Hemanth

IITB | Geospatial | DL
"Attention is all you need"
Looking for intellectual stimulus

[Hemanth @hemanth_why]

@yesith_thomas @ai_for_success At end of the day, people choose cheap and best like gpt 6 sol or luna

[Hemanth @hemanth_why]

@OpenAI I'm so excited to try gpt 6 luna, as my daily driver model being gpt 5.6 luna. Good see model costs are coming down

[Hemanth @hemanth_why]

If Chip design is automated through AI/RL becomes true, AI chip prices and consumer chip prices will fall immediately. I hope jeff deans new company works on it and solves chip design time duration bottleneck from 2 years to 4 months.

[Haider. @haider1]

Jeff Dean says:

"if you speed up chip design loops enough for RL to search them, you can compress the entire design cycle — 10 people in 3 months instead of 150 people and 2 years"

That means more specialized hardware and shorter compute bets, shrinking from 2-6 years to 6 months-4 years

[Srikanth Varma Chekuri @cskvarma]

If batteries explode, the onus lies with the battery maker. If a cloud DB has a bug which leaks data to hackers, the cloud provider is liable. What makes closed source AI mishaps any different? With open source, the onus lies with the InfoSec team of the enterprise deploying it.

[Hemanth @hemanth_why]

@maxinomics Want to see how oil prices & inflation in US effect the Trump, so that he or future presidents will be careful in meddling other countries war.

[Hemanth @hemanth_why]

@MiaAI_lab Is ox alpha model from SSI first model??

[Hemanth @hemanth_why]

@kimmonismus What's the pricing??

[Hemanth @hemanth_why]

@jcrajan00 Ethanol blending implemented such way that it tainted govt perspective to common people. I'm not sure, will it will continued to increase ethanol percentage?

[Hemanth @hemanth_why]

Every civilization has four bottlenecks (Energy, Materials, Information, Coordination)
- Steam engines solved energy.
- Concrete and Steel solved materials.
- Computers solved information.
- AI is primarily solving coordination.

[Hemanth @hemanth_why]

@scaling01 @grok @perplexity_ai tell me more about le chaton fat, which company developed it?

[Aviral Bhatnagar @aviralbhat]

India is being hit with a triple whammy:

- AI rise disrupting white-collar work
- Heat rise disrupting blue-collar work
- Oil rise disrupting all kinds of work

Hard forces that are going to make this a very tough year, could likely shave off 2-3% of potential GDP growth

[Hemanth @hemanth_why]

@AbhinavXJ You can use upi lite, which has higher limits

[Hemanth @hemanth_why]

@mayank_kmr @mumbaiheritage That's the reason the government is planning for water metro/aqua metro

[Nithin Kamath @Nithin0dha]

Watched the Dinosaurs documentary on Netflix this weekend. Highly recommend, especially with kids.

Dinosaurs ruled this planet for 165 million years. Then an asteroid they never saw coming ended it all in a geological blink.

Put that in perspective.

Humans with abstract thinking, art, and complex language, what we'd call truly modern humans, have existed for maybe 50,000 to 100,000 years.

Civilisation as we know it? Writing, cities, organised society? Maybe 5,000 years.

The version with industrial-scale technology, global trade, and the ability to reshape the planet? Barely 200 years.

And the version with nuclear weapons, AI, and the ability to end all of it? Less than 100 years.

We solved the asteroid problem, by the way. NASA can now track and deflect them. The thing that wiped out 165 million years of dinosaurs, we've figured that one out.

And yet billions are being spent daily on war and destruction, making an already bad climate situation even worse.

The threats coming from the universe are becoming manageable. The ones we're creating ourselves, not so much.

The dinosaurs had no choice. The asteroid just came.

We do. That's what makes what's happening right now so much harder to watch.

[Hemanth @hemanth_why]

@sickdotdev It will win ✌️

[Chubby♨️ @kimmonismus]

Indian model Sarvam-105b is really really good

Sarvam AI has open-sourced two India-built reasoning models, Sarvam 30B and 105B, positioning them as globally competitive open models.

The big unlock is not just benchmark scores like 98.6 on Math500 for 105B or strong local deployment efficiency for 30B, but the full-stack story: in-house data, training, RL, tokenizer design, and inference optimization built for both frontier GPUs and consumer devices.

[Hemanth @hemanth_why]

@cskvarma If the cost of software goes down, doesn't all enterprises want more software for making them efficient??
The same software engineer builds more and gets the same pay or better pay for his/her work

[Hemanth @hemanth_why]

@cskvarma @GeminiApp Are you working on designing those electro-mech systems or writing code for them??
For coding in the electronics part, I heard claude or gtp models better??

[Hemanth @hemanth_why]

@cskvarma Your thoughts are highly valuable, maybe that's the reason you feel like AI-written
In this age of slop, human written unclear thoughts will gain traction!