Grace Li

keeping up keeping up

3. DesignArena Why Elon Musk has been boasting Grok 4 to be a coding monster, it doesn’t look great on front-end tasks and no one comes closer to Claude 4 on that case, and not just topping - Grok 4 is not even in the top 5.

Seriously the most incredible team - seeing your trajectory of the last two years has been remarkable Can’t wait for what you’re cooking up next :)

revealing differences in underlying training datasets

can models have dissociative identity disorder

can models get depressed

so when we figure out continuous learning, will there be therapists for AI models? is that a post AGI plan, or is it turtles all the way down

absolutely love @threejs and have been comparing how different LLMs compare at 3D Design with it! thoughts in thread

So far this is a good battle arena to compare Vibe Coded apps with the same prompts. It has a leaderboards to track the best model as well. designarena.ai

R1 design dominance — genuinely unexpected

Interesting research, win-rates judged by real human votes. Claude 4 and DeepSeek R1 are basically neck and neck, Grok is surprisingly good. designarena.ai/leaderboard

🤣🤣🩴

Claude just swept the AI design Olympics. Llama showed up with flip-flops. 🫠 designarena.ai/leaderboard

grok 4… but can it make things pretty 🤨

okay help we need pagerduty.