im no expert, but just by looking at it you can clearly tell which models (claude) the chinese are distilling to train theirs
Here's an example of a creation by GLM 5.2 on @DesignArena, made from only one prompt.
Just six months ago, there was a big enough difference between open and closed weight models that we considered having two separate evaluations.
Open weights have come a long way. Congrats to the @Zai_org team for this achievement!
vlad retweeted
To all my vibe coders/non-designers…
Do NOT waste precious Claude code tokens on design again before you do this:
1. Go to banani .co (it’s FREE) to your Web/mobile app screen mockups.
2. Literally just submit a prompt & first output is stupidly good.
I feel like the Pablo Picasso of UX/UI with this tool. 😂
cool to see "generate next screen" interaction we explored early on at @BananiAI picked up by other gen UI design tools
BREAKING 🚨: Google keeps working on the Interactive Mode for Stitch, where AI will predict how the UI should look after navigating to the next page.
It is a glimpse into how the internet will work in the future. All UI will be generated on the fly and react to user interactions
Some other upcoming features:
- Image Mode "Generate design inspirations"
- Annotations
- Export to Jules
state of hiring in 2025: candidates don't even bother to remove prompts to LLM from their intro messages
summary is using other notion blocks, like checklists, time, mentions of people, that you can move around to other pages!
tested openAI’s new gpt-image-1 with low / middle / high quality settings. even "low" variation has its moments. some observations:
tried “Mickey Mouse playing poker in a casino”, aaand nothing. openAI's copyright filter still going strong even in the new image gen
