Design for technical founders: work your engineers can build
How technical founders can judge design work, and how design should reach engineering without a translation step: tokens, components and code.
- Guide
- October 2026
- 4 min read
If you can review a pull request but not a design proposal, judge design the way you judge code: does it solve the right problem, is it consistent, and can it be maintained? Then make sure design reaches engineering as tokens and components, not just pictures.
Many SaaS and AI companies are started by engineers. They can read a diff in seconds and spot a race condition in a review, but a design proposal arrives as a set of pictures, and the only available reaction is “looks good” or “I don’t like it”. Then the pictures go to engineering, someone translates them into code, and a few weeks later the product looks a bit like the design and a bit like whatever was easiest to build.
Both problems have the same fix: treat design as something with a spec, a structure and a build, like the rest of your product.
How do you judge a design proposal as an engineer?
Judge design the way you judge code: does it solve the right problem, is it consistent, and can it be maintained? Four questions cover most of it:
- Does it state the problem and the user it solves for? A proposal that starts with screens instead of a problem is like a pull request without a description.
- Are states covered: empty, loading, error, permissions? The happy path is the easy part. What a new user sees with no data, or an admin sees without access, is where products feel finished or not.
- Is it consistent: same components, same spacing, same patterns? A new button style for one screen is the design version of a copy-pasted function.
- Can it be built with what you already have? Good design reuses the system. If every proposal needs new components, ask why.
You don’t need taste to ask these. You need the same habits you already use in code review. Taste still matters, but it’s much easier to discuss once the basics are covered.
Design to engineering without a translation step
The translation step is where design gets lost. A designer draws a screen, an engineer reads it, guesses the values, rebuilds the components and fills in the missing states. Every guess is a small drift. Removing the guesses takes four things:
- Tokens in code from day one. Color, type, spacing, radius and shadow as variables in the repo, so nobody reads a hex value off a screenshot.
- Components named the same in design and code. If the design says “Banner, warning”, there should be a Banner component with a warning variant.
- Specs for behaviour, not just visuals. What happens on submit, on error, on a slow network, on a small screen.
- A working development environment where the design runs as real UI, not just a picture of it. Engineers can open it, click through it and copy from it.
Nothing should have to be translated from a picture.
This matters more now that many teams build with AI coding tools. A tool given tokens, components and written rules will extend your system. A tool given screenshots will approximate them, and every approximation becomes code someone has to maintain.
What should you ask a design partner?
Before you hire anyone, ask how the work reaches your engineers. The answers say a lot about how much translation you’ll be paying for later. For a broader list of questions, see Choosing a design partner when AI can make anything.
- How do you hand off: files, tokens or code?
- Can you work in our repo and our tools?
- Who keeps the design system up to date once the project ends?
- How do you cover states and edge cases, not just the main screens?
- Can you explain a design decision in terms of the problem it solves?
How we work
We hand over three things your engineers and AI coding tools can use directly: design tokens for every value, markdown files (like a DESIGN.md) that explain how and when to use each part, and a development environment where the design already runs as code. Nothing has to be translated from a picture.
- Design tokens: every color, type size, spacing step, radius and shadow, named by role.
- Markdown files: the intent behind the system, which component to use for what, and the rules not to break.
- A development environment: the design running as real UI, with the states covered, ready to read and build from.
We still explore in design tools when it’s the fastest way to compare directions. What reaches engineering is built to be used, not redrawn. For why one person designing and building removes so many handoffs, read The return of the generalist.
Where to start
For a new product, a Launch sprint (from US$12k, about four weeks) covers the core product flows, the identity system and a UI kit, handed over the same way. For a product that already shipped, a design partnership (from US$6k/mo) works through it one focus at a time, starting with the system. Details on our partner page.
If you want design that arrives ready to build, send your brief.
Asked & Answered
One that hands over tokens, written rules and a running development environment, and can talk in your codebase’s terms. If the handoff is only a set of static screens, your engineers will spend their time translating instead of building.
Design tokens, markdown files (such as a DESIGN.md) and a development environment where the design already runs as code.
Figma for exploring, tokens and components for building. Figma is a good place to try ideas and compare directions. What engineers build from should be values and components they can import, plus written rules for how to use them.
The way you review code. Does it state the problem and the user? Does it cover empty, loading, error and permission states? Is it consistent with the rest of the product? Can it be built with what you already have?
Keep reading
- Brand guidelines your AI coding tools can followHow to give Cursor, Claude Code, Lovable or v0 brand rules they can follow: design tokens for values, a DESIGN.md for intent, and what each one does.
- When should a startup hire its first product designer?When a startup should hire its first product designer, what to have in place before that hire, and how a design partner can bridge the gap until then.
- What a SaaS or AI marketing site needs at launchThe pages, homepage message and proof a SaaS or AI marketing site needs at launch, what to skip, and how to make it readable for AI assistants.