Founding Product Designer
The role of taking products from 0 → 1.
Then and now (with Claude).
AI changed the work. Now, we design sequences of trust: how people understand, evaluate, and get a result they previously thought impossible.
Selected guide
What Changed
AI reshaped the work itself. The opportunity is no longer just making screens: it is designing the systems that decide, generate, and explain.
Use this site if you are
The shift
What actually changed
01
Execution collapsed
AI collapsed the gulf of execution. The effort didn't disappear — it moved to the user, who now has to answer a harder question: did it do this well, and how would I know?
02
Trust must be designed
A probabilistic system can't earn trust through predictability, because it isn't predictable. Trust has to be designed, on purpose, surface by surface.
03
Instachurn is the failure mode
Users don't complain anymore. They try you once, decide silently, and never come back. That's the failure mode. Nobody files a bug report for distrust.
The job now
Three responsibilities. Not ten. Three.
- 01
Design trust into every surface
Evidence attached to outputs. Honest uncertainty. Preview before consequences, undo after them, graceful recovery when the system is wrong.
- 02
Keep the product simple so users stay motivated
Show the one thing that solves today's problem. Let mastery unlock the rest. Demotivation is churn.
- 03
Write everything down
Your team includes coding agents, and agents build exactly what your instructions say. Writing isn't how you explain the work anymore. It is the work.
Workbench → review bench
The surface you design moved
You used to design the workbench — the tools users operated to do the work. Now you design the review bench — the surfaces where users judge, correct, and approve work that arrives already done. Every execution surface AI absorbs spawns an evaluation surface someone has to design. That someone is you.
| The work | Execution surface (what you used to design) | Evaluation surface (what you design now) |
|---|---|---|
| Writing an email | A blank draft and a cursor | A finished draft with its sources linked, one fact flagged as uncertain, and recall after send |
| Editing an image | Lasso, mask, and layer tools | The AI's cut with before/after compare and edge-level correction — not redo-from-scratch |
| Shipping code | The editor | The diff, the tests that ran, and the agent's stated plan next to what it actually did |
| Analyzing data | The formula bar | The chart with every number traceable to its source rows in one click |
| Taking action (send, book, buy) | The form | Preview-then-commit, and one-step undo after |
Read the right column again. None of it is decoration — it's where the user answers the only question that matters: can I stake my name on this? Skip these surfaces and users judge in the dark. Dark judgment defaults to distrust.
Motivation loop
The three questions, in order
Get attention. Keep it. Cut cognitive load. In AI products the third also sequences trust — users graduate from low-stakes wins to real delegation.
Get attention
A real struggle × a promise specific enough to feel personal. The glance is free. Everyone gets one.
Keep attention
Fast, felt progress on the problem the user actually came with. Retention compounds. Advocates recruit.
Cut cognitive load
Progressive disclosure. In AI products this doubles as trust sequencing — users graduate from low-stakes wins to real delegation.
Capability bar
What you must be able to do
Build in real code
Streaming states, recovery flows, and confidence displays can't be judged as static frames. A working prototype this week beats a deck about it.
Decide what's next, and say why
There is no roadmap to execute. You help write it, arguing from the user's job, anxieties, and desired progress.
Write
Skills, conventions, instruction docs — markdown is how your judgment reaches humans and agents intact.
Prototype vertically
One feature built completely with realistic data, not 168 static states and a prayer. The prototype is the spec.
Then / now
The comparison
| Before | Now |
|---|---|
| Made the product easy to operate | Makes the product easy to judge |
| Designed screens and flows | Designs behavior, trust surfaces, and instructions |
| Shipped Figma files; engineers translated | Ships working slices; engineers and agents extend |
| Explained work in specs and meetings | Writes prose that machines also execute |
| Earned trust slowly through reliability | Designs trust explicitly, from session one |
| Fought for usability | Fights for motivation — demotivation is churn |
The Claude Skill
Your design judgment, installed in the agent.
Skills, conventions, instruction docs — write them once, then let the agent build what you meant.
Get the skill pack
$ git clone https://github.com/knurture/founding-product-designer-skills
Go deeper
Ten guides, one argument.
The Gulfs Have Flipped: Why Evaluation Is the New Foundation of Product Design
The core argument. Execution collapsed, evaluation became the bill, and verification design is now the moat.
Trust Is Not a Feeling. It's a Surface You Design.
The four trust patterns (evidence, honest confidence, reversibility, graceful recovery), progressive trust, and a 20-minute trust audit.
The Founding Designer's Stack: Figma, Claude Code, and Markdown
Figma for intent, Claude Code for behavior, markdown instructions for judgment — with a real example of what an instruction file looks like.
Nobody Can Build From 168 States: The Case for Vertical Prototyping
Why static state matrices fail, how vertical slices with realistic dummy data become the spec, and what this changes about handoff.
Attention Is Cheap. Motivation Is the Product.
Getting attention, keeping it through felt progress, progressive disclosure, and the one-screen audit that predicts retention.
Hiring (or Becoming) a Founding Product Designer in the AI Era
A decision guide for both sides of the table: scorecard, green/red flags, interview kit, self-assessment, and a first-30-days plan.
Chat Was Never the Interface: A Field Guide to Emerging AI Design Patterns
Seven interaction patterns worth stealing (receipts, ghost text, preview-then-commit, earned autonomy) and the anti-patterns to name and kill.
The Bottleneck Moved: How Iteration Changed When Building Became Free
Iteration when building is nearly free: the five-day loop, the kill rate, and why shipped slop spends user trust.
Taste Is the Last Moat: When Anyone Can Ship Anything, Judgment Wins
When everyone rents the same models, judgment is what's left to own — and how to train it and write it down.
Slow Is the New Broken: Designing the Wait in AI Products
Users punish silence, not latency. Streaming, named work, partial results, and honest progress — the wait is a design surface.
Bottom line
You are no longer the person who makes the product easy to operate. You are the person who makes it worthy of delegation — quickly, visibly, on every surface the user touches.
The foundations didn't change. Jobs-to-be-done and UX discipline are why any of this works. What changed is where the work bears weight.
