FIELD NOTE / X
Typing speed is a weaker hiring signal now.
The complete written thought and the evidence behind it. The video edition will follow its public release.
The written argument is here.
This approved X edition is on the journal now. Its video player and original platform link will appear after each public release is verified.
Typing speed is a weaker hiring signal now.
Video caption
Typing speed is a weaker hiring signal now. Judgment still needs a system model. Use role-specific anchors and technical checks. #EricFieldNotes
Full written post / accessibility read
If a coding agent drafts the patch, the interview should reveal what the engineer does with it. A green producer test does not answer whether a lagging consumer reads the new event shape. The job includes finding that boundary and deciding whether a release is defensible.
The candidate should trace producer, queue, consumer, replay behavior and rollout order. Ask them to explain what old messages do after deployment and what would trigger rollback. A persuasive process speech without data-flow knowledge does not pass. Neither does elegant code that silently changes a business contract.
Show the same one-page generated diff and green test summary to every candidate. Ask: what would you need before release? Then disclose that one consumer cannot upgrade yet. Listen for a changed compatibility plan, an independently defined replay check and a named owner for the window. Nobody needs to build a migration to answer.
Score architecture, falsifying-test choice and communication against prewritten anchors, with technical fundamentals checked directly. Do this because the useful signal is how the engineer reasons about the generated patch, not how much free build work they complete. Pilot the questions against this role's real decisions before treating the rubric as predictive.
#EricFieldNotes
Four-beat scene transcript
1. Typing speed is a weaker hiring signal now.
If a coding agent drafts the patch, the interview should reveal what the engineer does with it. A green producer test does not answer whether a lagging consumer reads the new event shape. The job includes finding that boundary and deciding whether a release is defensible.
Visual: The agent can produce the first diff.
2. Keep fundamentals in the score.
The candidate should trace producer, queue, consumer, replay behavior and rollout order. Ask them to explain what old messages do after deployment and what would trigger rollback. A persuasive process speech without data-flow knowledge does not pass. Neither does elegant code that silently changes a business contract.
Visual: Judgment still needs a system model.
3. Ask one open question, then add one fact.
Show the same one-page generated diff and green test summary to every candidate. Ask: what would you need before release? Then disclose that one consumer cannot upgrade yet. Listen for a changed compatibility plan, an independently defined replay check and a named owner for the window. Nobody needs to build a migration to answer.
Visual: The interviewer brings the artifact; the candidate reasons aloud.
4. Hire for the decisions around the diff.
Score architecture, falsifying-test choice and communication against prewritten anchors, with technical fundamentals checked directly. Do this because the useful signal is how the engineer reasons about the generated patch, not how much free build work they complete. Pilot the questions against this role's real decisions before treating the rubric as predictive.
Visual: Use role-specific anchors and technical checks.
Research and claim limits
- U.S. OPM: Work Samples and Simulations (S151)
- METR early-2025 experienced-developer RCT (S153)
- U.S. OPM structured interviews (S203)
- U.S. OPM structured interview guide (S204)
The examples identified as illustrative or simulated are design probes, not reported incidents. Vendor specifications do not establish workload performance.