# Chevis Zhou

> Chevis Zhou is a product design engineer (design technologist, technical product designer, UX engineer). He builds delivery systems and ships the interfaces they produce.

This index covers the public portfolio. The links below are faithful, generated Markdown
exports of the published case studies and essays. Use their original-section links when
citing evidence; inspect the canonical HTML page for visual or interactive context.

## Case Studies

- [AethrDesign: A brief-to-delivery product built around the decisions each client used to need me for.](https://cheviszhou.com/markdown/projects/aethrdesign.md): AI-directed delivery system.
- [Treevah: Rebuilding cloud file navigation, then building the design operation that could ship it.](https://cheviszhou.com/markdown/projects/treevah.md): DesignOps.
- [Solara: Developer documentation designed as a product, with a visual system built to earn trust.](https://cheviszhou.com/markdown/projects/solara-webproduct.md): Visual design & systems.
- [Solara: Sunrise identity and marketing site that replaced a developer tool's cold-blue uniform.](https://cheviszhou.com/markdown/projects/solara-marketing.md): Brand identity.
- [GMU: A graduate capstone testing whether assistive technology can rescue an inaccessible website.](https://cheviszhou.com/markdown/projects/ux-research.md): Academic research.
- [SUSTech: A MATLAB cognition experiment fitted with an algorithm borrowed from electronic design.](https://cheviszhou.com/markdown/projects/sustech-capstone.md): Research engineering.
- [Entagile: A financial staffing firm's site, redesigned around what makes the industry credible.](https://cheviszhou.com/markdown/projects/entagile.md): Credibility research.

## Writing

### Process

- [The tool I was an expert in couldn't be automated](https://cheviszhou.com/markdown/writing/framer-to-code.md): Framer approved me as an expert in April 2025. Within a year TasteLed had moved to code, because the canvas had no file a machine could write into.
- [Running client work with AI: coordination was the ceiling](https://cheviszhou.com/markdown/writing/coordination-was-the-ceiling.md): The first work I handed to AI was the record-keeping around design: status, handovers, the second copy of everything. That was the ceiling on how much I could carry.
- [Delegating design to AI: taste is a perishable asset](https://cheviszhou.com/markdown/writing/taste-is-perishable.md): The advice going around tells designers to hand over 100% of execution and keep only their taste. I've run most of that experiment for a year, and the last part is where it breaks.
- [I stopped typing code. I started directing it.](https://cheviszhou.com/markdown/writing/learning-to-direct-ai.md): A designer with no CS degree now ships production React. The method is a five-stage loop, and no stage closes until it hands over an artifact I can judge.
- [My portfolio had no client, so I invented the constraints](https://cheviszhou.com/markdown/writing/building-this-portfolio.md): A portfolio has no client, no deadline, no scope ceiling. So I manufactured all three, with AI as the engineering partner.
- [I learned enough code to read the AI's homework](https://cheviszhou.com/markdown/writing/read-the-homework.md): In February I predicted I could redo two months of design-system work in a week with AI. I was wrong in a way worth writing down: the gain was judgment, not speed.

### Craft

- [Cutting this site from five typefaces to two families](https://cheviszhou.com/markdown/writing/typography-five-faces-to-two.md): This site launched wearing someone else's five-face type system. Six prototype rounds later it ships two families, and the swaps that broke short headlines are gone.
- [Two logo systems: the mark was the cheap part](https://cheviszhou.com/markdown/writing/two-logo-systems.md): A client changed their mind about a finished logo. Swapping the drawing took an afternoon. Fixing everything the old drawing had been deciding took a lot longer.
- [A 2022 redesign: onboarding 98% of users abandoned](https://cheviszhou.com/markdown/writing/mentor-buddies-onboarding-redesign.md): Mentor Buddies lost 98% of new users before their first mentor. The redesign's highest-leverage move was letting them skip the profile form entirely.

### Tools

- [Claude Code vs. Cursor: one knows your code, one searches](https://cheviszhou.com/markdown/writing/claude-code-vs-cursor.md): Cursor resolves a clicked element straight to its source line; Claude Code has to go looking. That one mechanical difference decides which tool gets which job.

## Optional

- [Sitemap](https://cheviszhou.com/sitemap.xml): complete public-route inventory.
- [RSS feed](https://cheviszhou.com/writing/rss.xml): writing feed.