The Illusion of Read
In "Why Job Interviews Don't Work," Farnam Street explains that unstructured interviews let charisma and narrative smoothness substitute for actual signal. Founders are especially vulnerable: pattern-matching on founders they admire, they mistake fluency for competence. The research is blunt — a confident, likeable candidate produces almost the same predictive accuracy as a coin flip. The feeling of certainty after a great conversation is noise wearing a suit.
Work Samples Are the Eval
The antidote the essay prescribes is structured assessment: standardized questions asked of every candidate, scored on consistent rubrics, and real work samples tied to the actual job. This maps directly to how reliable agent systems are built. You do not trust an agent because it sounds confident during a demo; you design a test harness, measure outputs, and compare runs. Hiring deserves the same architecture. Define the task, observe the work, score it against criteria you wrote before you met the person.
Base Rates Keep You Honest
Farnam Street's essay also stresses base rates: most candidates for most roles fail to sustain early performance, and your confidence interval should reflect that reality before the interview starts. For founders, this is the approval-layer logic applied to people. The stakes of a hire — high, irreversible, trust-dependent — warrant more rigor, not less. Humans own the final call on anything hard to undo. A structured process is not bureaucracy; it is the mechanism that makes that judgment defensible when you are explaining it later.
Design Trust In, Not By Feel
An AI-native company is designed from the start around the principle that trust is earned through measurable evidence and clear accountability, not through vibes. That standard cannot stop at the agent layer and dissolve when a human walks in the room. Apply structure to every consequential, hard-to-reverse decision — and a founding hire is about as hard to reverse as decisions get. Match the rigor to the stakes.
