The short version
Two of your current targets will screen you out on code ability. Five more will raise it as a question. The rest will never ask. Sort them before you spend time on any of them.
What they're asking and what they mean
"Can you prototype in code?"
The question is about coding. The concern is about velocity: will this hire hand off static mockups and wait for engineering to make them real? Companies that ask this have been through that bottleneck. They want a designer who can make something run in a browser without filing a ticket.
Your portfolio at junochen.com contains no code-built prototypes. Your GitHub has zero public repositories. That is an evidence gap, and no framing eliminates it. What you can do is redirect the conversation at the subset of companies where the actual concern is speed, not syntax.
Which companies trigger this screen
I pulled every role in your active target list that mentions code prototyping, AI-assisted code tools, or shipped-code experience. Seven companies. They sort into three tiers based on how the screen functions in their process.
Hard filter — the interview tests the capability directly:
- Ashby (Design Engineer): Pair programming in the interview. TypeScript, React, CSS as stated minimums. Expects shipped code to hundreds of users. This is a design-engineering hybrid, not a design role. A verbal bridge cannot substitute for a live coding exercise.
- Luma AI (Product Designer): Requires functional prototypes built with Cursor or Claude Code, visible in the portfolio before the interview. Explicitly excludes managers.
Strong preference — addressable in conversation if you confirm the screening behavior first:
- Headway (Senior/Staff Product Designer, Provider): "Prototyping in code is core to the role." Names Claude Code and Cursor. But the experience requirements don't list coding as a qualification, and no public source confirms whether non-coding applicants are screened out before interview. Clarify with the recruiter before writing outreach.
- Revin (Senior Product Designer): First in-house designer. Expects functional prototypes with Claude Code or Cursor against a Next.js/React product. But the posting explicitly allows someone who doesn't write production code by hand.
- Mariana Minerals (Staff Product Designer): Expects code prototyping and live pairing with engineers. States production code is "a plus, not a requirement."
Technical fluency, not code output:
- Checkout.com (Staff Product Designer): Requires shipped experience with AI-assisted design or development workflows. Explicitly says this is not a front-end engineering role. Your Agentic Labs artifacts are partially relevant here.
- Docebo (Senior Product Designer, AI Experience): Expects high-fidelity AI-behavior prototypes before engineering commits. Interview includes a workflow review focused on AI tool use. Does not require production-code ownership.
What separates these roles from your Director+ targets is whether the posting describes a maker who builds or a leader who directs. Docebo has 900+ employees. Checkout.com's design org runs 30+ people. The filter is role type, not company size.
Most of your active pipeline — Anthropic, Stripe, Salesforce, roles at that scale — evaluates strategic judgment, org design, and craft leadership. Those interviewers will not ask you to open a code editor.
The verbal bridge
For Headway, Revin, Mariana Minerals, Checkout.com, and Docebo — roles where the posting leaves room for a non-coding designer who ships fast with engineers — you have a bridge grounded in delivery evidence.
Confidence: moderate. This addresses the velocity concern. It does not answer a work-sample test.
"I don't have a public code-prototyping portfolio today. What I do have is a delivery record in small, integrated teams where design and engineering shared sprint cadence and platform constraints. Red Cross at BCG Digital Ventures went from zero to national deployment in six months — one designer, two engineers, building on Salesforce SLDS. Equinox+ at BCG DV went from zero to MVP in three months with a shared token architecture. At TinyFish, I shipped AI-native enterprise tools in production. The question those timelines answer is whether I create bottlenecks for engineering. I don't."
The specifics matter if the interviewer probes. At Red Cross, designing on Salesforce SLDS meant working within the platform's component constraints — layout, interaction patterns, data models were shared territory between design and engineering, not artifacts handed over a wall. At Equinox+, the token architecture meant design decisions were implementation decisions; changing a spacing token changed the build. TinyFish adds the most recent layer: production-level AI tooling work where the product itself was technical infrastructure. Use it in past tense as the guideline permits, and let the named company carry the weight.
From Issue #7's objection discipline: the smallest claim that answers this is "I ship fast in engineering-integrated workflows, and here's the evidence" — not "I can code."
Use with caution when the interviewer asks to see a repository, requests code samples, or the process includes a live coding exercise. At Ashby and Luma, do not attempt this bridge.
The Agentic Labs as partial evidence
Your three Agentic Labs applications — Brand Pulse, Retail Velocity, Carrier IQ — are running systems you built solo. Multi-agent searches, ranking, extraction, verification, review workflows. They run and expose system behavior.
But no source code, build history, or documented development process is publicly visible. An evaluator can see that the systems work but not how you built them.
The systems also carry a velocity signal: you built them solo, and they run. That doesn't prove coding proficiency, but it shows you can move from concept to functioning product without waiting for engineering support.
At Checkout.com and Docebo, where the requirement is AI-assisted workflow fluency, the Labs are useful. Walk through system behavior live and describe your build decisions — what you configured, what you prompted, what you rejected, what you iterated. The live walkthrough compensates partially for the absent repository. Issue #7's AI-credibility dossier already established the boundary: the Labs demonstrate AI decision-layer credibility, not production-code participation. Stay on that side of the line.
Build Queue recommendation
Item: Carrier IQ Correction Lineage — Coded Work Sample
If Headway, Revin, Luma, or Mariana Minerals remain active targets, the smallest artifact that closes this gap is a functional prototype with an inspectable repository. Build the already-prioritized Carrier IQ correction-lineage flow in React. This avoids creating an unrelated demo project and exposes both code-prototyping practice and the correction-to-next-run sequence currently missing from the live app.
The repository needs a README that distinguishes AI-generated code from your behavioral, structural, and quality decisions. This doesn't position you as a front-end engineer. It shows you can use code tools to make a design idea run, and that you made deliberate implementation choices along the way.
Confidence: moderate, conditional on target mix. Count how many roles in your current pipeline fall into the technical-builder category. If these seven companies represent a meaningful share, build it. If your pipeline is overwhelmingly Director+ leadership roles, the investment doesn't justify the narrow applicability.
Quick-Reference Card
When you hear: "Can you prototype in code?" / "Do you use Cursor or Claude Code?" / "How technical are you as a designer?"
What they mean: Will you create a bottleneck for engineering?
Your gap: No public code prototypes, no visible repositories. Real gap. Don't pretend otherwise.
The bridge (moderate confidence): "Red Cross at BCG Digital Ventures to national deployment in six months, one designer, two engineers. Equinox+ at BCG DV to MVP in three months on a shared token architecture. At TinyFish, shipped AI-native enterprise tools in production. I don't create engineering bottlenecks."
Use at: Headway, Revin, Mariana Minerals, Checkout.com, Docebo.
Skip: Ashby Design Engineer, Luma AI. The interview directly tests coding ability. The bridge cannot substitute.
Show the Labs at: Checkout.com, Docebo. Walk through system behavior live, describe build decisions. Don't claim they prove coding proficiency.
Recruiter clarification for ambiguous roles:
"Does the interview process include a coded prototype review or a technical coding exercise, or is the code-prototyping expectation evaluated through portfolio discussion and workflow questions?"
- Ashby's role separation: The Design Engineer posting is a distinct department from Ashby's ordinary Product Designer roles, so the hard-filter conclusion should not be generalized if Ashby posts a non-engineering design seat.
- Retail Velocity's TinyFish attribution: The live app visibly references TinyFish technology, which carries an unresolved permission concern under the publication boundary — worth settling before using it in any live walkthrough with an evaluator.
- OpenAI's player-coach model: Their Growth design leadership posting explicitly accepts senior ICs who demonstrate mentorship and team impact, confirming that the code-prototyping screen and the IC+manager screen can overlap at frontier companies even when coding itself isn't required.
- NIST on deployed-AI monitoring gaps: Their March 2026 report flags unresolved questions around human-AI feedback loops, drift, and oversight burden, which strengthens the case for the Carrier IQ correction-lineage artifact as evidence of exactly the temporal design judgment these frameworks describe but few portfolios demonstrate.