Onboarding: The First Ten Days

Onboarding is where the hiring bar becomes the performance bar. The pillars that gated the interview — ownership, reversibility, via negativa, AI in the workflow, the barbell — are the pillars the new hire is now measured on, starting immediately. Done well, the first ten days both prove the hire and audit the team.

1. Ship outcomes, not elegance. Ruthless 80/20. In week one, land one flagged production change that moves a real metric.
2. Barbell risk, be antifragile by design. Many small reversible bets plus a few big ones; write the blast radius and rollback first.
3. Run a fresh-eyes audit. Your first ten days are an audit on us. Keep a blindspot log of at least seven items and eliminate one end-to-end.
4. Seek discomfort; talk to humans. Two non-engineering one-on-ones in week one; ship one improvement sourced from them.
5. Publish and challenge. A day-five memo with a few concrete start/stop calls and one bigger bet with kill criteria.
6. Ask forgiveness within guardrails. Use a green/yellow/red taxonomy for changes so autonomy has clear edges.

A handful of small rules make the principles concrete: the ninety-minute rule (stuck for ninety minutes means pair, escalate, or pivot — never silently grind); write the rollback before the change; subtract before you add; flag everything loudly; keep to a two-metric diet so you're not drowning in dashboards; prefer small PRs; and remember that done beats perfect. These are the same behaviors the review will later score, which is the point.

Day 0 — access, and read the map: service objectives, the last couple of postmortems, the top open tickets, and the product used as a real user would use it. Day 1 — pick one measurable, user-visible outcome to own, and start three living documents: a blindspot log, a risk ledger, and an impact map. Day 2 — ship two small, reversible fixes behind flags. Day 3 — hold cross-functional one-on-ones, add the one metric you found missing, and write a pre-mortem for your chosen outcome. Day 4 — ship v0 behind a flag and run a tiny chaos drill in staging. Day 5 — publish the top-of-mind memo and demo what you've done. Days 6–10 — iterate on real data, eliminate one blindspot end-to-end, and draft a longer-horizon barbell plan.

A crucial distinction that protects both the hire and the manager: a new person deserves explicit runway. Name an "unlimited ramp" period on a scoped problem before any performance judgment, and consciously separate "still onboarding" from "not going to make it." Most early wobble is ramp. Reading ramp as underperformance burns good hires; reading persistent underperformance as ramp lets a mismatch fester. The blindspot log and day-five memo give you early, concrete signal to tell the two apart.

By the end of ten days you have observed the new hire exercising ownership, reversibility, subtraction, AI-assisted work, and the barbell — on real work, not in an interview. That is the same evidence a performance review runs on. When onboarding, review, and hiring all point at one list of pillars, a manager is never starting from a blank page: the running record began on day one, and "how is this person doing?" always maps back to the same yardstick. That continuity is the whole reason to treat hiring and firing as one system — which is where the second half begins.