Welcome to Monday. Threads is continuing on its journey to gather as many users as possible and it doesn’t look to be slowing down anytime soon. The Twitter rival now has one-fifth of Twitter’s weekly active users. Do you think Threads will take the crown, or is it a hype train?
Have you binged Extrapolations on Apple TV+ yet?
Without giving too much away, in the not-too-distant future, the show depicts the use of virtual criminal court judges, “programmed to create a universal standard of justice reflecting legal and ethical traditions from across the globe.”
The future is now: Last week, Anthropic released Claude 2, a chatbot trained on rules inspired by the Universal Declaration of Human Rights. This enhanced version is now available in the US and UK, and it may be kinder, faster, and cheaper than ChatGPT.
Which is smartest? Depends on how you define smart. Claude 2 outperformed ChatGPT on writing and bar exams, but the latter apparently performs better on verbal and quantitative tests.
Claude’s je ne sais quoi: The feature that makes Claude a potential ChatGPT killer is its ability to handle and recall larger amounts of information.
Users can input up to 100K tokens, which corresponds to around 75,000 words. By comparison, GPT-4 is theoretically capable of context windows ranging somewhere between 8K and 32K tokens.
WTF is a context window? In generative AI, like the language models that power chatbots, the "context window" refers to how much information the model can "remember" and consider when generating its response.
Claude’s 100K context window capabilities mean “businesses can now submit hundreds of pages of materials for Claude to digest and analyze, and conversations with Claude can go on for hours, or even days.”
You can put an API, a front end or a piece of serverless AI onto Cloudflare's network without provisioning anything, picking a region, or thinking about a server again. Write it in JavaScript, Python or Rust, deploy once, and it runs in every location Cloudflare has, which is most of the places your users are. Scaling is not something you configure. The free tier is 100,000 requests a day, which is more than most side projects will ever see and enough to run something real before you pay anybody a cent.
