Working in something new | rocket scientist & QR | prev @fdotinc & @antlerglobal
Joined February 2025
- Tweets 453
- Following 290
- Followers 469
- Likes 645
76 Photos and videos
Pinned Tweet
Mar 10
The End of Human trading
Systems fully autonomous and capable of finding alpha and executing on this opportunities.
We are launching our SDK that can be used by builders or user of tools like Claude Code and OpenClaw.
Introducing the end of human trading.
Agents that adapt and evolve with the markets, without human intervention.
1
3
17
3,019
Jesus retweeted
Apr 16
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
Apr 15
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 🧵
21
20
513
37,293
Apr 16
The sharpe ratio is a clear overfit
The issue is using the agents for trading.
Even if you retrain, they have look ahead bias inside the weights, you cannot leave it.
Plus their nature of sampling make them unusable for trading
Apr 15
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 🧵
1
2
186
Apr 16
LLMs cannot be used for trading.
While they can simplify a lot of the work in hypothesis formulation or assist with coding, LLMs introduce risks like lookahead bias and overfitting in backtests.
Some of this issues cannot be avoided, and the newer models are worse at this.
4
2
4
241
Mar 28
If you think the $20k for Bloomberg terminal is the charts and look you should quit finance
Mar 28
OpenBB: A free alternative to the $20,000 Bloomberg Terminal
Available 100% free on GitHub:
1
6
530
Mar 28
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
Mar 28
🚨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 👇
1
3
9
2,275
Mar 28
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.
1
2
245
Mar 28
This repo has the same relationship to quantitative trading that a horoscope has to astronomy.
Next time, read the code before you post the repo.
1
203
Mar 28
Kreo slop affiliate not paying good, they want to swap for duolingo
4
298
Mar 27
Negative skew implosion incoming
Mar 27
Dream live is to have $1M and trade bonds on Polymarket
Buying at 99.9 at least once a day
That’s $360k a year, $30k a month
1
2
284
Mar 27
The core idea is great, it’s true that LLMs can I do backtesting but isn’t this just like using vectorbt with more steps?
Also by making the backtest vector using you need to convert the strategy to plug it into an exchange
Mar 27
I keep seeing people backtest with Claude/ChatGPT and it's painful to watch.
LLMs will:
→ Invent price data to satisfy your prompt
→ Look into the future
→ Ignore survivorship bias
→ Make wrong assumptions about splits & adjustments
They will make your backtest looks great but pnl will probably not follow.
I built Kwants to fix this.
It gives your LLM a real backtesting engine:
• Downloads & cleans your Polygon data automatically
• Enforces strict point-in-time data access
• Ignores late prints
• Runs fast with vectorization caching
• No hallucinations allowed at the infrastructure level
No code required. Works with any LLM.
Try it now:
kwants.dev
1
4
372