Skip to content

Spotlight · one maker a month

The craft behind the queue

A monthly interview with the person who submitted the most interesting thing to the queue — about the decisions, not the launch. 3 months are written so far, drawn from each maker's own submissions and gate results.

Honesty label: the interviewees are the sample roster personas used across the moderation demo. The answers are written from their submitted work to show the format a real interview would take — they are not quotes from real people, and we would rather say that than pass off invented voices.

mo

2026-09 spotlight

Monoflow

@monoflow · Checkout & flows · joined 2026-08-12

Maker page →

The 99 was luck. The checklist was not.

You hold the highest audit score in the queue. What did you actually do?

Focus order first, then colour. The contrast pass takes ten minutes and fixes the score more than any clever markup does.

Your challenge entry swapped a spinner for a skeleton. Why?

A spinner is a promise. The skeleton is the truth — the page is going to look like that in 400 ms, so show it.

What do you cut when a brief asks for more motion?

The second animation. There is almost always one that carries the meaning and one that carries the ego.

ka

2026-08 spotlight

Karina Sole

@karina_ui · Micro-interactions · joined 2026-07-15

Maker page →

Under 2 KB, every time.

How do you keep loader bundles that small?

I write the CSS before the component. If the state can live in a data attribute, it never needs a hook.

Your flip loader was picked apart by the judges on honesty. Fair?

Completely fair. I faked measured progress because it looked better. The bar that admits it does not know is the better component.

Advice for a first submission?

Ship the boring version with clean gates. A 78 with no warnings teaches you more than a 95 you cannot explain.

gl

2026-07 spotlight

Glyph RGB

@glyph.rgb · Prompt engineering · joined 2026-07-08

Maker page →

Prompts are specs, not wishes.

You write prompts like requirement documents. Why?

Because the model reads constraints better than adjectives. ‘Hero, 62 chars, two CTAs, no video’ beats ‘stunning modern hero’.

What do you do when a model fails a section?

I do not rewrite the prompt blindly. I find which constraint it dropped, then say that constraint twice, differently.

One line for someone who thinks prompt writing is not real work?

Then neither is writing a brief for a contractor. Same job: make the constraint unmistakable.

The interview also lives in the Learn section, alongside the build-alongs and essays — the same craft argument, told by the people doing the work.

Read in Learn →