Adi Kulkarni

CEO @ David Data | YC S26 | CS & Econ @ Yale

[Adi Kulkarni @AdiKulkarni6]

he was supposed to be working on our actual company 💀

[Alex Sima @sima_alexx]

Claude Code and Codex wasted days of my time so I ragequit. built Bullet as a passion project, a really fast coding agent.

[Adi Kulkarni @AdiKulkarni6]

@sima_alexx lol fr

[Adi Kulkarni @AdiKulkarni6]

"almost done thinking" is the new "5 minutes away"

[Adi Kulkarni @AdiKulkarni6]

do ppl still use bloomberg terminal? feels like its falling off cuz of all the new financial data providers

[Adi Kulkarni @AdiKulkarni6]

a digital salesroom for sales agents perhaps

[Alex Sima @sima_alexx]

Do you guys have this problem? Enterprise sales are so slow :( I feel like it's such a headache for a startup to try to coordinate a structured deal process for clients that is not confusing.

Is there some minimalistic "digital sales room" tool out there for startups? I would buy it.

[Adi Kulkarni @AdiKulkarni6]

agents setting up emails, api keys, accounts… they’re probs gonna be doing YC in the fall at this rate

[Adi Kulkarni @AdiKulkarni6]

anyone’s customer support inboxes getting flooded by users agents instead of actual ppl? trying to figure out how common this is

[Adi Kulkarni @AdiKulkarni6]

The model sees what a real analyst would on the day (prices, financials, earnings, estimates, news, filings), it’s point in time. For the prices, we don't simulate actual buyers and sellers. It's a model tuned to move like a real market (calm periods, volatility spikes, crashes clustering, reactions to earnings/macro), and events like earnings or shocks push the prices around.

[Adi Kulkarni @AdiKulkarni6]

Testing on real market data after the model’s knowledge cutoff could be a more accurate representation of real-world accuracy, but yeah the time window is much smaller and is only one path, while synthetic data lets us test across a much wider range of market conditions (and we validate the synthetic data on real panels).

[Adi Kulkarni @AdiKulkarni6]

LLMs are cooked in trading atm

[David Data @davidhfdata]

Frontier LLMs are impressive at many things. Trading is not one of them.

We had Fable, GPT-5.5, Gemini 3.5 Pro, and Grok 4.5 each run a portfolio for ~1.6 years on a synthetic stock market they've never seen.

All of them lost money.

Think you can beat the benchmark? Try it 👇

[Adi Kulkarni @AdiKulkarni6]

@ojaskandy @sima_alexx @davidhfdata @ycombinator the YC optical illusion 😯

[Adi Kulkarni @AdiKulkarni6]

@cha73066 @sima_alexx @davidhfdata @ycombinator Krishna's calling u out bro @sima_alexx lol

[Adi Kulkarni @AdiKulkarni6]

Superhuman AI traders won't be trained on the past. They'll be trained on every possible future.

That's what @sima_alexx and I are building at @davidhfdata.

Excited to join @ycombinator S26

[Adi Kulkarni @AdiKulkarni6]

@micahomnd @sima_alexx @davidhfdata @ycombinator Thanks Micah!

[Adi Kulkarni @AdiKulkarni6]

@ojaskandy In one world our model created, Goliath wins and David is cooked

[Adi Kulkarni @AdiKulkarni6]

@sensho @sima_alexx @davidhfdata @ycombinator we're working on it! 😄

[Adi Kulkarni @AdiKulkarni6]

@cdngdev "trust me bro" stat

[Adi Kulkarni @AdiKulkarni6]

https://t.co/i2iJmODMc3

[Adi Kulkarni @AdiKulkarni6]

15 TB for a Minecraft world? that's a war crime on grass blocks.

[Adi Kulkarni @AdiKulkarni6]

bro got portfolio mogged on an airplane💀