@shadcn Time for upgrading.

@peer_rich Vercel is like that, many star employees. People fear ex-founders as their ability to exit, their influence and different management. Ex-founder is more like a founding member than a typical employee. They do affect the company a lot.

@BillAckman @deepseek_ai @nvidia At what point have we become so weak that a startup with only a hundred employees and little funding can disrupt a tech industry worth trillions of dollars? Doesn’t this prove the tech industry is too weak to deserve such valuation? Wake up. Let's build and compete.

@PalmerLuckey No one cares that 5M. What matters is its open nature and size. People have been led to believe only big tech can compete. Now a startup can achieve the same. If it's closed-source, nobody cares. Improvement of society is far greater than benefiting only for-profit companies.

@gustaf Competing by action is common sense now. Many are worrying others like network, integration, data, and resources. But I don’t think hiding is a good strategy either. Being open and welcoming competition are proved to be great. DeepSeek wins like this. Otherwise no one knows it.

@rauchg Lol, maybe my need is for team. I have an IMAP setup for our company email support. Multi accounts and inboxes are a requirement. There is literally no single app on web or desktop or mobile that works for my case.

@AravSrinivas tbf, it is the first true open source model that matches leading models. The MIT license is far more open than the Llama license (which is not open-source). And it comes from an unknown startup.

@rasmickyy @Tempo_Labs @Kevin1Michael @petergok I like your video on cutting edge practices. I think you already have a great audience, and it's invaluable. Will you keep updating videos and share your thoughts frequently.

@Ahmad_Al_Dahle I believe even before AI, every company was already tagging people with labels. Google does the same for search. In general I think it's convenient. Just sometimes I actually don’t want a personalized answer.

@sama OpenAI needs to rethink its approach now. Others are making AI models as a commodity. Maybe it's wrong to try to profit from AI models. Other big tech could make AI free as they are not their core business. OpenAI is really struggling on this balance.

By hiding ideas and claiming others’ findings as independent, it feels truly wrong. It completely betrays the idea of “Open” AI. There’s no credit in making this claim, and it actually damages OpenAI’s reputation. All we know is that it’s truly for profit and not actually working to improve general society.

@bunjavascript @DrizzleORM Insane, it couldn't be easier to set up persistent data during development, and easily switching to a full-fledged Postgres setup.

@steveschoger Refreshing! The colors look vivid, and the text is clear. I really like the new landing page.

App Developer API enrollment cycle lags behind the speed of AI development. Considering new APIs rely on iOS system and take years to enroll, it is impractical for app devs to directly use the AI API, say offered by Swift. Providing more low-level APIs to allow higher-level AI libraries like PyTorch to leverage GPUs and NPUs would make more sense.

@RyanCarniato @vite_js I bet on Vite; however, Next.js currently shines in generating meta tags on the fly, like open graph preview images, and there is no good alternative for this. I do see the Vite ecosystem booming thanks to you Ryan, Tanner and many others!

@AndrewYNg I think this should be common sense for anyone in tech. I'm not sure why some people react differently—some even freak out and give up. Have you forgotten how much easier development has become thanks to the AI wave. Build applications, folks!

@avichal 😂Lol, ironic, hedge fund in China and nonprofit in US.

@karpathy Agree, it unlocks much wider AI applications, and it only makes AI more bullish. Wall Street's reaction is completely irrational and feels like they don't even know which companies are actually winning.

You forgot that Apple Silicon uses Unified Memory. Running inference on-device is much more accessible on Apple devices, including iPhone, MacBook, and Mac. Currently, the most affordable options for using a local AI model are the Apple Studio or Apple Mini. You can easily get 128 GB of VRAM at an affordable price compared to Nvidia.

@garrytan Feel the same! I actually think AI agents will have much wider applications. Apple Intelligence should be much more powerful. Deepseek is a good sign that this approach will become widespread.