One AI Question, Four Screens
"How deep is your AI experience?" is one question covering four separate screens: AI interaction judgment, code-assisted prototyping, production-code authorship, and evaluation infrastructure. You're strong on the first, you can close the second, and the last two are fit boundaries rather than preparation problems. This dossier maps which construct each target company is screening for, what evidence each one requires, and the diagnostic question that makes the interviewer show you which screen you're on before you answer the wrong one.
One AI Question, Four Screens
"How deep is your AI experience?" is one question covering four separate screens: AI interaction judgment, code-assisted prototyping, production-code authorship, and evaluation infrastructure. You're strong on the first, you can close the second, and the last two are fit boundaries rather than preparation problems. This dossier maps which construct each target company is screening for, what evidence each one requires, and the diagnostic question that makes the interviewer show you which screen you're on before you answer the wrong one.
Reading What an AI-Design Role Is Actually Screening For
"Agentic" is currently doing duty in postings that screen for at least three different capabilities. The word tells you nothing about the filter, and if you misread the filter you lead with the wrong proof. This maps three observable signals — posting language, team structure, company profile — to the five AI-design constructs actually in play right now, and points each one at the experience pillar and portfolio proof to lead with. Read it once for the patterns. Scan it before you send anything.
Reading What an AI-Design Role Is Actually Screening For
"Agentic" is currently doing duty in postings that screen for at least three different capabilities. The word tells you nothing about the filter, and if you misread the filter you lead with the wrong proof. This maps three observable signals — posting language, team structure, company profile — to the five AI-design constructs actually in play right now, and points each one at the experience pillar and portfolio proof to lead with. Read it once for the patterns. Scan it before you send anything.
Interaction Judgment Is Your Strongest Provable Construct
You have a published framework, three live applications, and production context from enterprise agent governance. Together they prove you can design the boundary between what an AI system prepares and what a human authorizes. Five roles on your current list screen for this as a first-order filter. This dossier establishes what each evidence pillar entitles you to say, where that entitlement stops, and which artifact to lead with at which company. It also names the gap you can't close yet and what to say when someone asks about it.
Three Live Apps and No Build Record
Three working applications are live under your name. An evaluator can use them right now. But Linear, Vercel, and Midjourney are screening for how you build with AI coding tools, not just what you shipped — and the apps prove output while the build process stays invisible. This is an evidence gap, and no verbal bridge closes it permanently. What follows: the bridge that carries initial conversations, and the single build that resolves it.
Production-Code History Either Exists or It Doesn't
You have not authored production code professionally. That is not a framing problem or an evidence gap — it is a qualification boundary, a new category distinct from the four objection types covered in earlier issues, where no artifact or reframe creates the missing credential. The useful work here: a diagnostic question you can ask in outreach or screening calls to surface whether a role actually requires code ownership, exact language for exiting cleanly when it does, and your current pipeline sorted by which response applies.
Which Version of "AI Experience" Are They Actually Asking About
"Tell me about your AI experience" is four different screens sharing seven words, and committing to the wrong one wastes your best material on a construct the interviewer was not testing. This piece is a real-time diagnostic for reading which construct is actually in play before you respond — based on who is asking, what words they choose, and what their follow-up reveals. For each construct: a clarifying move that simultaneously diagnoses and demonstrates competence, plus which of your four evidence layers to reach for once you know.