HypoForge

The blank page is the hardest part of research

A lot of research effort goes into the very first step: framing good, testable hypotheses worth investigating. It is open-ended, easy to get stuck on, and quietly shapes everything that follows. The hard part is rarely running the test, it is deciding what is worth testing in the first place.

From topic to candidate hypotheses

1. Frame. Give HypoForge a topic or dataset to think about.
2. Generate. A language model proposes candidate hypotheses worth exploring.
3. Explore. You scan the suggestions, keep the promising ones, and move faster into the real work.

A lightweight, browser-based tool

A Python and LLM core wrapped in a simple HTML and JavaScript interface, deployed as a live demo so the idea is something you can try in a browser, not just read about.

Live demo, open source

HypoForge has a working demo and is open source on GitHub. It sits at the intersection of my research background and my day job: using language models to accelerate the genuinely creative parts of technical work.