Practical quant resources on X

A small corner of X has turned into a workshop for aspiring quants. The posts are not polished courses or peer‑reviewed papers. They are threads. But they are unusually concrete. One thread from Quant Science walks through how to build a “mini” hedge fund in Python. Another breaks down a first trading bot. A third points readers toward Claude prompt patterns for finance work. The appeal is obvious. These are not abstract lectures about alpha. They are project blueprints that can be copied into a notebook, pushed to GitHub, and shown in an interview (threadreaderapp.com, unrollnow.com, anthropic.com). What makes the threads travel is that they sit on top of a real change in the tooling. Five years ago, a beginner trying to build a quant project had to stitch together data, backtests, execution, and reporting by hand. Now the open web is full of frameworks that promise to collapse that work into a starter stack. Quant Science’s own ecosystem markets a “Hedge Fund in a Box” workflow for retail traders learning Python, and its public material keeps pointing readers toward the same architecture: data layer, research layer, portfolio logic, and execution (quantscience.io, youtube.com). The most important fact here is not that the branding is slick. It is that the architecture is legible enough for newcomers to imitate. That is why the trading‑bot thread lands. It points readers to the Investing Algorithm Framework, an open source Python library that explicitly promises strategy development, backtesting, data management, order execution, analytics, and deployment in one place (github.com, coding-kitties.github.io, pypi.org). On April 5, 2026, the package was still actively shipping releases on PyPI, which matters because half the problem with “learn quant on the internet” has always been dead repos and stale tutorials (pypi.org). A thread that says “here is the bot, here is the library, now run it” is crude, but it clears the biggest beginner hurdle. It gives the project a spine. The Claude angle is newer, and more slippery. There is now a growing market for prompts that claim to turn an LLM into a research analyst, code generator, or stat‑arb assistant. Some of that is fluff. But the underlying direction is real. Anthropic has spent the past year pushing Claude deeper into finance workflows, first with a financial analysis product in July 2025 and then with added Excel and market‑data integrations in October 2025 (anthropic.com, anthropic.com). Its own documentation now emphasizes structured prompts, XML tags, tool use, and agentic systems rather than one‑shot magic commands (platform.claude.com, anthropic.com). So when X users share “prompt recipes” for quant work, the useful part is not the recipe itself. It is the idea that prompting has become another layer of system design. That shift is easiest to see in the projects these threads keep citing. TradingAgents, for example, is an open framework and research paper built around specialized LLM agents that play the roles of analysts, traders, and risk managers inside a simulated trading firm (github.com, arxiv.org, tauricresearch.github.io). It is a research toy, not a live fund. The paper’s reported returns do not mean a student can prompt Claude into free money. But as a portfolio project, it is almost perfect. It lets a candidate show data plumbing, model orchestration, evaluation logic, and enough market structure to sound like they know what a risk desk is for. That is why these X threads matter. They are not selling edge. They are selling scaffolding, one copied code block at a time.