Paramarisha Senani

Humble servant to arcane called Deep Learning.

[Paramarisha Senani @paramarishi]

Beaten GPT-5-Pro on first shot, @ListenLabs Challenge (https://t.co/qL1QB5MOq8).

Here is my code : https://t.co/qPIMR5Sy3P

I developed it by overall my own algorithm but used Claude Code + DeepSeekV4 Flash for code writing.

[Paramarisha Senani @paramarishi]

@kanpuriyaabhau @mxtaverse I am way more literate than good for nothing folks like you. LOL!

[Paramarisha Senani @paramarishi]

@ranaharshraj7 3. Lots of params are hard-coded instead of learnt.

Will update later.

[Paramarisha Senani @paramarishi]

@ranaharshraj7 The tech used is Denoising (Noise Gate + Wiener filter) followed by WebRTC GMM rough implementation. Lot of pieces are basic or missing.

Like...
1. I do not use QFM but simple FFT. For low spec cases, QFM is better (obv).
2. GMM is laughably rudimentry.

[Paramarisha Senani @paramarishi]

@kanpuriyaabhau @mxtaverse I do not know what you mean when you say that word. Spell out or shut up.

[Paramarisha Senani @paramarishi]

@ranaharshraj7 did they allow any ai generated code? like for doing wiener filter, noise gating or rms normalization?
OR
were you supposed to write all the code by yourself?

[Paramarisha Senani @paramarishi]

@kanpuriyaabhau @mxtaverse I dont know what you idiots mean when you say reform. Its an abstract noun after all. Tell me what do you mean by it?

[Paramarisha Senani @paramarishi]

@paraschopra India can do this too....
https://t.co/H97gVwySti

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

Alibaba Qwen3.7 slowly fading into irrelevance at the frontier due to proprietary stance.
In it's place we have Minimax M3 and... *checks notes* Rio 3.5 397b, made by the municipal IT company of Rio de Janeiro's city government.

https://t.co/JgIJYVhoEi

[Paramarisha Senani @paramarishi]

@paraschopra Meet this :
Rio-3.5 from Rio IT and Telecom firm of Rio de Jenario Municipal department. It is currently beating frontier models like GPT 5.5 and DeepSeek V4 Pro.

They built it without spending even 1 M USD by combining SwiReasoning and Opensource Qwen-3.5.

[Paramarisha Senani @paramarishi]

@paraschopra https://t.co/QhHIi5Dwc1

[Paramarisha Senani @paramarishi]

@aravind True greatness comes from maximizing impact with resources you have and not whining like bitches. India can very much have a frontier class model without billions of dollars in expense by being more innovative.
Hint : Combine these : Allen's data set, SwiReasoning and GPUs it has

[Paramarisha Senani @paramarishi]

@aravind While Indians are crying like bitches, a city government funded IT and Telecom company in Rio de Jenrio, Brazil releases an optimized version of Qwen 3.5 that is beating frontier models WITHOUT spending even 1 million dollars. Meet Rio-3.5-Open-397B.

https://t.co/H97gVwySti

[𝗭𝗲𝗻 𝗠𝗮𝗴𝗻𝗲𝘁𝘀 @ZenMagnets]

Alibaba Qwen3.7 slowly fading into irrelevance at the frontier due to proprietary stance.
In it's place we have Minimax M3 and... *checks notes* Rio 3.5 397b, made by the municipal IT company of Rio de Janeiro's city government.

https://t.co/JgIJYVhoEi

[Paramarisha Senani @paramarishi]

@kimmonismus Its not a completely new model, they improved Qwen 3.5 by integrating SwiReasoning. Still, its a great work.

https://t.co/QhHIi5Dwc1

[Paramarisha Senani @paramarishi]

@krrzz231 Absolutely can. Simple bullets and explosives are not hard to make.

[Paramarisha Senani @paramarishi]

@BMTheEquityDesk Six months old news. India since then has improved its data quality by revising its GDP numbers and being more transparent.

[Paramarisha Senani @paramarishi]

@VENKATESHzeeker @kanpuriyaabhau @mxtaverse India went from 2.7 (or 27) -> 39 over 2001 to 2025.
China went from 3.5(35) -> 43 over 2001 to 2025.

[Paramarisha Senani @paramarishi]

@VENKATESHzeeker @kanpuriyaabhau @mxtaverse India's present CPI is 39 (compared to 43 of China).

So I do not see corruption in India as that much of a problem, it is sign of relatively lower economic development. Once India reaches higher economic development, this will get much better.