Forger Labs HQ
We build open-source tools that make AI-assisted research reproducible, grounded, and shipable.
Our mission
Most ML experiments disappear. They live in a notebook someone ran six months ago, in a Slack thread nobody can find, in a mental note that evaporated overnight. The winning idea gets shipped; everything that failed — and why it failed — is forgotten.
We think that's a problem worth solving. Not just for reproducibility, but because the rejections are where the real knowledge lives. Knowing why something didn't work is as valuable as knowing what did.
ResearchForge is our first attempt at fixing this: an autonomous research loop that tracks every hypothesis, every experiment, every rejection — grounded in literature, verifiable by anyone, shippable with a single command.
Evidence first
Every hypothesis is grounded in literature. Nothing ships without measured proof against a frozen baseline.
Open source always
The full research loop — Apache 2.0, no lock-in, no hidden APIs. Use it, fork it, build on it commercially.
AI-augmented, human-approved
Claude Code and Cursor do the heavy lifting. You approve every hypothesis, every plan, every ship.
Team
Built ResearchForge after running experiments for a $50k Kaggle competition and realising there was no good way to track what was tried, what failed, and why. The tool that didn't exist became the first product of Forger Labs HQ.
Forger Labs HQ is a very early-stage organisation — just getting started in 2026. ResearchForge is our first open-source product. If you want to build with us, reach out.