
📸 AllScreenshots → @allscreenshot 📝 Paperstack → soon 💻 WeCodeLive → soon
Sydney, New South Wales
Joined July 2009
- Tweets12.4K
- Following785
- Followers2.1K
- Likes5.3K
GPT 6.1-sol is pretty good with UI and UX, but not out of the box. It requires some explicit nudging to turn UI from being very basic, with no formatting, no structure, into something which is actually quite decent.
Giving explicit instructions nudges GPT into spending actual attention and automatic verification on the user experience, with far better results!
Prompt below!
Treat UI/UX quality as an acceptance criterion alongside functionality.
Before implementing:
- Inspect the existing interface and follow its design conventions.
- Identify the user’s primary task and establish a clear visual hierarchy.
- Choose appropriate typography, spacing, alignment, grouping, and controls.
While implementing:
- Use the platform’s native components thoughtfully.
- Make primary actions obvious and secondary information quieter.
- Size content naturally; support wrapping and resizing.
- Design loading, empty, error, disabled, and selected states where relevant.
- Preserve accessibility, keyboard navigation, and theme compatibility.
Before finishing:
- Run the actual interface and inspect screenshots.
- Check realistic content, long text, and smaller window sizes.
- Review hierarchy, density, whitespace, alignment, contrast, and consistency.
- Fix visual problems and inspect the result again.
- Verify that the primary user flow still works.
Deliver a visually considered interface. Include a screenshot of the
finished result and identify any states you could not verify.
Improved Mojo twice more: one improvement for List/Span checking improvement of a factor 2.5x, and Array accesses at comptime has been improved by 20%.
github.com/modular/modular/p…
github.com/modular/modular/p…
Just fixed a compile time regression in Mojo 1.2. Improved performance by a factor 27-36x for structs with many Optional fields.
See github.com/modular/modular/p…
"I reproduced the slowdown and found where the time goes: AArch64 instruction selection, specifically the first SelectionDAG combining pass, consumed almost all of the roughly 54-second build. LLVM’s earlier optimization passes were fast. I’ll isolate that backend case next."
Expect performance improvements in Mojolang soon! @modular @clattner_llvm
Tesla should make every Cybercab unique, so you can collect rides like people collect Pokemon cards.
Use Taikai to Vibe maintain a Kotlin or Java codebase!
Vibe coding is fun.
Vibe maintaining a 100k-line Java codebase? Not so much. 😅
As AI generates more code, automated architecture checks become even more valuable.
Meet Taikai: architecture rules for your Java tests.
⭐ github.com/enofex/taikai
Super quick release by @mnhocktweets ! Taikai + Kotlin are the ingredients to force the LLM to do the right thing in the right way!
🚀 Taikai 2.6.0 now supports Kotlin, thanks @epragt!
Bring architectural testing to your Kotlin codebase with Taikai. Define and enforce architectural rules, catch unwanted dependencies, and keep your codebase clean and maintainable.
Release notes 👇
github.com/enofex/taikai/rel…
"When you use PHP, Ruby, or NodeJS, you're kinda asking to be vulnerable for CVEs. If you really want to be secure, use a safe programming language like Rust"
Meanwhile: github.com/zip-rs/zip2/secur…
Lots of more stuff coming to Mojo! @clattner_llvm
New Mojo Pro update 🔥 You can now type into the console while debugging, so programs that read stdin just work (turn it on in the Mojo Debug config). Change Signature also landed. Rename, reorder, add or remove params and every call site in the project gets updated in one undo.
plugins.jetbrains.com/plugin…
Nice, Kotlin support PR for Taikai got merged yesterday 🎉 Big thanks to @mnhocktweets for being so open to a pretty massive PR. Taikai now will work on Kotlin projects starting at the next release, with a Kotlin DSL too, so Taikai will be helpful to even more projects!
Taikai, the ArchUnit extension, now supports a Kotlin DSL and Kotlin projects! Taikai will help you define rules about your architecture in a deterministic way. This, together with other static analysers, is a must use in a the age of AI!
Replying to @mnhocktweets
github.com/enofex/taikai/pul…
It looks like a big PR due to the move of some files to Taikai core, but this would give a cleaner approach. Happy to make changes, just let me know.
This would have been amazing, but it seems to be more of fantasy than ever. @TeslaAUNZ , please reconsider this!
Our Markdown Pro plugin now supports Pele, by @kepano. Pele is serious good stuff, and to be fair, Kepano is producing a lot of amazing stuff (Knap, Pele, and of course, @obsdmd)
plugins.jetbrains.com/plugin…







