<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Chevis Zhou — Writing</title>
    <link>https://cheviszhou.com/writing</link>
    <atom:link href="https://cheviszhou.com/writing/rss.xml" rel="self" type="application/rss+xml" />
    <description>Chevis Zhou is a product design engineer (design technologist, technical product designer, UX engineer). He builds the pipeline a product comes out of, then ships that product&apos;s interface to production.</description>
    <language>en-us</language>
    <item>
      <title>The tool I was an expert in couldn&apos;t be automated</title>
      <link>https://cheviszhou.com/writing/framer-to-code</link>
      <guid isPermaLink="true">https://cheviszhou.com/writing/framer-to-code</guid>
      <pubDate>Fri, 28 Aug 2026 12:00:00 GMT</pubDate>
      <description>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.</description>
    </item>
    <item>
      <title>Running client work with AI: coordination was the ceiling</title>
      <link>https://cheviszhou.com/writing/coordination-was-the-ceiling</link>
      <guid isPermaLink="true">https://cheviszhou.com/writing/coordination-was-the-ceiling</guid>
      <pubDate>Sat, 22 Aug 2026 12:00:00 GMT</pubDate>
      <description>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.</description>
    </item>
    <item>
      <title>Cutting this site from five typefaces to two families</title>
      <link>https://cheviszhou.com/writing/typography-five-faces-to-two</link>
      <guid isPermaLink="true">https://cheviszhou.com/writing/typography-five-faces-to-two</guid>
      <pubDate>Wed, 19 Aug 2026 12:00:00 GMT</pubDate>
      <description>This site launched wearing someone else&apos;s five-face type system. Six prototype rounds later it ships two families, and the swaps that broke short headlines are gone.</description>
    </item>
    <item>
      <title>Delegating design to AI: taste is a perishable asset</title>
      <link>https://cheviszhou.com/writing/taste-is-perishable</link>
      <guid isPermaLink="true">https://cheviszhou.com/writing/taste-is-perishable</guid>
      <pubDate>Sat, 15 Aug 2026 12:00:00 GMT</pubDate>
      <description>The advice going around tells designers to hand over 100% of execution and keep only their taste. I&apos;ve run most of that experiment for a year, and the last part is where it breaks.</description>
    </item>
    <item>
      <title>Two logo systems: the mark was the cheap part</title>
      <link>https://cheviszhou.com/writing/two-logo-systems</link>
      <guid isPermaLink="true">https://cheviszhou.com/writing/two-logo-systems</guid>
      <pubDate>Wed, 12 Aug 2026 12:00:00 GMT</pubDate>
      <description>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.</description>
    </item>
    <item>
      <title>I stopped typing code. I started directing it.</title>
      <link>https://cheviszhou.com/writing/learning-to-direct-ai</link>
      <guid isPermaLink="true">https://cheviszhou.com/writing/learning-to-direct-ai</guid>
      <pubDate>Thu, 16 Jul 2026 12:00:00 GMT</pubDate>
      <description>A designer with no CS background now ships production React. The method is a five-stage loop, and no stage closes until it hands over an artifact I can judge.</description>
    </item>
    <item>
      <title>Claude Code vs. Cursor: one knows your code, one searches</title>
      <link>https://cheviszhou.com/writing/claude-code-vs-cursor</link>
      <guid isPermaLink="true">https://cheviszhou.com/writing/claude-code-vs-cursor</guid>
      <pubDate>Thu, 09 Jul 2026 12:00:00 GMT</pubDate>
      <description>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.</description>
    </item>
    <item>
      <title>A 2022 redesign: onboarding 98% of users abandoned</title>
      <link>https://cheviszhou.com/writing/mentor-buddies-onboarding-redesign</link>
      <guid isPermaLink="true">https://cheviszhou.com/writing/mentor-buddies-onboarding-redesign</guid>
      <pubDate>Wed, 08 Jul 2026 12:00:00 GMT</pubDate>
      <description>Mentor Buddies lost 98% of new users before their first mentor. The redesign&apos;s highest-leverage move was letting them skip the profile form entirely.</description>
    </item>
    <item>
      <title>My portfolio had no client, so I invented the constraints</title>
      <link>https://cheviszhou.com/writing/building-this-portfolio</link>
      <guid isPermaLink="true">https://cheviszhou.com/writing/building-this-portfolio</guid>
      <pubDate>Wed, 01 Jul 2026 12:00:00 GMT</pubDate>
      <description>A portfolio has no client, no deadline, no scope ceiling. So I manufactured all three, with AI as the engineering partner.</description>
    </item>
    <item>
      <title>I learned enough code to read the AI&apos;s homework</title>
      <link>https://cheviszhou.com/writing/read-the-homework</link>
      <guid isPermaLink="true">https://cheviszhou.com/writing/read-the-homework</guid>
      <pubDate>Sat, 14 Feb 2026 12:00:00 GMT</pubDate>
      <description>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.</description>
    </item>
  </channel>
</rss>