How research moves forward

Check whether someone has published it, open-sourced it, patented it—or already discovered that it rolls better as a square. A quick literature search beats rebuilding someone else's failed prototype.

See who's on the case and how many investigations are underway. Join an existing effort, compare approaches, or try something completely different. Science isn't a group project where everyone has to copy the same answer.

Sketch it, simulate it, test it, blame the boundary conditions, then actually check them. Even reusable rockets need iteration. Your first model is allowed to have a character arc.

Bring a paper, explanation, code snippet, result, replication—or evidence that your breakthrough depended on one rogue spreadsheet cell. AI can polish the explanation; evidence still gets the final say.

Record what worked, what flopped, which assumption quietly set the lab on fire (metaphorically, please), and what to try next. “Works on my machine” is not a reproducibility plan. Future you is not a reliable archive.