Interview

Taste Labs founder Thais Castello raises $18.5M to build AI for aesthetic taste and content quality

Jun 18, 2026 with Thais Castello

Key Points

  • Taste Labs raises $18.5M and launches publicly to build AI infrastructure that optimizes for aesthetic taste rather than technical correctness, targeting retailers and creators tired of indistinct AI outputs.
  • Castello argues that solving taste requires continuous retraining as trends shift, fundamentally different from one-time capability fixes like fixing six-fingered hands that fade from notice once solved.
  • The startup is recruiting creatives who accept AI as inevitable and want to shape its output, with ambitions to codify subjective taste across visual design, code style, and editorial judgment.
Taste Labs founder Thais Castello raises $18.5M to build AI for aesthetic taste and content quality

Thais Castello's Taste Labs has raised $18.5M and launched publicly this week to build AI infrastructure for aesthetic taste — targeting retailers and content creators who need more than technically correct outputs.

The core problem Castello is solving is one that anyone drowning in AI-generated imagery already feels: the outputs are competent but indistinct. Raising the quality ceiling, she argues, is fundamentally different from solving capability problems like coding or math. Unlike those domains, taste requires continuous retraining because trends shift, repetition kills value, and what reads as fresh today becomes noise tomorrow. You can't train it once and ship it.

The inbound of the last forty eight hours has been absolutely insane... it's actually the reason we took so long to launch, the reason we took so long to launch is actually because we already had more demand than honestly I think we could serve before even the launch... 18 and a half million.

The capability vs. ceiling gap

Castello thinks the capability layer — getting basic visual fundamentals like alignment, typography, and proportions right — will largely be solved within a year. What follows is harder and longer: personalization, localization, and staying ahead of trend cycles. The comparison to six fingers is instructive. That problem quietly disappeared from frontier models, and almost no one noticed because the conversation had already moved on to the next flaw. The ceiling always rises.

Her framing of "intentional diversity" is worth noting. Cranking up model temperature to produce randomness is not the same as understanding how to force meaningful variation. Taste Labs is trying to codify the difference.

Broader ambitions

The infrastructure Castello is building isn't narrowly visual. Castello herself is opinionated about how she writes code — and sees the same logic applying to creative writing, and any domain where there are multitudes of legitimate preferences rather than a single correct answer. Chip design, research taste, editorial judgment — all of these involve the same underlying challenge of turning something subjective into something that can be measured and trained toward.

Demand and community

Launch week demand was, by Castello's account, more than the team could serve — and that was true even before the public announcement. Building the creative community behind it is trickier. The creative world is divided on AI, and Taste Labs is deliberately recruiting the segment that has accepted AI as inevitable and wants to shape what it produces rather than fight it. The company's founding designer has a hat that says "picky," which is probably the right hiring signal for what Castello is building.

Taste Labs is headquartered in San Francisco with team members in New York and Brazil, where Castello is originally from. Both cities, she notes, carry strong concentrations of creative talent she's trying to pull toward SF.

Every deal, every interview. 5 minutes.

TBPN Digest delivers summaries of the latest fundraises, interviews and tech news from TBPN, every weekday.