Jai
@JSupa15
ML Engineer on Twitter to see cool stuff and meet interesting people. Founding Dev at @NousResearch
You might like
Not how it works, 50 year will be the only one affordable after it causes houses to become more expensive to level off from the change.
Sure, it's totally for mosquitoes and not assassinations guys why would you think that ahahaha
Tornyol (@tornyolsystems) is building micro-drones that kill mosquitoes. They use smartphone microphones, car park assist sensors, and some clever DSP and control to transform 40-gram toy drones into mosquito killers.
Kimi-K2 apparently cost $4.6M to train but GPT-5 is obviously better
We have just finished training Hermes 4 on ByteDance's 36B model decentralized on psyche but did not do this one ^_^ psyche.network/runs/hermes-4.… Doing some other ablations on that as well rn on there - will share when released!
imagine if it was Claude "Oh I see now! You're absolutely right!" *blows up a random wedding party with a reaper*
JUST IN: U.S. Army general admits he uses ChatGPT to make "key command decisions"
Every time I chat with the @NousResearch I get more amped. Psyche Network on @solana next! Nous Con Next Week in SF! @theemozilla @dillonrolnick @karan4d @Teknium1 @JoeLi5050 @nullvaluetensor @max_paperclips @rogershijin @bha_ku21 @artemsya @Aboozle @ropirito @JSupa15…
I've never seen the Python community embrace any tool faster than they did with uv. uv is likely the best Python tool of the last few years. If you aren't using it yet, stop what you are doing and look into it. If you are already a user, check out the attached cheatsheet.
Meta just did the unthinkable. They figured out how to train AI agents without rewards, human demos, or supervision and it actually works better than both. It’s called 'Early Experience', and it quietly kills the two biggest pain points in agent training: → Human…
Interesting experience from @karpathy on this new LLM training repo.
Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,…
The technology that is best baked into the LLM have the real moat because people (kids especially) are going to end up gravitating to what the LLM knows best. Python will never die because LLMs can get python scripts to work ootb so easily.
Some crazy people on Reddit managed to extract the "SynthID" watermark that Nano Banana applies to every image. It's possible to make the watermark visible by oversaturating the generated images. This is the Google SynthID watermark:
Please don't call yourself a backend engineer if you haven't read this book. The secret behind how Google runs production systems.
Everything you need to know about RL environments in atropos is right here:
Next up we have Nous' very own @rogershijin, explaining the Atropos project, RL Environments inside of it, and how to get started. Check out the Atropos repo here and get started: github.com/NousResearch/a…
JUST IN: Microsoft, $MSFT, has withdrawn its proposal for a data center in a Milwaukee after community outcry, per FORTUNE
RL env for OF simulation, I'm gonna call it AlphaHo
Wonder why IBM's Granite model thinks it's name is Hermes 😘
Dude I am extremely bullish on private solar panels. Own your own means of producing electricity. It's not as efficient but you reduce the risk of being priced out by these data centers that are popping up.
If you bring your own power and dont suck up the grids' I think it will be acceptable but no one can afford the power prices as they are for their measely homes sir
United States Trends
- 1. #WWERaw 53.2K posts
- 2. John Cena 51.7K posts
- 3. Packers 42.8K posts
- 4. Eagles 67.4K posts
- 5. #RawOnNetflix 1,383 posts
- 6. GRAND SLAM CENA 12.4K posts
- 7. #GoPackGo 4,495 posts
- 8. Jalen 14.9K posts
- 9. Cade Horton 1,114 posts
- 10. Tush Push 10.1K posts
- 11. Shipley N/A
- 12. Drake Baldwin 5,125 posts
- 13. Dirty Dom 1,544 posts
- 14. Go Birds 17.5K posts
- 15. Green Bay 10.4K posts
- 16. Monday Night Football 6,229 posts
- 17. Triple H 5,486 posts
- 18. Dominik Mysterio 10.1K posts
- 19. Edgerrin Cooper N/A
- 20. #ANDNEW 2,586 posts
You might like
Something went wrong.
Something went wrong.