Deepak Silaych

building @usesidecar
open source, gsoc 25
@iitbombay 26

[Deepak Silaych @DeepakSilaych]

1/12 (worth it)

ok it's real, I graduated from IIT Bombay and honestly when I joined I was pretty new to all of it. coding, dev, tech in general, I was kinda just figuring things out as I went.

four years later it's wild to look back. I have freelanced, taken a shot at building a startup, gotten to work with some genuinely amazing founders, and shipped way more than I ever thought I would.

I grew up a lot here, mostly by building random things, breaking them, and sitting at my desk at 3am trying to understand why something ran on my laptop and nowhere else.

so instead of a generic thank you post, lemme walk you through everything I built over these four years. the good, the messy, all of it.

[Deepak Silaych @DeepakSilaych]

@buildwithsid this is called being unmissable

[Deepak Silaych @DeepakSilaych]

currently now @SarvamAI 's only real pitch is language

Interesting product, but honestly any big AI lab can just open an India office, hire 30 people, collect data and train an India specific model with a new tokenizer on already proven architecture. That alone would beat @SarvamAI by a huge margin

But @SarvamAI isnt only a frontier model play. They r building the application layer to and that is where they become the real saviour in long term.

Wonder if govt will end up becoming their biggest customer

[Shubhangi Tiwari @ShubhangiYaps]

My 2 cents on Sarvam AI - India’s newest AI unicorn

[Deepak Silaych @DeepakSilaych]

@coolcoder56 broooo, twitter have light mode?

[Deepak Silaych @DeepakSilaych]

A chess GM shipping a full AI business before GTA 6 dropped wasn't on my 2026 list, but here we are.

Earlier, building was the hard part.
Agents made building the easy part.
Now the only thing that matters is the insight. He didnt predict the best move, he predicted the human move.

Thats the whole game, BUILDING is not the moat anymore, THINKING is.

[Vidit Gujrathi @viditchess]

Chess engines tell you the best move.

But grandmasters are human, they don’t always play it.

So I built "Kibitz": a human move predictor for chess broadcasts. I trained this model on my Nvidia RTX 5080.

Then I made it run as a business by itself.

A channel buys the overlay, Hermes onboards them, charges via @stripe test mode, runs the broadcast, narrates with @NVIDIAAI Nemotron, tracks inference cost, and books its own P&L.

I build. Hermes operates.

This is my demo and entry for the @NousResearch × @NVIDIAAI × @stripe Hermes Agent Accelerated Business Hackathon.

[Deepak Silaych @DeepakSilaych]

Today HSR saw a rainbow.
Standing on my terrace, I realized this isn't just a weather phenomenon.

It's a masterclass in B2B AI startups.

Here are 7 lessons:
1. Rainbows only appear after storms.
Just like PMF only comes after burning investor money.
2. Everyone sees the same rainbow.
Yet every startup claims it's a unique moat.
3. The rainbow is actually an optical illusion.
So is "10x engineer replaced by AI."
4. The rainbow disappears if you chase it.
Just like enterprise customers after the pilot ends.
5. There are two rainbows today.
Which reminded me to launch an AI wrapper around another AI wrapper.
6. Nature didn't call it "RainbowGPT™."
Branding matters.
7. The rainbow needed sunlight and rain.
Your startup needs OpenAI credits and a TechCrunch article.

[Claude @claudeai]

New in Claude Code: Artifacts.

Interactive pages built from your session, like a PR walkthrough or a living project dashboard, shared with your team at a private link.

Available in beta on Team and Enterprise plans.

[Deepak Silaych @DeepakSilaych]

Remember when people said AI would kill open source? 🙃

[Z.ai @Zai_org]

Introducing GLM-5.2: Frontier Intelligence, Open Weights

- Significant improvements in coding and agentic tasks
- Strong long-horizon capabilities with a 1M context window
- Two levels of reasoning effort: GLM-5.2 (max) pushes the limits, while GLM-5.2 (high) strikes a strong balance between performance and token efficiency
- MIT-licensed open weights
- Same API pricing as GLM-5.1

Tech Blog: https://t.co/LAsxUdN0JZ
Weights: https://t.co/g0A1C4UWx4
API: https://t.co/Kc3E22cbN7
Coding Plan: https://t.co/Nk8Y98HNhU
Chat: https://t.co/WCqWT0qCQb

[Yuzhen Mao @Mao_Yuzhen]

What happens when multi-agent systems stop relying on a central “controller” agent? Can agents coordinate by sharing results directly with each other?

Introducing Decentralized Language Models (DeLM): we let agents coordinate asynchronously through a shared context. Agents claim tasks from a queue and write back compact, verified results as they finish, making progress visible to all workers without requiring a main agent to merge, filter, and rebroadcast it.

New paper with @azaliamirh!

[Deepak Silaych @DeepakSilaych]

12/12 (huhhh)

thank you to every founder who took a chance on hiring a student who showed up with projects and no pedigree.

thank you to the ITC web team for letting me break things and own them.

thank you to the teammates who stayed up with me, the mentors who gave real feedback instead of just encouragement, and the friends who listened to me ramble about whatever I was building and somehow stayed interested.

IITB gave me a lot of things but the most useful one was just four years of space to build and break stuff until I figured out what I was actually good at.

if you're working on something interesting and want to talk, I'm around.

[Deepak Silaych @DeepakSilaych]

1/12 (worth it)

ok it's real, I graduated from IIT Bombay and honestly when I joined I was pretty new to all of it. coding, dev, tech in general, I was kinda just figuring things out as I went.

four years later it's wild to look back. I have freelanced, taken a shot at building a startup, gotten to work with some genuinely amazing founders, and shipped way more than I ever thought I would.

I grew up a lot here, mostly by building random things, breaking them, and sitting at my desk at 3am trying to understand why something ran on my laptop and nowhere else.

so instead of a generic thank you post, lemme walk you through everything I built over these four years. the good, the messy, all of it.

[Deepak Silaych @DeepakSilaych]

11/12

and honestly the part that still feels like redemption: in final year I got placed at BrowserStack through campus placements.

the exact same process that wouldn't even give me an interview a year earlier because of my branch and my CPI.

kinda funny how that works out. turns out all the projects and the late nights ended up mattering way more than the numbers on my transcript.

(though this offer got withdrawn later on 😅)