I’m launching a new YouTube channel: PaperX, a companion to Vincent Codes Finance. It posts explainer videos of recent papers in finance, AI/ML, and anything at the intersection of finance and technology.
Why a new channel?
The literature I follow moves faster than I can read it. Every week there are new working papers on SSRN and arXiv that I want to understand at more than the level of the abstract, but that I won’t have time to read carefully for a while (if ever). So I started making explainer videos for myself, as a way to get the substance of a paper in about 20 to 30 minutes. They turned out to be useful enough that I decided to share them.
What’s in a PaperX video?
Each video walks through one paper at a graduate level (think MSc students in finance, math, econometrics, or statistics). It doesn’t assume you already know the paper’s specific subfield, so it defines the key concepts the first time they matter. A typical video covers:
- the motivation and background concepts,
- the data and the method,
- the main results, shown with the paper’s own tables and figures,
- a recap of the findings and of the scope and limits that the authors themselves state.
The videos are explainers, not reviews. They stick to what the paper says and attribute interpretations to the authors. Every number on screen comes with a source line pointing to the table or figure it’s from.
I ask the authors for permission before posting a video about their paper. The one exception is papers from large AI companies, which I may cover without asking first.
How the videos are made
PaperX is faceless, and the narration is an AI voice. Recording a human voiceover takes a lot of time, and using an AI voice is what makes it possible to put out these videos regularly. I will keep recording the videos on the main Vincent Codes Finance channel myself. I pick the papers; the production pipeline does most of the rest. It’s built with AI coding agents and open-source tools: the paper is extracted to text and its exhibits re-rendered with MinerU, the script and the animated scenes are drafted by agents and built with HyperFrames, and the voice is synthesized with text-to-speech. I review and approve each script before the final voiceover is generated using ElevenLabs.
Because the explanations are AI-generated, they can contain mistakes. Any errors are ours, not the authors’, so please read the paper itself (linked in every video description) before relying on a specific result. If you spot an error, let me know in the YouTube comments.
If you’re curious about building this kind of pipeline, I plan to cover parts of it on the main Vincent Codes Finance channel.
First video: Can LLMs discover novel economic theories?
The first video covers Can LLMs Discover Novel Economic Theories? by Alejandro Lopez-Lira (University of Florida), Sina Seyfi (Wilfrid Laurier University), and Yuehua Tang (University of Florida).
The authors build AutoTheory, a system that uses large language models and evolutionary search to generate, derive, code, calibrate, and verify economic theories, with LLM referees scoring the candidates that survive. They apply it to three asset-pricing puzzles. For the scaling of price multipliers, 1,000 runs produce two distinct verified equilibrium mechanisms. For the other two (the equity term structure and the currency premium puzzle), the top-scoring candidates rely on “private dials,” parameters whose only effect is on a single target. The system’s deterministic checks flag these, and the authors do not consider those puzzles solved under their current standard. A recurring message of the paper is that the verification layer is what lets the search tell genuine mechanisms apart from sophisticated curve fitting.
What’s next
I’m not dropping the original channel: Vincent Codes Finance will keep its tutorials and videos on coding for finance research, and PaperX is an addition alongside it. New PaperX videos will come out regularly, on topics ranging from machine learning in asset pricing to how AI is changing research itself. If there’s a recent paper you’d like to see explained, leave a comment on the channel. And if you find the videos useful, subscribe to PaperX to get the next ones.