Omansh Sharma

[Omansh Sharma @OmanshSharma3]

@Nolaan_boy @Online_Jester @tenderizzation it's 2029, Anthropic has dominated the market, everyone uses that and then there is this guy running an open model on their hardware as a gesture of rebellion. It's hacky and rebellious so a bullet being fired using lighter instead of a gun. What's vague in this?

[Omansh Sharma @OmanshSharma3]

@Online_Jester @tenderizzation there is nothing vague about this post

[Omansh Sharma @OmanshSharma3]

@_Bauda @rogesterone lmao true! still wouldn't call it moat though

[Omansh Sharma @OmanshSharma3]

@cuteshruti18 @OrienRex @kkspassingtime in that case OP definitely meant LLMs, not MLMs cz he said "Dude cant differentiate between MLMs and AI wrapper model." He is already referring to Dhruv's offering as AI wrapper model, and calling Sarvam's offering 'MLM' which just doesn't make sense in this context.

[Omansh Sharma @OmanshSharma3]

@rogesterone aah right. Completely missed that the guy was referring to harnesses and not the models. Thanks for clarification!

[Omansh Sharma @OmanshSharma3]

@cuteshruti18 @OrienRex @kkspassingtime Is this an offering Sarvam is working on? Never heard of it... what's the novelty? All harnesses I know of let you switch models... is it like an automatic router or something that uses a system one model?

[Omansh Sharma @OmanshSharma3]

@cuteshruti18 @OrienRex @kkspassingtime oh as in LLM that supports multiple regional languages? that makes more sense.

[Omansh Sharma @OmanshSharma3]

@OrienRex @kkspassingtime Do you mean LLM? MLM brings "Multi-level Marketting" to mind lol

[Omansh Sharma @OmanshSharma3]

@kkspassingtime at least Sarvam is trying make an innovative product. Pretty sure all of Dhruv's revenue was due to the marketting and not because the product was actually good. In long run Sarvam would definitely catch up.

[Omansh Sharma @OmanshSharma3]

@loresjpg Deepseek v4.1 Flash 💔💔

[Omansh Sharma @OmanshSharma3]

@cmichelio @AtharvaIngle7

[Omansh Sharma @OmanshSharma3]

@alokbishoyi97 would love to join

[Omansh Sharma @OmanshSharma3]

@grok @thdxr why

[Omansh Sharma @OmanshSharma3]

@lydiahallie oh thank God!

[Omansh Sharma @OmanshSharma3]

@themahis @AdamHoltererer That's just a toy project, what's important is that it's possible!

[Omansh Sharma @OmanshSharma3]

@jacsebl now make it write a 3T param LLM :P

[Omansh Sharma @OmanshSharma3]

@paraschopra bruhuh this is so cool!!

[Omansh Sharma @OmanshSharma3]

I wonder how long frontier labs will allow this 👀

[Rikiya Takehi @rikiyatakehi]

We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.

We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: https://t.co/bOWCcFhCVq