Start from the logo

The delivery project’s design was stuck: starting from Stitch mockups produced page after page of “no aguanta, nothing special”. So the process got inverted: make a great logo first, and pull everything from the logo ▸ 24:18. Precedent supports it: Pickyto’s design flowed once its logo existed; Zenota flowed from the idea instead ▸ 24:26. Different projects have different natural anchors; the mistake is starting with none, shooting blindfolded and hoping to hit style.

The logo hunt itself produced two lessons worth keeping. First, against the “AI does everything now” narrative: Gemini’s first six brainstormed prompts were all rejected on sight ▸ 25:03. What unlocked it was human: one mediocre output (“Celestial Guardian”) happened to evoke city lights at night, and feeding that association back produced the winner, an eagle-box logo textured like a city seen from a plane, dense points and lines that read as a galaxy from afar and as streets up close ▸ 26:06. The model generates volume; the person supplies the spark that recognizes and redirects. Neither alone made the logo.

seis prompts malos, una asociación humana, un logo →

Second, a small blow against design relativism: taste converged. Out of the whole batch, Julia picked her two favorites without hearing his, and they were exactly his two ▸ 22:32. Either the visual has more common sense than people admit, or two people who share everything converge on one eye ▸ 21:54; both readings are useful, and the second is a real asset for a two-person design team: a shared taste is a fast veto.

The teaching, portable: when a design won’t start, don’t push wider, push denser. One artifact that fully carries the identity, a logo, a font, a single hero image, gives every later prompt something to obey…