About
Why "Null Result"
In science, a null result means you tested something and it didn't work. Those results rarely get published, so everyone keeps rediscovering the same dead ends.
AI content has the same problem, only louder: demos that work once, benchmarks nobody can reproduce, tools that are "game-changing" until you try them on real work.
Null Result is the opposite. Each experiment starts with one question, documents the setup, the cost and the numbers, and publishes the code. If it fails, that's the result.
Ground rules
- Every claim comes with numbers and a way to reproduce it.
- Costs are real: money, hardware and hours.
- Sponsors never decide results. Sponsored content is labelled.
- Failures get published.
Contact: hello@nullresult.ai
Get each experiment by email
One experiment every two weeks: what was tested, what it cost, what actually happened.
Subscribe