jaykr_
Re:Gemini 2.0 Flash Thinking – Audio Reasoning
on Saturday February 08, 2025 @08:47PM (#42987800)
Seems like Gemini 2.0 Flash Thinking got silently updated in AI Studio to accept audio input, as well as image and text, making it the first reasoning model I'm aware of that works across audio. Trying it out on a few audio tasks (transcription, sound analysis) seems to perform a little better than 2.0 Flash or even Pro. Curious what you guys make of it!
Re:Show HN: Llama 3.2 Interpretability with Sparse Autoencoders
on Thursday November 21, 2024 @04:10PM (#42208694)
This is awesome! I really appreciate the time you took to document everything!
Show HN: Jargon, a Chrome extension to turn your browsing into language learning
https://chromewebstore.google.com/detail/jargon/gghkanaadhldgmknmgggdgfaonhpppoj
Hello HN! Jargon's a Chrome extension we've developed to turn any time browsing the internet into time learning languages.Jargon automatically picks sentences from the content you're browsing to generate a small practice quiz in a language of your choosing, helping you improve your reading and vocabulary skills. Simply click on the blue underlines as they appear to attempt a quiz. The...
Show HN: A better interface to edit written text with LLMs
https://www.llm-peer-review.com/
Editing essays or other text through existing LLM chat UI's feels clunky and doesn't give you much control. Inspired by the collaboration features in Google Docs that allow you to suggest edits that the original author can either accept or reject, as well as the ability to comment on specific portions of text, I built this small web-app to allow for a similar experience with LLM’s. Just input a...