About the lab
I am an independent researcher working toward artificial general intelligence. This site is the public record of that work: the questions I ask, the experiments I run, and the things that do not work.
Every experiment states a question and a hypothesis before the result, reports what happened including failures, and lists what it does not prove. Experiments get harder over time, so the archive shows the climb from fundamentals to original research.
The people behind the lab
The mark
- Loops. Each ring is one research cycle: question, experiment, learning. Every loop is wider than the last because each cycle starts from a bigger question.
- Rising nodes. Milestones climb from the first question at the centre toward general intelligence.
- Amber node. AGI, the long-term goal. The outer loop shades into amber as it converges on it.
- Ticks. Measurement. Nothing counts until it is measured.
Principles
- Failures are published. A failed experiment is a result.
- Claims stay as small as the evidence. Every experiment lists its limitations.
- Start small enough to inspect everything, then scale what is understood.
- Measure against a baseline, on data the model has not seen, with more than one seed.
How every experiment works
Question
State one thing I want to understand.
Hypothesis
Predict the answer in a way that can be wrong.
Experiment
Build the smallest test that could refute it.
Result
Record what happened, including failures.
Analysis
Separate what the data shows from what I hoped.
Learning
Write down what changed in my understanding.
Next question
Turn the surprise into the next experiment.
The next question starts the next loop.
Start with the experiments or the research articles.