Check out our latest research on data. We're releasing 24T tokens of richly labelled web data. We found it very useful for our internal data curation efforts.
Excited to see what you build using Essential-Web v1.0!
II-Medical is now sota across open models & beats almost all proprietary models for health knowledge
Works locally on < 8gb RAM
On track so every medical diagnosis is checked by AI, everyone has access to free medical knowledge & we can move from sick care to health care
🙏
The jobs of the future are public sector jobs.
AI CEOs & companies will outcompete non-AI ones.
The surplus they generate will be distributed by the public sector, with jobs guarantees & more.
Safest, highest status jobs will be public sector.
Need to improve public sector🙁
seems kinda janky that there are 118 known physical elements and yet none of them have been given names starting with J or Q. V and U and K also feel underutilized
the issue is imo not a generational thing, but an issue with labels. every Thing becomes cringe and uncool eventually because it’s the cringe and uncool people who cling loudly fastidiously to the labels/markers/signals of the Thing and end up representing it to the world
I’m looking for new AI writers and content creators to follow
What I don’t want:
• Theory or prognostication
• Exaggerated claims or hype
• Utopian or dystopian predictions
What I do want:
• Concrete, practical experiments
• Thoughtful analysis based on real experience
• How-to advice on using LLMs effectively
Got any recommendations?
Wow.. Now you can transcribe 60 minutes of audio in just 1 second with a completely open-sourced model 🤯
@nvidia just open-sourced Parakeet TDT 0.6B V2, a 600M parameter automatic speech recognition (ASR) model that tops the @huggingface Open-ASR leaderboard with RTFx 3380
It's open-sourced under CC-BY-4.0, ready for commercial use.
⚙️ The Details
→ Built on FastConformer encoder + TDT decoder, the model handles up to 24-minute audio chunks with full attention and outputs with punctuation, capitalization, and accurate word/char/segment timestamps.
→ It achieves RTFx 3380 at batch size 128 on the Open ASR leaderboard, but performance varies with audio duration and batch size.
→ Trained using 150K steps on 128 A100 GPUs, then fine-tuned on 500 hours of high-quality human-transcribed English data.
→ Total training data spans 120K hours, combining human-labeled and pseudo-labeled sources, including LibriSpeech, Fisher, YTC, YODAS, and more.
→ Available via NVIDIA NeMo, optimized for GPU inference, and installable via pip install -U nemo_toolkit['asr'].
→ Compatible with Linux, runs on Ampere, Blackwell, Hopper, Volta GPU architectures, requiring minimum 2GB RAM.
→ Granary dataset used for training will be made public post Interspeech 2025.
good reply game in the wild seems to be becoming even rarer than before. i think it has to do with people feeling overwhelmed by info/feeds etc and so they rush and make less of an effort on each reply. sadly this likely makes people's experience worse imo
Saw a quote today:
“To inspire people, don’t show them your super power. Show them theirs.”
People think inspiration comes from showing others what you can do. But that’s not how it works. The best way to inspire isn’t to dazzle them with your talent, it’s to reflect something dormant back to them. You don’t need to be the hero in their story. Just the mirror.
In startups, in writing, in love, people rise when they feel seen. So the real superpower isn’t being exceptional. It’s making others believe they are.
CNN asked if @DOGE is a success. Yes, unequivocally.
Even my liberal counterpart admitted as much: “There is no question that what Elon Musk managed to do, no one thought he was going to be able to do.”