Research
Tests of AI agents, retrieval and quant models on real financial data, with the numbers and the method behind every result.
I tested Vibe-Trading, the open-source AI trading agent with 37,000 GitHub stars
An open-source agent with 90+ finance skills that analyses stocks, codes and runs its own backtests and convenes a four-agent investment committee, for 11 cents to $2.80 a run. It's an interesting tool, though some features are still limited and its backtests have small timing problems.
Nov 2, 2026 · 7 min read
New DeepSeek model competes with Claude and GPT-6.1 while costing 90% less: a “deep” analysis
DeepSeek V4.1 Flash 1017 reports 77% on DeepSWE for about $0.20 a task. Independent boards put it 14th of 50 on the Vals Index and 17th of 80 on finance, ahead of GPT-6.1 Sol, at $1.20 per million output tokens against $10.
Oct 19, 2026 · 9 min read
I tested AI Hedge Fund, the open-source repo with 64,000 GitHub stars
How the multi-agent fund turns five investor agents and a quant model into a portfolio, the math behind each step, and a live paper run on DeepSeek where one weight in a strategy file decided two of five stocks.
Oct 13, 2026 · 7 min read
GPT-6.1 Sol is OpenAI's best deal, but it disappoints in finance, ranking 30th of 72
OpenAI's newest model is the best value on Vals's general index, 91% of the top score for 15% of the cost. On Vals's finance benchmark it ranks 30th of 72, and 17 cheaper systems beat it.
Sep 30, 2026 · 6 min read
Is Jev efficient for RAG? A test on 10-Ks from Apple, Microsoft, Nvidia and Amazon
A decision model that only returns numbers found the evidence for all 50 analyst questions, with one LLM call per answer and about 100 times fewer tokens than an agentic pipeline.
Sep 27, 2026 · 10 min read
A 12-agent AI investment committee for $0.09. Here's what it decided.
10 AI agents build the case, 2 decide, and neither judge ever reads the data. Inside GitHub's most starred trading project (108,000 stars), tested over 68 runs with every log published.
Sep 24, 2026 · 7 min read
AI finance skills are limited by design. Build your own.
Anthropic ships seventeen finance plugins. Daloopa ships twenty-three skills with 487 stars behind them. Reading the code shows where each one stops, and both packages hardcode the same number.
Sep 21, 2026 · 9 min read
Machine Learning and Deep Learning in Finance: A Complete Guide to 10 Models
Trees, neural networks and real market data, with the code for every one. Three of the ten lost to a baseline that costs nothing, and that result stays in.
Sep 9, 2026 · 15 min read
