Jeremy Fox 🦊

@mc_anthropic It’s so true haha

this man can't help but ship

plz go use this thing i built <3 code.claude.com/docs/en/channe…

Claude go brrrrrr

@sam_d_1995 Is this actually true? I had heard that approvals for new housing weren’t actually happening as much in nyc anymore (eg I had heard that much of the current downtown Brooklyn construction was based on changes under Bloomberg). Would love for it to actually be the golden age!

New York’s YIMBY golden age is upon us. this single building has more homes than were approved in the entirety of San Francisco in 2024

@martinamps Oof

omg just discovered that 6 years after acquisition not only can you not send an e-mail with your Twilio API key, but you can't even receive SendGrid webhooks via Twilio Functions 🤯

@svpino Hey there! I’ll see what I can do to fix this for you. If you don’t mind, would you dm me what you’re trying and what issues you’re running into?

I want to move permanently to Claude, but their desktop experience is nowhere near ChatGPT. You can’t upload large/multiple documents to Claude because it can’t process them. ChatGPT ingests whatever you throw at it. This is the only thing keeping me from canceling my ChatGPT subscription.

@JerusalemDemsas Thanks Jerusalem 🙏. Let us know where we’re falling short and we’ll work to fix it.

I really enjoy using ChatGPT and think it is genuinely more useful than other chatbots for various tasks but I can't in good conscience keep paying for it. I downloaded my data and ended my pro account today. Claude makes it easy to import your memory over. I recommend that people make this switch. claude.com/import-memory

To all the new Claude users today — welcome!

It’s probably easy to get caught up living your day to day, waking up in the morning, and just proceeding to exist without challenging reality and what you believe in. I’m fortunate in that I’ve probably never been put in such a difficult position; though I hope I’d make the right decision. So in many ways, what do I know? But everything is temporary and you are the sum of your actions. I’m really proud of being part of a company that cares as much as we do about the future of humanity and actually puts it into practice. I encourage folks at AI labs and other institutions to seriously consider what we know about this situation, what is truly the right thing to do without finding workarounds and concessions, and what are the consequences of actions at such a pivotal moment in history. We all only live once.

We’re #1! Thank you everyone for you support 🎉

I've always been proud to work at anthropic! in the last four years the ev of anthropic has been increasingly positive, and this is a very clean example of why you might believe that to be the case notdivided.org

@PranavDixit That’s actually a good idea…maybe we should make that

I need a Claude skill to keep me up to date with everything Anthropic is doing with Claude each day.

Very proud today to work for this company.

@mckaywrigley Thanks McKay ❤️

including a 🇺🇸 emoji in all requests to claude today. proud to be an anthropic customer. and even prouder to be an american.

anthropic. the life, liberty, and pursuit of happiness company. ai allows us to raise our ambitions. but as the technology advances, as a country we should never lower our standards. let freedom ring. 🇺🇸 🇺🇸 🇺🇸

we cannot in good conscience accede

@akshatk7 @mikeyk It’s just as thoughtful of a response! It’s the same model, but the tokens are coming out much faster. Depending on your use case, it could be worth it to be able to work much faster (and it’s certainly a rush)

Can someone smarter than me (maybe @mikeyk?) tell me why should I use fast Opus? I go for the extended reasoning one all the time assuming that generates a better response. My assumption is faster means less thoughtful response.

Claude can give you incredible health insights if it has access to the right context (genetic data, in this case)

@simonw @burkov Simon you are correct — we are not using embeddings for this. Just bash! Very cool use case @burkov ! Thanks for sharing.

@burkov What makes you certain it's using embedding for this as opposed to grep-style seaches?

If you zip your entire app (no matter the size in tokens), paste the zip file into the Claude chat UI, and ask it to fix something, it will do it. Under the hood, it will unzip your app, embed your fix description and all code files into vectors, find the files that are cosine similar to your fix description, only add the relevant files to the LLM's context, and provide the fixed files. You don't need any IDE at all.

@inductionheads The implementation in the chat app is actually different than in Claude Code — we added in some new aspects like taking the removed transcript and making this available in the file system for Claude to read back in. If you find this valuable, please let us know!

I assume Claude Code does this too? Can anyone confirm? support.claude.com/en/articles/11…