Michael Filipiuk

New work: Massive Good 🌿

Massive Good is a charitable organization that helps people case by case, pairing structured intake technology with human caseworkers who verify each need and implement the right solution.

I was able to help with designing & building the website, copywriting, and preparing assets.

Every few weeks someone on here announces that designers are finished. This week it was 'software is solved' and 'designers still working in Figma will be gone by 2027'.

The teams furthest ahead with AI are seeing something else. Anthropic's head of product design says their engineers got about 10x faster with agents, and their designers 3 to 5x. One of their teams stretched its standups from 30 to 90 minutes, and spent the extra hour arguing about what to build.

When building gets that cheap, the slow part of a product is deciding what it should be. That was always the designer's real job. It used to be hidden behind weeks of drawing screens, and now it's most of what takes time.

I said this last year and I still think it: some founders will be happy with whatever the agent gives them, the same way some paid $100 on Fiverr for a mobile app. Others will keep paying $10,000 or $100,000, and what they're paying for now is mostly the deciding.

Design tip: dark mode needs less contrast than you think.

Even Apple put an icon next to every menu item, then took them all out a year later. Their guidelines now use those menus as the example of what not to do.

An icon has to earn its spot. Next to text that already says everything, it's noise.

The most saved design post on X this week was a skill that makes your agent's UI look more polished. People passed it around as someone open-sourcing their taste.

Skills are great for a specific kind of thing: easing curves, hit areas, focus rings, how a drawer drags, what a button does when you press it. Those should feel the same in every product, and it's good that the best version is one command away.

How your product looks is a different job. If thousands of people install the same taste, they get the same product. We've done this before with UI kits, and again with everyone shipping the same component library on its defaults. In one study, Claude Code picked the same library for 90% of its component choices.

What works for me is a small starter file of my own decisions: the spacing I always end up with, the components I actually use. Agents copy whatever is already in the codebase, so the first screens I make by hand become the template for everything after.