When the résumé says agentic systems and you need to know if it’s true.
A scored, evidence-based read on whether a candidate or incumbent engineer is actually operating in the AI-native register the role requires — versus rehearsing the vocabulary. Six dimensions. One of four verdicts: Real, Rehearsed, Misleveled, or Unverified.
Want to feel the methodology yourself? Take the Engineering Signal Scorecard — 21 questions, ~7 minutes, scored against the same six dimensions.
Why the verdict is the verdict.
Assessment is priced independently so the verdict stays independent. If the evidence says Real, we say Real. If it says Rehearsed, we say Rehearsed. If it says Misleveled — meaning the work is real but the level is wrong — we say the correct level. If it says Unverified — meaning the claims could not be confirmed within scope — we say that. There is no downstream search fee for us. That’s the test.
Assessment is not a coding screen, a take-home, or a reference check at scale. It is an evidence standard applied to a specific engineer against the version of their role the next 18–36 months will demand. No shipping evidence = no score.
A scored report across the six dimensions of the Engineering Signal Lens, with cited shipping evidence per dimension, a synthesized assessment, and one of four verdicts: Real, Rehearsed, Misleveled, or Unverified. Delivered to the hiring sponsor first. Rollout sequence chosen by the sponsor.
The verdict
Every assessed finalist receives one of four verdicts.
VERDICT·REAL
The work is theirs, at the level claimed.
VERDICT·REHEARSED
Fluent vocabulary, thin production evidence.
VERDICT·MISLEVELED
Real work, wrong level — under- or over-leveled.
VERDICT·UNVERIFIED
Claims could not be confirmed within scope.
Fixed fee. Fixed timeline. A dossier you can put in front of a board.
Six dimensions. Each backed by shipping evidence.
- 01
Shipping Evidence
What they have actually built in production, against what users, at what scale.
- 02
Production Discipline
How they reason about latency, cost, eval, and failure under load.
- 03
Failure Decomposition
How they explain what broke, what they tried, what they learned.
- 04
Learning Velocity
How fast they absorb new model capabilities and integrate them into shipped systems.
- 05
Stack Reality
What they actually use vs. what they describe — the gap between vocabulary and tooling.
- 06
Cross-Functional Translation
How they ship with research, PM, and infrastructure as collaborators rather than blockers.
Two to three weeks per candidate or cohort. Structured technical interviews focused on what has actually been shipped, against what production constraints. Artifact review — code, architectures, postmortems, eval suites. Cross-functional signal from PM, research, infra. Final report delivered to the hiring sponsor. The candidate is informed of the engagement in a sequence the sponsor approves before kickoff.
Fixed fee. Fixed timeline. Confirmed in writing before kickoff. Paid independently of any downstream Search engagement.
- A finalist who looks right on paper but doesn’t feel right after the coding screen.
- An incumbent engineer whose role is being rewritten by what their team now ships agentically.
- A founding-engineer hire where the cost of a wrong call sets the team’s direction for a year.
- A succession decision on a Distinguished Engineer or AI Platform Lead.
- A PE portco where the engineering org’s AI-native posture is part of the value-creation thesis.
Not a coding screen. Not a take-home. Not a personality assessment. Not a reference check at scale. Assessment tests whether a specific engineer can operate in a specific AI-native role under a specific set of production conditions.
Request an Assessment.
Confidentiality is the default. Sensitive engineering hiring questions, replacements, and finalist assessments are handled directly by Jason and shared only with agreed stakeholders.