Bikram Sarkar

Infrastructure 🖥, Incident and change management. Currently learning how to automate repetitive tasks, generative ai, cloud migration.

[Akshay 🚀 @akshay_pachaar]

DeepSeek just fixed one of AI's oldest problems.

(using a 60-year-old algorithm)

Here's the story:

When deep learning took off, researchers hit a wall. You can't just stack layers endlessly. Signals either explode or vanish. Training deep networks was nearly impossible.

ResNets solved this in 2016 with residual connections:

output = input + what the layer learned

That "+" creates a direct highway for information. This is why we can now train networks with hundreds of layers.

Recently, researchers asked: what if we had multiple highways instead of one?

Hyper-Connections (HC) expanded that single lane into 4 parallel lanes with learnable matrices that mix information between streams.

The performance gains were real. But there was a problem:

Those mixing matrices compound across layers. A tiny 5% amplification per layer becomes 18x after 60 layers. The paper measured amplification reaching 3000x. Training collapses.

The usual fixes? Gradient clipping. Careful initialization. Hoping things work out.

These are hacks. And hacks don't scale.

DeepSeek went back to first principles. What mathematical constraint would guarantee stability?

The answer was sitting in a 1967 paper: the Sinkhorn-Knopp algorithm.

It forces mixing matrices to be "doubly stochastic," where rows and columns each sum to 1.

The results:

- 3000x instability reduced to 1.6x
- Stability guaranteed by math, not luck
- Only 6.7% additional training overhead

No hacks. Just math.

I've shared link to the paper in the next tweet.

[Elon Musk @elonmusk]

The Grok Singularity

[Sourav @Dutta_Souravd]

A good MF portfolio for most people:

- SBI smallcap
- Axis smallcap
- HDFC Flexicap
- PPFAS Flexicap
- Niftybees

[Sourav @Dutta_Souravd]

Netflix is fun, reading annual reports is boring.

Boring makes money. Fun kills brain cells.

[Rust Language @rustlang]

Rust 1.85.1 has been released, fixing a few regressions that accidentally slipped into Rust 1.85.0. 🐛🪲🐞

Check out the announcement for details: https://t.co/3weX01FJzo

[Rust Language @rustlang]

Rustup 1.28.0 has been released with many improvements. ✨

For those using rustup in CI: Note that rustup will no longer automatically install the toolchain from `rust-toolchain.toml`. Make sure to run `rustup toolchain install` if necessary.

Blog post: https://t.co/9mnOoWLpci

[Karan Vaidya @KaranVaidya6]

INSANE: Grok 3 + Replit Agent is a Superpower!!

Using the pair, I built an AI agent that reviews my Pull Requests

Here's the flow:
• Grok 3 generates PR-reviewing agent code using Composio + LlamaIndex
• Replit Agent automatically crafts the front-end interface
• Get a ready-to-run workflow where you paste the url to your PR, and the agent generates a code review instantly with suggestions.

Built with @grok @Replit Agent
Powered by @composio @llama_index @GroqInc

repl link: https://t.co/Xwpc4ehhnz

[DogeDesigner @cb_doge]

Grok 3 beats ChatGPT, DeepSeek and Gemini. 🦾

[Bikram Sarkar @biksar04]

Join me at Monster Scale Summit (free + virtual) for 40+ tech talks on NoSQL strategies, innovations & insights for data-intensive applications. 🎫 Register for free at https://t.co/WcmX599s4r #ScyllaDB #monsterscale

[Bikram Sarkar @biksar04]

Google pays up to rehire AI pioneer https://t.co/nIE1OOnt1z

[OpenAI Developers @OpenAIDevs]

Just 5x'd rate limits again:

o1-preview: 500 requests per minute
o1-mini: 1000 requests per minute

[Johann Rehberger @wunderwuzzi23]

🔥 Microsoft fixed a high severity data exfiltration exploit chain in Copilot that I reported earlier this year.

It was possible for a phishing mail to steal PII via prompt injection, including the contents of entire emails and other documents.

The demonstrated exploit chain consists of techniques that didn't even exist 2 years ago. 🔥

In particular, it involves:

1. Prompt Injection 💉
2. Automatic Tool Invocation (without human in loop) to bring PII into chat context ⚙️
3. ASCII Smuggling 🫣
4. Rendering of benign link + invisible text 👀
5. (Optional) Conditional instructions to only trigger when certain users view the content ☝️

Discussing two demos (stealing sales data and MFA codes), including the videos I had shared with MSRC in February.

@simonw @goodside @llm_sec

https://t.co/G0xpiwYfOZ

[nixCraft 🐧 @nixcraft]

truth

[Abhay patil @AbhayPatil100x]

Big Announcement - I will not post anymore questions asked in Interview here🥲
The reason - I have curated all questions asked in a repo, I just want it to be reposted, star the repo, and please refer it, here is the link please support and share - https://t.co/KNQvbmsCQ6

[Bikram Sarkar @biksar04]

What does it take a build a collaboration tool like Slack or MS Teams or Zoom from scratch?

[Gad Saad @GadSaad]

https://t.co/pHoLTcibp2

[Bikram Sarkar @biksar04]

@ankushdharkar Try Mattermost

[Bikram Sarkar @biksar04]

Learn DSA from Google https://t.co/e9frId42fZ

[Bikram Sarkar @biksar04]

https://t.co/RPuulZhbMv

[json @JsonBasedman]

Ok hold up, pump the brakes

Are you telling me that Lisa Su, the CEO of AMD, and Jensen Huang, the CEO of NVIDIA, are COUSINS????

I'm sorry, WHAT????