Hire the circle

Hire the Circle

Values into scorecards, scorecards into interviews, and the circle's rings staffed across humans and agents — with the two non-delegables in writing.

Gate

None — team design is legitimate at any stage. But the output lands in GROWTH.md ## Team, so create the file from the template if absent.

Inputs

• GROWTH.md (what work the rings currently demand), current team and agent inventory, and the role being hired or the lane being delegated.

The traits (score evidence, not vibes)

1. Horsepower — can actually do the analysis, not just commission it.
2. Purpose over pressure — won't buckle into short-term optimization.
3. Competitive, idea-brutal, person-respectful — attacks the claim, never the colleague.
4. Quality bar near perfectionism — a win of 10 basis points asks for 20.
5. Courage to reset — can take something working and say "not here, start again."
6. Integrity under score pressure — reports the failed experiment plainly.
7. Hires better than themselves — and isn't threatened when it works.
8. Value over perception — builds the real thing rather than the narrative of the thing. The trait the whole system stands on.

Steps

1. Write the scorecard. Per trait: what evidence looks like in this candidate's history (a busted-belief story with the data; a metric they retired; a hire who outgrew them). One trait = one story minimum.

2. Build the interview kit. Questions that elicit stories, with listen-fors and red flags. Red flags worth naming: metric talk with no user in the sentence; wins narrated without luck acknowledged; contempt for "boring" experiments; K-factor enthusiasm before retention exists.

3. Map the rings. For each ring, assign: owner, operators, and boundary. Agents excel at the monotony — instrumentation, lore queries, loop bookkeeping, ledger hygiene, probe suites. Humans hold judgment calls: keystone succession, context-reset verdicts, publication.

4. Write the two non-delegables. The alienation test and consent decisions never go to agents, regardless of capability. Every autonomous lane gets: a ledger it writes to, attribution on every entry, and a named human veto. Ambiguity here is where spam machines are born.

5. Record. Scorecard, kit, and ring map to GROWTH.md ## Team, dated. Revisit the map when the team or the model generation changes.

Conventions

• Culture-capability conflicts get flagged explicitly: a brilliant candidate who optimizes perception fails the card, and the card wins.
• Score agents like hires: an agent lane earns autonomy by ledger history (clean pre-registrations, honest kills), not by model name.
• The growth owner must be senior enough to engage core product strategy — growth abstracted away from product value becomes channel management.

Edge Cases

• Solo founder, no hires planned. The ring map is still the deliverable: founder + agents is a team, and the non-delegables still need writing down.
• A failing growth hire. Re-run the scorecard against observed behavior; the gap names the problem (often trait 8, hired on trait 1).
• An agent lane wants more autonomy. Grant it ring by ring against its ledger history; never grant the non-delegables, which is what "earned autonomy" tempts you toward.