| ▲ | phkahler 6 hours ago | |
>> This is the second ChatGPT shared conversation I've seen today that is truly fascinating. We recently had some bugs fixed in the geometry kernel of solvespace. Not much conversation, but the analysis from the AI was amazing: https://github.com/solvespace/solvespace/pull/1729 https://github.com/solvespace/solvespace/pull/1730 https://github.com/solvespace/solvespace/pull/1731 From the Validation section of PR 1730: "The model family was reconstructed programmatically (parameterized cuboid stack) and swept over 2,304 configurations — extrusion directions, workplane-normal orientations, sketch windings, D's plane/height/depth/extent, including all the exact-coincidence heights. Zero failures with the fix; 576 failing configurations without it. The generator is available on request." It looks like it wrote a python script to generate test cases in our file format for testing. Just... you know, as a side quest. | ||
| ▲ | versteegen 20 minutes ago | parent [-] | |
> It looks like it wrote a python script to generate test cases in our file format for testing. Just... you know, as a side quest. On the one hand, agents have done this sort of thing for a year+, if you pushed them to check their work. On the other, I absolutely can feel Fable and Sol have crossed a threshold where they can be trusted far more than before. Huge difference between plans written by Opus or Fable. Accumulated AI slop can simply be cleaned up by better models. Real cost of technical debt is shrinking due to the the inflationary devaluation of code! | ||