I might have my gripes @cluely but this thread by @im_roy_lee shows his understanding of business. Well written , good clarity.

I don’t know who wants to hear this. But @anirudhofficial doesn’t miss when it comes to Thalaivar . Absolute sambavam.

Lmao this is just next level…GDM suffering from success.

Absolutely Wonderful thread.

The final boss of math has spoken & it’s truly amazing how much thought is given by him to study this OAI claim. Fact is he prolly will benefit the most if we get a superhuman level AI , something finally to assist him. Even then he doesn’t give into the hype & is highly neutral.

This can work extremely well imo with tasks where a lot of expert data is available . And guess where we have literally millions of data points ? That’s right math proofs since the last 100 years.

OptionGAN combined with LLMs used as reward policy judgement mechanisms to train the inverse RL. Like a RL-ception.

“General purpose RL”-hmmmm. Looks like we have a DeepSeek moment here (GRPO). I’m guessing a new SOTA based method or idk…but it’s a stretch but inverse RL+choosing optimal policies? Basically OptionGAN-Cc- @tokenbender.

A lovely read . I didn’t know Belagaum dealt with photonics lasers and fighter jet helmets….

Added to this the general and absolute lack of real world utility. It’s like web3 people took the lack of respect for hardcore engineering (robotics , aerospace , chemical etc ) from software folks & mixed it with technical jargon to look cool & tell everyone this field has money.

Idk it might be a hot take , but I am yet to see definitive proof of any use case for web3 and its hype. I get crypto and why it might be useful but b/w the rugpulls and scams+massive centralisation, the concept of decentralised-anything doesn’t make sense.

This is just a prime meme. IFYKYK oh boy the trauma with digital electronics turned into a nice joke.

TIL the hard way once again of a mathematical rigour.A 9 page proof invalidated because my prof found an approximation that was not justified properly. I am grateful to her cause how else will I develop that muscle & the eventual paper would be top tier. Missed the paper deadline

Zuck & most top SF sharks (Elon , Jobs , Sam) have such clarity of thought & eloquence it's breath taking. They write like they think & they think they write . It's such an insane loop. The thoughts seem so simple but these are the directives for a T-10 market cap behemoth .

How tf am I running into limits everytime I use claude(pro plan) for any serious work for half an hour , then it locks me out for 4 hours. This is so frustrating , how are yall dealing with this?.

If MAXAR can pull this shit , I wonder what the military capabilities are of the US DoD. Also , I am giggling inside because I can understand whatever data in these 2 images fully( having a ADCS bg is nice sometimes). ECI- Earth centre inertial frame , qx,qy,qz are quaternions.

The absolute chasm of daylight between India and China that exists in terms of research output at ICML 2025 , is just heart breaking , on the flip side @HiSohan india ML tracker is great website and a tool

Bangalore sunsets are something.

Hmm , seems like Grok 4 has caught up and surpassed SOTA in some places(still skeptical for IRL use). Whats amazing apparently this is RL’ed version of Grok 3 and they are seeing RL scale well. Plus the pace of their innovation is crazy. Am I missing the RL LLM train?

Here goes nothing. dear @grok , please list my top 10 mutuals ranked by IQ. please and thank you.