Taste Labs Launches With $18.5M to Fix AI's Aesthetic Problem
Taste Labs launched out of stealth on June 16 with $18.5M in seed funding to build the data and infrastructure layer that gives AI models and agents genuine aesthetic judgment, starting with design.
Taste Labs launched out of stealth on June 16, 2026, announcing $18.5 million in seed funding to build what founder Thais Castello Branco describes as a data and infrastructure layer specifically designed to give AI models and agents taste.
The round was led by CRV and Amplify Partners.
What Taste Labs Is Actually Building
The core problem Taste Labs is targeting is one most creative professionals already feel in their bones: generative AI can produce almost anything, but it consistently produces work that feels generic, off-brand, or aesthetically hollow.
Castello Branco put it directly in the company's launch post: "AI has [democratized access] and made it easy to generate anything. But it still feels off. Now, the challenge is judgement. What fits, what feels like you, what's GREAT."
The company is building post-training data and reinforcement learning environments focused on codifying subjective aesthetic judgment into measurable form, starting specifically with design and visual output.
Taste Labs says it is already working with frontier foundation model labs on evaluation pipelines, and with app-layer companies on brand-context and verification tooling.
Competitive Context
The dominant AI training and evaluation ecosystem has been built around performance benchmarks tied to logic, reasoning, and code accuracy.
Aesthetic judgment has largely been left to user feedback loops, which tend to reward novelty over craft.
Taste Labs is positioning itself in the gap between raw generative capability and the kind of on-brand, curated visual output that creative teams and video producers actually need for professional work.
No direct competitor is currently operating at the infrastructure layer with this specific focus on taste as a trainable, measurable signal — though larger model labs have been experimenting with aesthetic preference tuning internally.
The Signal in the Noise
For filmmakers and video producers using AI tools in production, the promise here is real even if the product is still early-stage.
The persistent complaint about generative AI in visual work is not that it can't produce something — it's that what it produces rarely fits a specific creative vision or brand identity without significant manual correction.
If Taste Labs succeeds in making aesthetic judgment a trainable layer that model developers and app builders can actually plug into, the downstream effect on tools like AI-assisted color grading, motion design, or on-brand asset generation could be meaningful.
That's a significant if, and $18.5M in seed funding is a starting point, not a finished product.
The company is pre-product in any publicly accessible sense, and its current work is happening at the infrastructure and partnership level.
Castello Branco's stated mission — "to end AI slop" — names the right problem. Whether the approach holds up under the complexity of translating subjective taste into RL reward signals is the question that the next funding round will likely answer.
Specs & Pricing
Taste Labs has not announced a public product, pricing structure, or availability timeline. The company is operating at the infrastructure and model partnership level. Follow developments at Taste Labs' official channels.
