@w1nklerr Did anyone bother to fact check this? That YouTube video was posted 4 years ago: https://t.co/8YsY5e4DYl
The views and revenue are legit, but the story is cap. This was posted to YouTube 1.5 years BEFORE ChatGPT launched, AI had nothing to do with it.
@zostaff@horizon_trade_x Little too slow for my taste. @AlphaCIO_ai will test thousands of strategies in seconds, against multiple markets, validate it and add it to a portfolio of uncorrelated strategies.
@apoorva_mehta I built @AlphaCIO_ai which does exactly this, an end-to-end quant fund. Idea generation > Research > Validation > Optimization > Allocation. Users define a mandate once, agents run the research pipeline and find strategies that generalize on investable universe.
Been working on my auto-research platform for quant trading over the last couple of months with solid results. After seeing @karpathy drop #autoresearch and get a lot of attention, figured it might be a good time to open my app to users. DM if interested.
Crazy seeing this days after implementing the same basic process to my systematic strategy research. This is obviously orders of magnitude more broadly applicable and technical than what I've done, but the core process design is essentially identical
I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then:
- the human iterates on the prompt (.md)
- the AI agent iterates on the training code (.py)
The goal is to engineer your agents to make the fastest research progress indefinitely and without any of your own involvement. In the image, every dot is a complete LLM training run that lasts exactly 5 minutes. The agent works in an autonomous loop on a git feature branch and accumulates git commits to the training script as it finds better settings (of lower validation loss by the end) of the neural network architecture, the optimizer, all the hyperparameters, etc. You can imagine comparing the research progress of different prompts, different agents, etc.
https://t.co/YCvOwwjOzF
Part code, part sci-fi, and a pinch of psychosis :)
It's also self-correcting, there's a research feedback loop, memory, my own "Principles" (thanks Ray), and a forward monitor to quickly swap out strategies that are breaking down. Took some time to build but finally happy with the evolution
A lot of people focus on finding one strategy and trying to optimize it to beat buy and hold. My former employer said it best - the holy grail of markets is...
Built a simple interface to manage various teams or desks with different mandates and now I have myself a quant-fund-in-a-box. Next step is to hook this up to https://t.co/gTdjd07alJ for execution and that's it.
Working on converting backend to Go - I think this is the best language for agent orchestration - will try some different models like @grok, and will likely add a self-improvement agent that researches & develops its own library of signals. Otherwise keeping it simple.
These are example portfolios the system generated (won't trade live) but only cost ~ $2 to run for 30 minutes. Not using expensive models, just Gemini-2.5-flash, simple 3 agent system - Discover, Validate, Allocate. Execution will be handled by https://t.co/gTdjd07alJ. Solid ROI