Toadoum

Move fast AND rigorously through the full AI/ML research loop: framing a testable hypothesis, reviewing the literature, reproducing a baseline, designing leak-free experiments, tracking runs, analyzing results honestly, and writing up. Use this whenever the user is doing machine-learning or AI research — reading or comparing papers, planning or running experiments or ablations, reproducing a result, debugging why a metric looks suspiciously good, choosing seeds/baselines, or drafting a paper or related-work section — even if they never say the word "research". Includes a self-improving lesson log: whenever you catch and correct a mistake, you record it so the skill gets better every time it runs.