@antiflasher I experimented with 0.5px borders several times, but ultimately abandoned this approach because while it looks good in Figma, it rasterizes to subtle color 1px in the browser in some cases
We at @evilmartians recently got a request from a client:
“Make my devtool landing page look like it’s 2026, not 2016”
So I revisited the examples I analyzed a year ago (https://t.co/wLW658awzj) to see what’s new on the scene. And there’s quite a bit 👇
@pablostanley I believe delivering joy/delightfulness is a crucial skill for designers nowadays to survive, as it adds value to AI. But it seems this is a stage of product development that not every company can reach
Five years ago, we posted an article on writing custom RuboCop rules (called "cops"). Since then, RuboCop had a major release with new features that make your custom rules way better.
This article by @viralpraxis revisits our previous post and explains how to write custom cops today: https://t.co/4xPngYjgpZ
PostCSS creator, @sitnikcode, shares how to make your open source popular.
He also destroys the false idea that writing open source will make you famous, or that you should write open source to show your expertise. Plus, he explains how to avoid burnout when maintaining your products.
https://t.co/fY56manqaH
Optimistic UI is a lie. And the modern web has trained everyone to expect interfaces that lie convincingly.
In this new post, see how with Inertia Rails, telling that lie takes remarkably little code because of how Inertia handles state:
https://t.co/zNT6wttX4j
I’ve been using Sonnet 4.5 for the last months as a core coding model, but it seems ChatGPT 5.2 performs even better. Almost no loop-like errors, quick expected turnaround
a week in design looks like:
mon: claude code in cursor, as weeks before
tue: nothing new
wed: get $250 tokens from claude code and try their research preview ui
thu: install google antigravity to try out
what's on friday?
thanks god for design-in-code, no figma vendor lock!
I wrote this because the evals debate feels stuck in abstractions and theory.
Here’s what it actually looked like inside GitHub Copilot, what worked, what didn’t, and what we learned building evals in the wild after that with @QuotientAI. Lessons learned and unrequested advice if you are building an AI product today.
@julianeagu One assumption I have. In tech our wins come in a passive state – sitting, typing, still. The body learns it can get dopamine w/o moving. Over time it stops valuing effort that requires hands, legs and sweat. So you end up coding a pet project for soul on weekends 🥲