Log inSign up
James profile banner
user avatar

James

@atomiksdev
@Base_UI dev
Australia
github.com/atomiks
Joined June 2012
294
Following
1,632
Followers
RepliesMedia
  • user avatar
    James
    @atomiksdev
    Jul 7
    This summarizes my experience with GPT-5.5 (left) vs. Fable (right) Even if GPT-5.5 can pass tests, its solution is almost always excessively verbose and it doesn't find the minimal one
    GPT-5.5, 25 line diff
    Fable, 3 line diff
  • user avatar
    James
    @atomiksdev
    Jul 4
    One thing you get for free moving to Base UI from Radix is speed. I benchmarked the areas that most affect perf (opening a dialog on a big page, mounting lots of selects, opening big menus, sweeping over menu items etc.) In each of the six scenarios, Base UI was the fastest.
    Bar chart titled “Popup performance comparisons” showing lower-is-better timings for Base UI 1.6.0, Radix UI 1.4.3, and React Aria Components 1.19.0. Base UI is fastest for dialog open at 112 ms, select open at 275 ms, menu open at 285 ms, tooltip trigger mount at 24 ms, select trigger mount at 147 ms, and menu item highlight at 2 ms. Radix UI measures 280 ms, 484 ms, 421 ms, 25 ms, 473 ms, and 23 ms respectively. React Aria measures 268 ms, 960 ms, 814 ms, 40 ms, 410 ms, and 58 ms respectively.
    Image
    user avatar
    shadcn
    Vercel
    @shadcn
    Jul 3
    Starting today, we are making @base_ui the default component library in shadcn/ui. First, a bit of history. When shadcn/ui launched in 2023, it was built on Radix. At the time, nothing else came close. Headless. Accessible. Composable. Fast forward a few years and the team who
  • user avatar
    James
    @atomiksdev
    Jun 13
    Fable was great at code reviews I compared 4 models on the quality of their reviews of a complex React component (same prompt for each). Was nice while it lasted... Even still, they each found something the others didn’t
    Weighted score chart alt text: Horizontal stacked bar chart comparing weighted review scores. Consortium scores 129, Claude Fable 5 XHigh 82, GPT 5.5 Pro Extended 46, GPT 5.5 XHigh 45, and Claude Opus 4.8 XHigh 40. Scores are split by issue type.
    Unique issue chart alt text: Horizontal bar chart comparing unique issues found. Claude Fable 5 XHigh found 21, Claude Opus 4.8 XHigh found 7, GPT 5.5 Pro Extended found 7, and GPT 5.5 XHigh found 4.
  • user avatar
    James
    @atomiksdev
    Aug 16, 2025
    The @base_ui Combobox is looking to ship as multiple different components: 1. Combobox 2. Autocomplete 3. FilterableMenu Some of the components you can build with these:
    Multiple select combobox
    Input inside popup combobox
    @ mentions textarea combobox
    Emoji picker grid combobox
  • user avatar
    James
    @atomiksdev
    Aug 10, 2025
    Tailwind users: do you still use Tailwind v3, or have you upgraded to v4?
    Tailwind v324.1%
    Tailwind v458.6%
    Results17.2%
    58 votesFinal results