Trigger
Product Design Lead, Growth-Codex posted August 18 — 17 days old. Confidence on trigger: high. Ian Silber's August 16 Lenny's Podcast interview dropped two days before the posting went live: composable primitives, hands-on prototyping, systems thinking, a developed AI-native point of view. That interview is the buyer's evaluation criteria stated publicly. Confidence that Silber is the hiring manager for this specific seat: moderate — he's the most visible design authority at OpenAI, but reporting line is undisclosed.
Window
You are in weeks two and three. Early-pipeline candidates are already screening. The prior two cards flagged the same pattern: research consuming the window that research was supposed to inform. Move this week.
What this means for the design org
The product surface shifted since the first card. In July, OpenAI launched ChatGPT Work for non-coding tasks — research, analysis, documents, presentations. The current quickstart routes users into three modes: Chat, Work, and Codex. But the documentation still tells users who previously used Codex for non-coding work that they can stay in Codex or move to Work. Axios called the resulting product lineup a "jumble."
That tells you what the Growth-Codex designer is actually solving. A first-time business user has to pick a product mode, decide on execution location (local vs. cloud), connect source context, grant permissions, and express a task outcome with a definition of done — all before the product does anything useful. The growth mandate is likely about collapsing that sequence so users reach value before the choices overwhelm them.
The Trust essay's first handoff — Intent-Setting — is this problem stated as a design principle.
Your angle — and the build-or-move decision
The Intent-Based Interaction artifact would be the strongest possible lead here. It would show, visually, how you'd design the layer between "I want to do something" and "the system routes me to the right mode with the right context." It would make the first handoff concrete and specific to a product shaped like Codex.
It doesn't exist yet — concept-only, no visual build progress. Five cycles of build queue specifications have produced zero published forward-looking artifacts, and waiting for this one means waiting indefinitely.
If you get to an interview, build a bounded Intent-Based Interaction sketch for the presentation — one interaction model showing how mode selection could collapse into an intent-expression layer. Build it for the interview, not for the portfolio site.
- Trust essay — the first handoff (Intent-Setting) names the design problem this role exists to solve. Open with it.
- Carrier IQ — the Agentic Labs app demonstrating all five handoffs in a regulated domain. Production proof the framework works outside a PDF.
- TinyFish cold-start — usable as practitioner context in outreach (role evidence, not portfolio proof) but not linkable. You can reference the blank-input-box redesign you shipped — it shows you've solved a version of this in production — but you can't point to a URL. Go deeper on the mechanics in conversation.
Competitive landscape
Three candidate archetypes are likely in this pipeline.
Internal transfers and AI-native designers (Google DeepMind, Meta AI, current OpenAI ICs moving laterally) will outperform you on brand pedigree and internal context. Your counter: none of them have published a framework that names the design problem this role exists to solve. The Trust essay is published and public; their design work is internal and invisible.
Enterprise design directors (Salesforce, Spotify, Figma alumni) bring org-scale credibility but will struggle to demonstrate AI-native product depth. You match their scale proof with Alibaba and exceed their AI credibility through TinyFish production experience and Agentic Labs.
Senior IC candidates from FAANG design orgs bring deep craft portfolios. They typically lack function-building proof and 0→1 speed — your BCG DV track record covers both.
Your primary vulnerability: the missing Intent-Based Interaction artifact. The interview-build strategy above is how you close that gap. (Confidence on competitive assessment: moderate — OpenAI's internal transfer pipeline is opaque.)
Who to contact
Primary: Ian Silber — Head of Product Design, OpenAI. LinkedIn.
Public fingerprint: The August 16 podcast is the sole current source. No LinkedIn posts, replies, or reshares between August 16 and September 4 are publicly indexed. Every informed candidate will reference the podcast. Your differentiator is specificity about the Codex/Work routing problem — showing you understand the product surface, not the podcast talking points.
Warm path: None surfaced. If you identify a connection (BCG DV alum, design community overlap, ADPList), use the warm message below. Otherwise, cold outreach.
Channel: Cold LinkedIn DM. If you find an email through your network, email is better — less noise. But LinkedIn is the available channel from public signal.
Secondary: Ed Bayes — design lead for Codex and ChatGPT Work. Identified in an April interview and a September conference announcement. His involvement in Growth-Codex hiring is unknown, but he's the closest design lead to the product surface. If Silber doesn't respond within five business days, Bayes is the second route.
Outreach messages
Warm message to Silber (assumes a mutual connection made the introduction):
[Name] suggested I reach out — I've been working on the design problem underneath Codex's growth challenge: how you help a user express what they want without requiring them to understand the product architecture between Chat, Work, and Codex.
I published a framework for this — the intent-setting handoff in "Trust Is the New Interface" at junochen.com — and shipped a version of it at TinyFish, where I redesigned the cold-start experience for an enterprise agent platform. Carrier IQ on my site demonstrates the full trust loop in production.
Would 20 minutes on how Growth-Codex defines activation today be useful?
Cold message to Silber:
The Codex/Work split created a mode-routing problem that's going to define growth design at OpenAI for the next year — how do you help a user express what they want without requiring them to understand the product architecture underneath?
I published a framework for exactly this problem — the intent-setting handoff in "Trust Is the New Interface" — and shipped a version of it at TinyFish, where I redesigned the cold-start experience for an enterprise agent platform so users could reach a working agent without prior context. Carrier IQ on my portfolio site (junochen.com) demonstrates the full trust loop in production.
I'd like 20 minutes on how Growth-Codex defines activation today and where intent-routing fits. Would that be useful?
What not to say:
- Don't describe the current Codex onboarding as a "blank box." The product has changed since earlier reporting. Getting the product state wrong signals you haven't looked at the current documentation.
- Don't quote the podcast back to him verbatim ("composable primitives," "systems thinking"). He knows what he said. Reference the ideas through your own framing.
Cover letter draft
The hardest growth design problem at OpenAI right now is intent routing. A business user opening the desktop app faces a mode decision (Chat, Work, or Codex), an execution-location decision, a source-context decision, and a task-framing decision before the product delivers any value. Collapsing that stack without removing user agency is the design problem underneath the Growth-Codex role.
I've been working on this from two directions. "Trust Is the New Interface," published at junochen.com, introduces a five-handoff framework for agentic systems — the first handoff, Intent-Setting, addresses how to help users express what they want without requiring them to understand the system's routing logic. Carrier IQ, one of three Agentic Labs apps live on my site, demonstrates all five handoffs in a regulated domain (InsurTech quote automation), including the intent layer that translates a user's situation into structured agent action. Most recently as Head of Product at TinyFish, an enterprise web agent platform (Series A), I shipped the agentic platform from 0 to 1 within three months — including redesigning the cold-start experience so first-time users could reach a working agent through a progressive intent-setting flow rather than an undocumented input field.
I moved to product to build AI-natively from zero and understand the full lifecycle from requirements to production deployment. I'm returning to design because I want to apply that technical depth — agent traces, auditability, governance — to a specific, high-stakes product surface where the design decisions shape whether millions of people trust an agent enough to delegate real work. Growth-Codex is that surface.
My portfolio at junochen.com includes the Trust essay, three live Agentic Labs apps, and case studies at Alibaba ($50B+ GMV platform redesign as Head of Design and Research) and Thermo Fisher ($20M+ margin, 0→1 in 12 months as Product Design Director at BCG Digital Ventures). I'd welcome the chance to discuss how intent-routing design could accelerate Codex adoption for business users.
Resume emphasis guidance
- Lead with Agentic Labs (Carrier IQ especially) and Trust essay. Direct-match evidence for the role's core problem.
- Follow with TinyFish as most recent role context: "Head of Product at TinyFish (enterprise web agent platform, Series A). Shipped agentic platform 0→1 in 3 months. Agent traces, auditability, governance in enterprise production."
- Third position for Alibaba. Enterprise scale credibility, but the role is growth-focused — don't lead with it.
- Trim or move down Red Cross and Equinox+. Neither maps to growth-at-AI-scale.
- Keep Allē accessible. The dual-surface model and 3.2× redemption rate are growth-relevant metrics if conversation turns to activation and retention proof.
The non-obvious thing
Ed Bayes leads design for both Codex and ChatGPT Work. The mode-routing problem — which product a user should be in — sits in his portfolio as a design leadership question, not a product-management decision. If you reach Bayes, the intent-routing angle applies to his entire mandate, not a single headcount. That gives the conversation scope beyond one role.
Warning signs
The Growth-Codex mandate may be narrower than the posting suggests. If the role owns conventional growth metrics (sign-up conversion, day-7 retention, A/B testing velocity) without latitude to redefine what activation means for an agent product, the intent-routing framing won't land — and the role becomes a growth-execution seat with limited scope expandability and craft depth. The prior card named the evidence gate: How does the growth team define activation for Codex today, and how much latitude does this role have to redefine it? Ask it in the first conversation. If the answer is "we have a defined funnel and need someone to optimize it," recalibrate your interest.
Post-outreach signals
AI-native startup rules. Response within 48–72 hours is healthy. Silence past five business days means the message didn't cut through or the pipeline is further along than posting age suggests. If Silber doesn't respond by day five, send one follow-up to Bayes with the mode-routing hook adjusted for his Codex + Work dual scope. If neither responds after two attempts each, close the loop and monitor for a posting refresh — a repost with changed language would reopen the window with new intelligence about what the first search taught them.
- Codex's knowledge-worker surge: Axios reported that knowledge workers represent roughly one-fifth of Codex users and are growing more than three times as fast as developers, which reframes the Growth-Codex mandate as a non-developer adoption problem more than a developer-retention one.
- Sara Culver's growth-design history: Culver's 2025 conference biography identified her as leading content design and growth design at OpenAI, and her hiring posts described the Growth team's A/B-testing cadence — if she's still in scope, she may be a closer evaluator than Silber for this seat.
- OpenAI's Identity role as signal: The separate Product Designer, Identity posting covers agent permissions, access, recovery, and human-to-agent authorization, confirming that agent authority is an explicit design surface across OpenAI beyond Growth-Codex.
- Homepage evidence conflict: A September 4 direct-server check found junochen.com presenting TinyFish as the first item under "Selected work" with no visible Brand Pulse, Retail Velocity, or Carrier IQ on the homepage — resolve the routing before any outreach that links to the site root.