder).
The honest claim: the assessment matches self-perception on the dimensions that matter and surfaces the same archetypes a $100 paid tool would, given the same person. It is not a clinical instrument. It will not diagnose anything. It is a self-knowledge exercise with an AI co-pilot.
MIT. Use it, fork it, sell it, give it away. Just don't claim it's a clinical diagnostic — it isn't.
Based on public-domain psychometric instruments:
- Big Five — IPIP (International Personality Item Pool, public domain)
- Attachment — items adapted from Hazan & Shaver (1987) and ECR-derived public items
- Life Satisfaction — Diener et al. SWLS (1985)
- Zone of Genius — preference items inspired by Shaan Puri / Andrew Wilkinson framing of solo-vs-team operator wiring
The closed-source tool that inspired this (deeppersonality.app by Andrew Wilkinson) aggregates ~28 validated assessments behind a paywall. This repo strips the assessment down to its highest-signal items and lets an AI agent do most of the work, in the open.