1. The Objection
"You built these apps with AI tools — how do I know what's your design judgment versus what Claude produced?"
Surfaces most reliably at AI-native companies (Anthropic, OpenAI) during portfolio review, and at growth-stage companies with design-invested leadership (Headway, Ramp) in final rounds. Screening risk is likely low; the apps clear initial filters on their own. It comes up mid-process, after an evaluator opens Carrier IQ and registers how much architecture is sitting in it.
2. What They're Actually Asking
Nobody on your target list cares that you used AI tools. Headway's current posting explicitly asks designers to prototype with Claude Code or Cursor. The concern is that your design judgment is not independently inspectable. An evaluator can see that someone made deliberate architectural choices in Carrier IQ — four differentiated review states, each implying a different operator workflow and evidence threshold. What they cannot see is which alternatives were on the table, what got tried and thrown out, why four states rather than three or six. Without that trail, there is no way to separate "she designed a review-state taxonomy from risk differentiation" from "the model proposed four states and she kept them." That separation is what they're hiring on.
3. The Strongest Case Against You
- Your homepage claims you "built the agents to prove it works" without saying which decisions were yours (architecture, interaction model, information hierarchy) and which were generated and accepted as-is.
- None of the three Labs apps (Brand Pulse, Retail Velocity, Carrier IQ) exposes a version history, a rejected alternative, or a rationale annotation. There is no public record separating your choices from generated output.
- Anthropic's candidate AI guidance preserves independent-reasoning assessment: take-home work must be completed without Claude unless expressly permitted, and live interviews run without AI assistance unless stated otherwise. A portfolio that can't demonstrate independent judgment in advance pushes all of that evaluative weight onto the live round.
- An earlier version of your site used the phrase "built alone with Claude Code." It's gone from the current version, but cached renders and search indexes may still serve it. An evaluator who hits the old phrasing, goes looking for a repository or a decision trail, and finds neither will read the gap as a credibility problem.
4. The Grain of Truth
There is no public record — on junochen.com, in the live apps, anywhere an evaluator can reach independently — of what you decided, what you rejected, or why. The judgment is in the apps. Carrier IQ's differentiated review states, Brand Pulse's agent-trace visibility model, and Retail Velocity's declared infrastructure dependency are design decisions, not defaults. But the evaluator has to take your word for that, and at this level taking your word is not a competitive position.
5. Classification
Evidence gap. The portfolio has the products and not the decision trail. The correction-lineage artifact from Issue #8 covers one dimension of this (failed output, diagnosis, intervention, result). The broader need is a design-decision record for at least one Lab: what was proposed, what you changed, why, and what the change produced.
- Build Queue item 1: design-decision trail artifact for Carrier IQ (status: not started). Carrier IQ is the strongest candidate because its four review states are the most visible architectural judgment in the set. The trail would show why those states exist and what alternatives were evaluated.
- Build Queue item 2: forward-looking visual design artifacts across four domains (status: in progress). These come at the authorship question from a different direction. Original design thinking about problems nobody has solved yet — inference-aware UX, intent-based interaction, human-agent system design — is harder to attribute to a model than an implemented product is, because there's no established pattern to generate from. Each one that publishes to junochen.com moves the evaluator from inspecting what you built toward engaging with how you reason. Check the Design Context File for current artifact inventory and status.
6. The Verbal Bridge
Until the decision-trail artifact ships, make the architecture legible in conversation. Carrier IQ is the lead example.
Walk the four review states as decisions rather than features. Bindable: evidence cleared every threshold, operator can act without further review. Normalize: coverage data needs standardization before comparison. Referral: risk profile requires a specialist. Call review: automated evidence is insufficient and a human conversation is required. The states differ in operator workflow, in evidence threshold, and in how much trust the system extends before a person intervenes. That taxonomy came out of how insurance underwriting actually works: match review intensity to risk signal. (Confirm the state names and the workflow logic against your actual implementation before you rehearse any of this. The architectural reasoning holds either way; the labels need to be yours.)
Pair it with the Trust essay's five handoffs framework. The framework is a design-reasoning artifact in its own right — it documents how you think about the boundary between human authority and agent capability. An evaluator who reads it and then sees it reflected in Carrier IQ's review states can follow the reasoning without a formal decision log. Be precise about sequence. If the framework codified thinking you developed while building the apps, say that. It's equally strong evidence, arguably stronger, because it means you recognized the pattern in your own work and formalized it. A causal claim that doesn't match the timeline is exactly the detail a sharp evaluator catches.
Confidence: moderate. This bridge works with evaluators who engage deeply enough to see the architectural logic. It does nothing for evaluators who screen on inspectable artifacts before the conversation starts. The decision-trail artifact is the only fix for that second group.
Tier Notes:
- AI-native (Anthropic, OpenAI): Highest risk. These companies assess independent reasoning explicitly. Lead with the Trust essay as a reasoning artifact, then walk Carrier IQ's review states as implementation of that reasoning. The forward-looking artifacts carry particular weight here, because original thinking about problems these companies are actively working on is the strongest available proof that the judgment is yours.
- Growth-stage (Headway, Ramp): Moderate risk. Headway's posting asks for design rationale and multi-sprint experiment arcs. Prepare specific iteration examples you can narrate verbally.
- Enterprise (Salesforce, Atlassian): Lower risk. Portfolio review focuses on scale and org impact. The authorship question is less likely to surface at all.
7. What to Say
"I designed Carrier IQ's review-state architecture — four states based on how insurance risk signals should route operator attention. That's design judgment grounded in the Trust framework I published. I'm building a detailed decision trail for the portfolio, but I can walk you through the reasoning behind any architectural choice in these apps right now."
8. What Not to Say
- "I built everything myself." This restates the inspectability concern as a defense. The question is whether your judgment is visible in what was built, not who typed it.
- "AI was just a tool." Dismissive of a legitimate concern, and it contradicts the seriousness your own Trust essay brings to human-agent attribution. An evaluator will notice the inconsistency.
- "I can show you my prompt history." Prompt history shows what you asked for, not what you evaluated, rejected, or redesigned. It also invites a forensic review you don't want.
9. Early Signals
- The evaluator asks you to "walk through your process" on a specific app rather than asking about the outcome. Process questions at this level are authorship probes. They want to hear a decision you made, not a result you shipped.
- The interview includes a live design exercise with AI tools explicitly prohibited. Both Anthropic and OpenAI reserve the right to assess independent reasoning without AI assistance. If you're told the exercise is AI-free, the concern is already active.
Objection: "How do I know what's your design judgment versus what the AI produced?"
Classification: Evidence gap
Real fear: Design judgment is not independently inspectable — no decision trail separates your choices from AI-generated output.
Lead with: Carrier IQ's four review states as deliberate design decisions grounded in the Trust essay's framework.
Say: "I designed Carrier IQ's review-state architecture based on how insurance risk signals should route operator attention. I can walk you through the reasoning behind any architectural choice in these apps."
Avoid: "I built everything myself" — restates the inspectability concern as a defense.
Build Queue items: (1) Design-decision trail artifact for Carrier IQ — rejected alternatives, design rationale, iteration history. Status: not started. (2) Forward-looking visual design artifacts across four domains — original design thinking on unsolved frontier problems moves the evaluator from inspecting past builds to engaging with current reasoning. Status: in progress.
- Cached site still diverges: The direct homepage no longer says "built alone with Claude Code," but the older cached rendering may still serve that language through search indexes and social previews — worth checking what Google, LinkedIn, and recruiter databases currently display.
- Headway's interview structure unpublished: The Senior/Staff Provider posting demands design rationale, multi-sprint experiment arcs, and code prototyping, but no public source confirms which round those probes appear in or how the portfolio review is structured.
- Atlassian splits the authorship question: Their design interview handbook separates IC portfolio reviews (probing design choices) from management reviews (probing how you led the team and shaped the result) — useful for anticipating which version of the authorship question you'll face.
- CHI 2026 hiring-creativity tension: A recent study of 17 executive design managers found that managers valued candidates who challenged assumptions and reframed problems, while portfolios often failed to show either creative departure or competent system-bound production — relevant context for how decision-trail evidence might land differently than polished final output.