Why I built this.

the story

I trade systematically and I read papers to find edges. For a long time my research process was a folder of SSRN PDFs and a rough reading plan. On a good day I got through ten or twenty pages, and half of that was background reading just to understand a paper’s premise. I once spent the better part of a month on market microstructure and orderbook convexity. I got through six papers. Two were usable, and the strategy I built on them was mediocre, because two papers is not a foundation.

When that strategy failed, the lesson was not that the idea was bad. It was that the process could not scale. The edges I wanted lived somewhere in hundreds of papers I was never going to read one at a time, and pasting abstracts into a chatbot produced fluent answers with citations that did not exist.

So I built the tool I wanted. I point it at my corpus and generate research programs, expand the ones worth expanding, trace which papers connect, and find the central ones and the niche ones I would have missed. The synthesis takes minutes. The reading I still do is deeper, because it is aimed.

Outsample is that tool, made multi-tenant so you can point it at your own library.

the name

Every quant has been burned by a backtest that looked perfect in-sample. The only test that matters is the one on data the idea has never seen. The product is named after that standard because it is built to apply the same skepticism to research: cited claims, adversarial review of its own output, and a plain statement when the evidence is not there. In-sample results lie. The name is a reminder, mostly to me.

one person

Outsample is built and run by one person, from Toronto. That means support answers come from the person who wrote the code, the changelog is real, and the product improves in the order that users actually need. It also means I will not pretend there is a sales team or an office. There is me, the corpus, and a roadmap.

the crawler

The corpus pipeline is open source. It is the same crawler I used to build the included library, and you can run it yourself on the literature you care about. Today the product takes individual paper uploads; bulk import of a crawled library is being built.

Crawler repo on GitHub: link pending (goes public before launch).

contact

Questions, bug reports, or an argument about a finding the adversarial review got wrong: [email protected]. It reaches me, not a queue.

If you read papers for a living and want to push on the product before launch, the beta list is open: