Hey PH 👋 Building in the AI logo space — here’s what’s been harder than expected

Hey everyone 👋

Quick intro: I’m Yalan, and I work on LogoAI.

I’ve been deep in the generative AI space for a while now, and there’s one weird disconnect that keeps bothering me:

We can ask ChatGPT to write complex code or generate incredible artwork in seconds, but if you want a clean, usable logo for a real project, a lot of “AI logo tools” still feel surprisingly close to template pickers from a few years ago.

The deeper we get into this space, the more I realize that a good logo is fundamentally different from a good-looking image.

Nobody wants to write a 40-word design prompt.
Non-designers shouldn’t need to understand kerning, negative space, composition, or visual balance just to get a decent result. Ideally, the product should do more of the heavy lifting.

“Minimal logo” can mean 50 different things.
Ask a law firm, a matcha cafe, and a developer tool for something “modern and clean,” and you’ll probably get three completely different directions.

Getting enough context to understand what someone actually wants — without forcing them through a 15-step design questionnaire — is surprisingly hard.

Micro-editing is still the final boss.
Generating 20 new ideas is easy. Tweaking the one you already like is where things start to break.

You say, “make this stroke slightly rounder” or “keep everything the same, but make this part a gradient,” and the model may change three other things at the same time.

For real logo work, precise editing can be just as important as generation.

A lonely icon isn’t a brand.
Once someone has a logo, they usually need colors, typography, logo variations, social assets, and usable exports pretty quickly.

And ideally, all of those should feel like they came from the same visual system.

We’re exploring a lot of these problems with LogoAI 2.0, but we’re still very much in the cycle of building, breaking things, and learning.

Curious how everyone here handles branding when bootstrapping a new project:

Do you use Adobe or Figma, prompt ChatGPT, hire someone, or just pick a clean font and call it a day?