Building | Rocket scientist & ML Engineer | Ex-Quant | Ex @salesforce | prev @fdotinc
Joined February 2025
- Tweets334
- Following339
- Followers525
- Likes613
A programming language only LLMs can understand
It still looks like normal code, but the text is just a view. The real program is a typed graph of content-hashed definitions.
- Edits by name: 71 tokens vs a 221-token Python diff
- Breakages flagged before anything is stored
- 89% of planted bugs caught automatically
- 39/39 sandbox escapes blocked
Three models solved 60/60 tasks from the spec alone, with no training.
I have been working in a model inspired by @thinkymachines Interaction model.
Still a prototype, but allows you to achieve similar results while being able to use any model, first I have made a @OpenRouter adapter.
I will open source soon.
I have already sent an email to them so it can be fixed, but in case someone is interested, here is the code
How you can fix it:
1. Never trust output from its process
2. Isolate it: separate process, state in / actions out
3.Recompute the score from a log it can't touch
4. Allowlist operations, denylisting names gets bypassed (string fragments)
5. Reject impossible values (inf, NaN, > max)
Again, sorry to @paradigm, you have an email with the fix and more in depth in case you want to see.
Jesus retweeted
None of these AI slop startups seem to be run by finance people, they are run by LLM people who don't understand that hedge funds don't just run pattern recognition on candlesticks, and don't appreciate the instability of distributions in financial data.
This xkcd is still true
We built a hedge fund where every single trade is made by AI.
No human portfolio managers.
No manual research.
No one writing trading code.
Just a reasoning model that reads the world, thinks for itself, and trades before humans can.
Early results in testing: 2.55 Sharpe ratio. 2.79 Sortino ratio. Market neutral. $50M+ capacity.
Here's how we got here 🧵
I read the code. And it's garbage.
Spoiler, they even use gpt-4o as the model.
Let me show you what "high-frequency trading" means to them.
Another repo down
🚨BREAKING: Researchers from Stony Brook, CMU, Yale, UBC, and Fudan just open-sourced a multi-agent LLM system built specifically for high-frequency trading analysis.
It's called QuantAgent and it runs four specialized AI agents simultaneously each analyzing a different dimension of the market then synthesizes everything into a single actionable trade decision with entry points, exit points, and stop-loss thresholds.
Link in the comments + details about this 👇
This is a Flask app that fetches Yahoo Finance data, renders matplotlib charts, and asks GPT-4o to write trading opinions.
12 files. Four LLM prompts in a line. A web UI.
That's QuantAgent.