Back in 2001 I was working on an app used in planning optimal routes for city garbage trucks. We quickly found out that this was equivalent to the Traveling Salesman Problem and the initial brute force solution was not fast enough to get results in useful time.

A colleague stumbled upon a paper on Ant Colony Systems and gave it to me to try to implement it and to check if the results were as good as the paper promised.

It was amazing how the solutions were so close to optimal ( we checked the results from the ACS algorithm against the brute-force solutions that took hours to compute ) and how fast the algorithm stabilized on the solutions.

It was also the first time I had used an algorithm that does not really stop, it just converges to a solution and another heuristic decides when to accept a final result.

I learned a lot from it.