Tabu Search, Live
Part 9 of my Algorithms, Live series. A search algorithm that escapes dead ends by remembering them instead of rolling dice: it bans the moves it just made, even when they look best. Solve N-Queens in your browser with it here:
https://t.co/eEWetGbpyS
Why do spam emails try to trick us, in 2026, with McAfee antivirus scams?
Is my AOL account infiltrated?
Were my MSN account details leaked?
Will I receive a CD-ROM in the mail with 100 free hours of internet?
Genetic programming evolves the program itself. Watch a population of expression trees breed a formula that fits a hidden curve, live in your browser. Part 8 of Algorithms, Live.
https://t.co/GSlu9lMUM5
Differential evolution has no step-size setting. Each move is the gap between two of its own population members, so the steps shrink as the population closes in. Part 7 of Algorithms, Live.
Most optimisers crawl along a long, thin, tilted valley because every step they propose is the wrong shape. CMA-ES watches which directions keep paying off and restretches its search cloud to match. Circle, tilt, needle. Part 6 of Algorithms, Live.
In 2021 we turned Factorio into an optimisation benchmark: a real game, driven by a real optimiser, scored on what came out the other end.
You can now run that problem in your browser. Watch two algorithms lay belts across a walled grid and argue about which one is smarter.
My website is static. No server, no database, nothing running.
It still knows what I'm reading and what I played last.
A robot commits to the repository every Monday at 6am while I sleep.
In 1995 a social psychologist and an electrical engineer sat down to animate a flock of birds.
They deleted rules to see which mattered. Collision avoidance went. Velocity matching went.
What was left could not fool a pigeon. But it could find the optimum of anything.