Jaskaran Singh

Humanity is Awesome! | Engineer

[Jaskaran Singh @jasksing]

This is aligned with my personal bias. But science is not looking to prove or finding any truth, it doesn't even know if truth exists or not. Science is just trying to measure the reality accurately.

[farbood @farbood]

God is the truth.
Science is the search for it.

[Jaskaran Singh @jasksing]

@farbood This is aligned with my personal bias.
But science is not looking to prove or finding any truth, it doesn't even know if truth exists or not.
Science is just trying to measure the reality accurately.

[Ritu Joon @ritujoon2j]

- Joined early Google (2000) 
- Rewrote Google spell-check engine 
- Built first Google Calculator
- Optimized core AdWords algorithms
- Pioneered Mixture of Experts
- Co-invented the Transformer (2017)
- Created multi-head attention mechanism
- Built Meena and LaMDA
- Quit Google over chatbot restrictions
- Co-founded breakout https://t.co/HTvc3gFwqM
- Rehired by google via $2.7B licensing deal
- Returned to co-lead Gemini 
- Now joined OpenAI(2026)

[Jaskaran Singh @jasksing]

@paraschopra Here intelligence of the humanity should be count, not individual, as in you innovate by standing on the shoulders of giant.

I guess next Einstein might require a good population scale.

[Jaskaran Singh @jasksing]

New Delhi when?

[SemiAnalysis @SemiAnalysis_]

SITUATION DETECTED: The city of Rio de Janerio has post-trained a model.

Based on Qwen 7/2, Rio 3.5 Open 397B adds SwiReasoning on top of the base Qwen model — a framework that dynamically switches between standard chain-of-thought and latent-space reasoning, guided by entropy-based confidence signals, so the model only "thinks out loud" when it needs to and otherwise reasons silently in hidden space for better token efficiency.

[Jaskaran Singh @jasksing]

Not having soverign LLMs, will repeat history.

The countries that adopted the Industrialization first in the 1900s, exploited the whole world who were late.

Soverign LLMs - does not mean simply fine-tuning on open LLMs, but having access to the whole recipe reproducible.

[Harveen Singh Chadha @HarveenChadha]

I wish I could do a podcast with people who were against building frontier models on why they didn't foresee this coming

[Jaskaran Singh @jasksing]

@erogol I think he meant less data for a particular problem.
Like learning to drive in less than 18 hours (training data).

[Gaurav Aggarwal @fooobar]

If there was ever any doubt, we cannot afford to not have our own models that compete with the best - never too late!

[Jaskaran Singh @jasksing]

@xwang_lk Diffusion

[Jaskaran Singh @jasksing]

@uvyie @shaashvats30 @agrawalmanindra If you assume the dist p1-p2 of mean 0 ( same skills of student) with some variance of like 20 (p1 could be easier than p2)
And now apply chebyshevs on it.

[Siddhartha Saxena @siddsax]

Michael Scott, the most qualified person to judge Claude vs Codex, has entered the chat.

[Jaskaran Singh @jasksing]

@drummatick TTS or STT?

[Jaskaran Singh @jasksing]

@ishowspeedsui

[Jaskaran Singh @jasksing]

Amazing!

[Dongmin Park @dongmin_park11]

Raon-OpenTTS paper is finally out! We fully open-sourced 615K hours of TTS data and a 1B model competitive with Qwen3-TTS-1B and Voxtral-TTS-4B. Like DCLM and DataComp, our work closed the gap towards SOTA closed-data models in TTS, which will help push the TTS community forward!

[Jaskaran Singh @jasksing]

@piersonmarks @GeminiApp @ElevenLabs @JellypodAI Same!

[Siddhartha Saxena @siddsax]

Anthropic onboarding day: Michael Scott introducing Karpathy like he just signed Wemby in free agency.

[Jaskaran Singh @jasksing]

The space of AI is becoming gamma contractive

[Jaskaran Singh @jasksing]

@ID_AA_Carmack This is the content I want!

[Jaskaran Singh @jasksing]

@annanay tmux new -s omnara

[Jaskaran Singh @jasksing]

@HarshalsinghCN Pretty good, now try qwen-tts and voxcpm2