
Models are great at writing code. They still struggle with design judgment. We hope to fix that. Introducing Fieldwork: fieldwork.software/writing/i…

Building Fieldwork.software. Led Design & Research @Coinbase, supported design systems @Airbnb, and teams @Facebook

Models are great at writing code. They still struggle with design judgment. We hope to fix that. Introducing Fieldwork: fieldwork.software/writing/i…

What felt fresh two years ago is today’s norm. The map keeps moving and models struggle to keep up.

The catch is that the best work usually sits outside where most work clusters. If you’re only good at the average, you never find the higher peaks.

Models struggle with this. They train on the most popular work on the internet and collapse toward it. Ask for a landing page and you get the average landing page.

Our working definition of “taste” at Fieldwork: knowing where your work sits in a solution space that shifts as preferences change. Humans need it, and models do too.


What if you could do a design crit with your agent?
How can audio models help designers? An internal tool that lets you quickly leave feedback and make small edits on any website:

At Fieldwork, we’re working to measure design variety so we can improve it Designers already push well beyond these defaults with better prompting, skills like @impeccable_ai, and tools like @withgraphical Still, it’s striking how often different models arrive at a very similar solution All outputs and methodology: fieldwork.software/eval/web-… Text-to-image results: fieldwork.software/eval/web-…

Text-to-image models show more visual variety on the same task but the results have their own flaws

Interestingly, Haiku 5.5 shows some surprising variety across three takes on this travel booking site

How much more creative could models be by default? We asked frontier and open-weight models to design a landing page with “unique personality and taste” three separate times using the same business brief and images It almost looks like we got the same site

It's the Super IC era. I put together some thoughts to help you navigate it: gabrielvaldivia.com/notes/in…

The best tools live in your codebase.
Introducing Graphical. It's a simple, powerful, and fun tool to design visual languages, style components, and work with coding agents to create interfaces that look memorable and feel unique. I hope you'll check it out at graphicalui.com!

Why does this matter? Good icons make apps usable and distinctive. This process also offers a glimpse into evaluating models at logo design At Fieldwork, we're building better evals and data for design tasks. If you’re trying to improve your model’s design capabilities, reach out. We’d love to help Full benchmark with prompts used and outputs on our website: fieldwork.software/eval/icon…

Here are the results, graded by humans on recognizability, process, consistency, and execution
