Talking to an engineer
The live call. A claim you cannot verify technically can still be evaluated on its shape, and the candidate's own questions carry as much information as their answers.
Lessons
- How do you hold a technical conversation with an engineer? The one guide to the call: what you can evaluate without technical judgment, how to test a claim by its shape, what unanchored numbers hide, and what a candidate's questions reveal.
- What can you ask a developer about their work without a technical background? The release pipeline as a question generator. Everything in it can be discussed with no technical judgment required to evaluate the answer.
- What does "I architected the system" mean, and how do you test the claim? The shape of the answer carries the information rather than its correctness. The tell is texture, and by now you have enough vocabulary to hear texture.
- Is "reduced latency by 40%" impressive? Every impressive number needs a baseline, a scale, and a consequence. Reusable on every metric on every resume forever.
- What does it mean when a candidate asks how often you deploy? Candidate questions as health checks on the pipeline they would live in, and what the honest answers reveal about the client.
- What does "the codebase is a mess" tell you about a candidate? The statements that sound like flags and mostly are not, and the ones that are real fit data.
- What is the difference between a big tech engineer and a startup engineer? Two different sets of scars. Neither disqualifies, and each raises a specific concern the other answers.
- How do you tell real AI experience from a resume keyword? The builds-and-uses axis run through the probing method. Every candidate claims GenAI experience now.