<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>Glitterball: Design after generation</title>
    <link>https://www.glitterball.dev/blog/</link>
    <description>Writing by Mark Boyes-Smith on evals for product design.</description>
    <language>en</language>
    <item>
      <title>Prototypes got cheap. Judgement didn't.</title>
      <link>https://www.glitterball.dev/blog/prototypes-got-cheap/</link>
      <guid>https://www.glitterball.dev/blog/prototypes-got-cheap/</guid>
      <pubDate>Thu, 01 Oct 2026 09:00:00 GMT</pubDate>
      <dc:creator>Mark Boyes-Smith</dc:creator>
      <description>Generating an interface now costs minutes. Deciding whether it's any good still costs a meeting. That gap is the most important problem in product design right now.</description>
    </item>
    <item>
      <title>AI slop is what the average looks like</title>
      <link>https://www.glitterball.dev/blog/slop-is-the-average/</link>
      <guid>https://www.glitterball.dev/blog/slop-is-the-average/</guid>
      <pubDate>Thu, 01 Oct 2026 09:00:00 GMT</pubDate>
      <dc:creator>Mark Boyes-Smith</dc:creator>
      <description>Generated interfaces converge on the same purple gradients, glass cards and &quot;Get started&quot; buttons. That isn't a bug in the models. It's what you get when nobody states a point of view, and it's measurable.</description>
    </item>
    <item>
      <title>What design should steal from how AI labs test their models</title>
      <link>https://www.glitterball.dev/blog/steal-from-model-evals/</link>
      <guid>https://www.glitterball.dev/blog/steal-from-model-evals/</guid>
      <pubDate>Thu, 01 Oct 2026 09:00:00 GMT</pubDate>
      <dc:creator>Mark Boyes-Smith</dc:creator>
      <description>AI labs spent years learning how to tell whether a new version is actually better. Five of their habits translate directly to product design. One of their failure modes does too.</description>
    </item>
    <item>
      <title>Measure what you can. Judge what you must.</title>
      <link>https://www.glitterball.dev/blog/measure-then-judge/</link>
      <guid>https://www.glitterball.dev/blog/measure-then-judge/</guid>
      <pubDate>Thu, 01 Oct 2026 09:00:00 GMT</pubDate>
      <dc:creator>Mark Boyes-Smith</dc:creator>
      <description>My first design eval flagged so much that nobody read it. The fix was to stop asking a model questions a ruler could answer, and to let the judge say &quot;I'm not sure&quot;.</description>
    </item>
    <item>
      <title>Design systems were written for people. Agents need rules.</title>
      <link>https://www.glitterball.dev/blog/design-systems-for-agents/</link>
      <guid>https://www.glitterball.dev/blog/design-systems-for-agents/</guid>
      <pubDate>Thu, 01 Oct 2026 09:00:00 GMT</pubDate>
      <dc:creator>Mark Boyes-Smith</dc:creator>
      <description>The most prolific contributor to your product's UI may soon be an agent that has never read your design system. Here's what changes when the standard has to be machine-readable.</description>
    </item>
    <item>
      <title>The design leader's job is now to write the eval</title>
      <link>https://www.glitterball.dev/blog/leaders-write-the-eval/</link>
      <guid>https://www.glitterball.dev/blog/leaders-write-the-eval/</guid>
      <pubDate>Thu, 01 Oct 2026 09:00:00 GMT</pubDate>
      <dc:creator>Mark Boyes-Smith</dc:creator>
      <description>When agents produce most first drafts, a design leader's taste can't be applied one screen at a time any more. It has to be written down, versioned and run. That changes the job.</description>
    </item>
  </channel>
</rss>
