Knowing when to build it twice

I built a parking lot layout tool in May. In July, I built it again.

Nothing about the first one was actually broken. It read a boundary and laid out stalls, drive aisles, and ADA spaces, live preview and all, and mostly it did that job well. Tight corners were the part I never quite trusted. A boundary that pinched into a sharp angle could leave a sliver of pavement too small to use, or a row of stalls packed unevenly against one edge, and fixing that cleanly is a harder geometry problem than the rest of the tool.

Between May and July, the model I was building with got meaningfully better. So I went back and gave it the one part I had left alone the first time.

The second build kept everything that already worked, the same boundary pick, the same live preview, and added real handling for those corners: carving out the awkward slivers on purpose, clipping and filleting the polygon properly, centering the interior rows instead of packing them from one edge. Same tool, same job, and the part that had been the hard part stopped being the hard part.

I don't think every rebuild needs a better model as the reason. Most of the time the reason is that the first version taught you what you actually needed. This one was different. The tool didn't get built twice because it was wrong. It got built twice because I finally had something capable enough to fix the part I already knew was rough.

Be Better - go back for the part you left alone the first time.