Trigger
Banu Guler installed as Chief Design Officer July 24, following Midjourney's acquisition of Co-Star. Design Lead posting live on Ashby, $285K–$350K cash. Today is day 50 of Guler's tenure. Confidence: high — both events are first-party confirmed.
That piece said wait until the Intent-Based Interaction artifact shipped. The reasoning was defensible then. I'm reversing it. Day 50, a live posting, and a parallel Product Designer req mean the slate is forming now. Waiting for a perfect artifact inventory while the team fills around you is the failure mode this publication exists to prevent.
Window
2–3 weeks, narrowing. Guler is building her first Midjourney design team. Two design postings running simultaneously signal team formation, not a single backfill. Post-Labor Day restart means hiring managers who paused screens last week are reopening calendars now. The first-90-days window closes late October. You want to be in conversation before Guler's shortlist solidifies.
Rubric
| Dimension | Score | Note |
|---|---|---|
| AI centrality (bonus) | 3 | AI is the product |
| Stage & equity | 1 | Self-funded, profitable, ~60 people. No disclosed equity |
| Design influence ceiling | 3 | New CDO building the function; Design Lead reports to her |
| Trajectory | 3 | V8.2, Edit Model alpha, acquisition, multi-role hiring |
| Company total | 10 | |
| Total comp | 2 | $285K–$350K cash confirmed. Historical profit-sharing reported; equity unconfirmed |
| Scope expandability | 3 | First design leadership hire under a new CDO |
| Craft depth | 3 | Hands-on prototyping and review alongside team building |
| AI exposure quality | 3 | The Edit Model is the human-AI control surface |
| Portfolio value | 3 | Consumer AI at massive scale. Differentiated case study |
| Role total | 14 |
Verdict: Act. Company 10, Role 14. The equity gap is the only drag, and it's an unknown, not a confirmed absence.
What this trigger means for the design org
Midjourney has been engineer-and-founder-led since inception — their careers page says so explicitly. Guler's appointment is the first structural investment in design leadership. She brought roughly two dozen people from Co-Star, and at least two product designers were already at Midjourney, so this isn't a cold start. But the posting's language about "recruiting the product-design team" and "establishing operating rhythms" tells you the function doesn't have structure yet. Guler builds the org. The Design Lead builds the practice underneath her.
The V8.2 Edit Model — natural-language image editing, inpainting, outpainting, reference images, session-based edit history — went into alpha in late August. Midjourney explicitly warns the interface is broken in places and changing rapidly. This is the product surface the Design Lead will own, and the control problems are live and unsolved.
Your angle
Narrative fit: The Edit Model has a trust problem at its center. A user types "make the lighting warmer" and the model interprets that instruction without preview, explanation of what will change, or graduated control between precision and exploration. That maps directly to the Trust essay's Intent-Setting and Output Review handoffs — and the mapping is tighter here than in analytics or enterprise products because creative output is higher-variance. The trust calibration is harder when the user can't fully specify what they want.
Lead with Trust Is the New Interface. The five handoffs framework describes the interaction architecture the Edit Model needs: how intent gets communicated, what the user sees while the model works, how the user evaluates and redirects output, where the user decides to commit or revise, how the system learns from that cycle.
Support with Agentic Labs — Brand Pulse and Carrier IQ demonstrate live AI products where you designed the control surface between user intent and agent action. Secondary: Allē (Product Design Director at BCG Digital Ventures) — 30M members, 3.2× redemption, $42 CAC — proves consumer scale with craft depth.
Artifact timing note: If the Intent-Based Interaction artifact ships before your first conversation with Guler, it becomes the lead evidence — it directly addresses how users communicate intent when they can't fully specify what they want, which is the Edit Model's core problem. Accelerate that build in parallel with outreach, or send now and reference it verbally if it ships before the call.
Competitive landscape
Three archetypes are likely in this pipeline. Confidence on all three: moderate — Midjourney's design hiring is new enough that the candidate pool is forming alongside the search.
Creative-tool designers (Adobe, Figma, Canva, Procreate alumni). They know creator workflows and visual-tool interaction patterns. Your vulnerability: you haven't designed a creative tool. But the Edit Model's hardest problem is AI trust calibration, not traditional tool interaction — and the Trust essay names the problem their domain experience doesn't address.
Generative AI designers (Runway, Stability, DALL-E team alumni). Direct experience designing for image generation. They have domain pedigree in the exact product category, which you don't. Most, though, designed internally without publishing frameworks. The Trust essay is a public artifact Guler can evaluate before a conversation — published design thinking at this level is rare from candidates who've been heads-down inside a single product.
Taste-forward consumer designers (design-led consumer companies Guler would respect). Strong visual portfolios, cultural sensibility that matches Guler's Kawakubo-citing, art-and-literature orientation. Taste is demonstrated through the work, and your portfolio is enterprise-weighted — that's where you're exposed. Agentic Labs and Allē demonstrate consumer craft, and the intellectual depth of the Trust essay is a differentiator this archetype rarely brings.
Where you're most exposed: a candidate who has both creative-tool domain experience and strong taste will feel like a lower-risk hire to Guler. Make the trust-calibration problem feel like the primary design challenge — it is — and frame domain experience in creative tools as secondary to understanding how AI interpretation works.
Who to contact
Banu Guler, Chief Design Officer. LinkedIn. Holz publicly directed design and product inquiries to her.
Public fingerprint: Limited post-appointment. She described using Midjourney as "uncovering something subliminal and new about yourself." Deeper design philosophy from Co-Star interviews: draws from "art, literature, therapy" rather than other apps, cited Rei Kawakubo as an influence, and made a specific decision not to use traditional astrological charts in Co-Star's mobile app because the form didn't suit the medium — comparing it to asking "what a spoon can do that a scalpel cannot." This is a taste-first, medium-specific evaluator. She will read your portfolio before your resume.
Holz's role in evaluation: The careers page says projects have historically been led by engineers and the founder, and that portfolios matter more than resumes or pedigrees. Expect him in later rounds. His evaluation criteria likely differ from Guler's — he's been the de facto design authority for four years and will test whether you can operate at the pace and ambiguity of a founder-led product culture. Your TinyFish 0→1 speed (agentic platform shipped in three months) is the proof point for him. Your Trust essay is the proof point for her.
Warm path: Cannot assess without authenticated LinkedIn network view. Holz's public direction to contact Guler makes cold outreach legitimate.
Channel: LinkedIn cold DM to Guler.
Email subject line (if the conversation moves to email): The Edit Model's control-surface problem
Outreach messages
Cold message to Guler:
Your description of Midjourney as uncovering something subliminal about yourself is the sharpest framing I've seen for what makes creative-AI control different from other AI design problems — the user isn't trying to get a correct answer, they're trying to get closer to something they can't fully articulate yet. That's a trust calibration problem, and it's what I've been writing and building around for the past year.
I published a framework called "Trust Is the New Interface" that maps five handoffs between human intent and AI action — intent-setting, in-progress visibility, output review, decision gates, and feedback loops. The Edit Model's current alpha is where all five converge in a creative context, and I think the interaction design is the hardest unsolved problem in the product right now.
I'd welcome 20 minutes on how you're thinking about the control surface between exploration and precision in the editor. junochen.com/trust-is-the-new-interface has the framework; junochen.com has the production work underneath it.
If warm path exists:
[Mutual connection] suggested I reach out — I've been following the Edit Model alpha and think the control-surface design between user intent and generative output is the most interesting unsolved interaction problem in AI right now. I published a framework for exactly this class of problem (junochen.com/trust-is-the-new-interface) and have been building AI products in production for the past two years. Would love 20 minutes on how you're approaching it.
What not to say:
- Do not reference Co-Star's astrology product as a design precedent for Midjourney. Guler acquired into a new context; treating her as "the astrology person" misreads the room.
- Do not describe the Edit Model's experimental state as problems to fix. She knows. Framing yourself as someone who will "clean up" the product signals you don't understand the pace.
Cover letter draft
The hardest design problem at Midjourney right now is the control surface between what a user intends and what the model interprets. In the Edit Model, natural-language instructions create a gap: the user says "warmer lighting" and has no way to preview, constrain, or understand the model's interpretation before it executes.
I've spent the past year building a framework for this class of problem. "Trust Is the New Interface" maps five handoffs between human intent and AI action — from how intent gets communicated through how the user evaluates and redirects output. At Agentic Labs, I built three live AI products (Brand Pulse, Retail Velocity, Carrier IQ at junochen.com) that implement these handoffs with live users and working agents in production. As Product Design Director at BCG Digital Ventures, I designed Allē (30M members, 3.2× redemption rate, $42 CAC) — consumer scale with craft depth.
Most recently as Head of Product at TinyFish, I shipped an enterprise agentic platform from 0 to 1 in three months, working daily with agent traces, auditability, and governance challenges when AI acts on a user's behalf. I moved to product to build AI-natively; I'm returning to design because I want to apply that technical depth to a specific, high-stakes creative product.
The framework, the production work, and the case studies are at junochen.com. I'd welcome a conversation about how you're thinking about the interaction architecture for the editor.
Resume emphasis guidance
Craft depth: 3. AI exposure quality: 3. Hands-on AI-native role.
- Lead with: Agentic Labs (all three apps — live AI products, solo-built, demonstrate the Trust handoffs in production). Trust Is the New Interface essay immediately after.
- Follow with: Allē (consumer scale, dual-surface, craft depth — Product Design Director at BCG DV) and Alibaba (enterprise trust design at $50B+ GMV — proves you've operated at scale).
- TinyFish: Header line only. "Most recently, Head of Product at TinyFish (enterprise web agent platform, Series A). Shipped agentic platform 0→1 in 3 months." One bullet on agent traces and governance in production. Past tense throughout.
- Trim: Red Cross and Thermo Fisher to single lines. They're regulated-platform proof, not creative-product proof. Equinox+ can stay as a 0→MVP speed signal.
- Frame the transition: "Moved to product to build AI-natively from zero. Returning to design to apply that depth to a specific creative product." One line. Don't over-explain it.
Post-outreach signals
AI-native startup, ~60 people. Response within 48–72 hours is healthy. Silence past 5 business days means the message didn't cut through or the role is further along. If Guler engages the design problem you named — the control surface, the trust gap, the Edit Model — rather than sending a generic acknowledgment, that's a strong signal. Respond within 24 hours and deepen the thread. One follow-up at day 6 if no response, referencing a new product signal (the next Edit Model update, a new alpha changelog entry). Never "just checking in."
The non-obvious thing
The Product Designer posting — the junior role running alongside yours — lists required proficiency in "Figma, Claude, and Cursor." Claude and Cursor are AI coding and design tools. Midjourney's design team is expected to build with AI, not just design AI products. Your TinyFish experience shipping production software using AI-assisted workflows is directly relevant, and most competing candidates from traditional design backgrounds won't have it. Don't lead with this. Be ready to speak to AI-as-design-tool fluency when it surfaces in conversation.
Warning signs
Equity is unconfirmed. The posting lists cash and benefits only. Historical reporting mentions profit-sharing in lieu of stock packages, but nothing confirms equity for this role. At $285K–$350K cash, the floor is strong. Without equity clarity, you're evaluating a high-cash role at a company with no disclosed path to liquidity. Ask about compensation structure in the first substantive conversation — not the cold outreach, but the first real call. If the answer is profit-sharing only with no equity, recalibrate whether the scope and portfolio value justify the absence of upside.
Founder-led culture with a new CDO. Guler is 50 days into changing a culture where projects have been led by engineers and the founder. Whether she has real authority to build a design-led practice or is operating within Holz's creative orbit is unknowable from public signal. The acquisition of her entire team is a strong investment signal. But ask in conversation: what decisions has the design team made in the last month that shipped? The answer tells you whether Guler's authority is operational or aspirational.
Guler's attention is split. Co-Star remains under her direction alongside the Midjourney design org build. How much of her time goes to each is undisclosed. The Design Lead may end up with more autonomy than the reporting structure suggests — which could be a feature or a gap depending on how much air cover you actually get.
- Midjourney's Edit Model changelog: The September 2 alpha release warns that parts of the editor are broken and changing rapidly — any new update before your follow-up window gives you a specific, current hook for a day-6 message to Guler.
- Draft Mode as intent signal: Midjourney's Draft and Conversational Modes let users generate 24 low-resolution alternatives from a single prompt, which is a concrete exploration mechanic worth referencing if the conversation turns to how users narrow intent.
- Gusto's org-transformation case: Gusto published a detailed account of converting its design team from Figma deliverables to code-based building within one quarter, including internal resistance and tooling decisions — useful context if Guler asks how you'd stand up AI-assisted design workflows for a team that hasn't used them.
- Figma's 2026 AI survey data: The Figma AI report found designer participation in development rose from 21% to 41% year over year, which gives you a market-level data point for the "designers who build" expectation visible in Midjourney's Product Designer posting.