I love polars so much

Don’t (really) know why they explicitly state the “no training on chats or files” - my understanding of their privacy policy is that they don’t do this for normal Claude, either. Now normal users are scared/wondering

Why do all chatbot apps (ChatGPT, Claude, Poe etc.) think a search over your chats is not an essential feature?!

Wait, are there stealth models in the artificialanalysis image arena? There seem to be three new models in testing, with Saturn-Next being really good.

My most used AI feature on Kagi (the summarizer) also got a makeover, featuring their amazing mascot ^_^

>We plan to make it more broadly available later this year. lmao i just made an extension for replicating this . link below

Xeophon claude.ai/recents lets you search.

Someone is generating training data

How do I unsubscribe from web dev twitter

Comparison LiveCodeBench Jan-Sept: Sonnet 47.1 Turbo 44.5 4o 43.2 🐋-2.5 41 🆕 405B-FP8 40.4

Pay 2K to access the latest and greatest features* * if you are in a small, selected group and have a MacBook. Maybe.

The used models:

all the unc engineers ( who started before nov 2022) have an effort estimator built-in from years of writing code without llms , that estimator needs to be overridden at the split second compile time. you can make bigger things now , just need to stop blocking yourself .

Today I’m trying out the Gemini API wish me luck chat

Two release from the whale DeepSeek 🐋 Oblig Teortaxes▶️ YouJiacheng called it huggingface.co/deepseek-ai/De…

It was literally a thing of 2 minutes. The two gripes I have is that I cannot create a new project in the Google AI studio and the python sdk downloads soooo many dependencies, what the actual hell

Really good video by Internet of Bugs TL;DW: Implement a simple HTTP server from a coding site (Codecrafters) Results: Cursor > Codeium >>> Jetbrains > Copilot

After verifying the required setup (with system prompt, no prefilling), I can safely say Reflection does not do well on BigCodeBench-Hard, at least. Complete: 20.3 (vs 28.4 from Llama3.1-70B) Instruct: 14.9 (vs 23.6 from Llama3.1-70B) The CoT/thinking/reflection process