# shibo - @Seve If I want to make changes to the ... Channel: #contributor Source: https://discord.com/channels/1233487248129921135/1530599881058681085 Started: 2026-07-25T15:37:29.317000+00:00 Last activity: 2026-08-07T16:49:03.571000+00:00 ## shibo — 2026-07-25T15:37:28.815000+00:00 <@757706909351411845> If I want to make changes to the analog simulations, do I PR to the previous RFC or make a new one? ## Seve — 2026-07-25T15:38:02.037000+00:00 Either way is common! ## Seve — 2026-07-25T15:38:17.307000+00:00 Usually if the rfc is new, edit ## shibo — 2026-07-25T18:04:22.778000+00:00 <@757706909351411845> https://github.com/tscircuit/rfc/pull/11 ## Seve — 2026-07-25T20:13:14.267000+00:00 left some comments ## shibo — 2026-07-31T14:31:37.223000+00:00 <@757706909351411845> I split the fix for sample 4 to two parts, I understood part 1 and make the visualization clear, part 1 fixed sample 4's static reachability issue but it doesn't pass yet, the PR increased dataset 23 DRC completion by 48% https://github.com/tscircuit/tscircuit-autorouter/pull/1799 ## shibo — 2026-07-31T14:32:51.279000+00:00 the issue is that we failed to detect bga shapes if the component has other pads outside the bga pattern ## shibo — 2026-07-31T14:32:57.019000+00:00 the PR body explains further ## Seve — 2026-07-31T14:53:09.978000+00:00 Sg! ## shibo — 2026-08-01T14:42:26.711000+00:00 <@757706909351411845> fix in repair solver 03, improves DRC pass rate for the kicad dataset 24 from 0% to 20%, I noticed no regressions: https://github.com/tscircuit/tscircuit-autorouter/pull/1810 ## Seve — 2026-08-01T17:39:48.817000+00:00 cool ## Seve — 2026-08-01T17:39:50.613000+00:00 nice ## shibo — 2026-08-04T16:46:35.262000+00:00 <@757706909351411845> new sample from OpenAI's dataset passes DRC: https://github.com/tscircuit/tscircuit-autorouter/pull/1902 ## shibo — 2026-08-04T16:46:45.881000+00:00 repair 03 fix ## shibo — 2026-08-05T16:57:23.096000+00:00 <@757706909351411845> new sample pass DRC in both OpenAI and kicad datasets: https://github.com/tscircuit/tscircuit-autorouter/pull/1927 ## shibo — 2026-08-05T16:57:54.725000+00:00 DRC rate is 25% and 30% respectively ## Seve — 2026-08-05T16:58:16.475000+00:00 *partner datasets 😉 ## Seve — 2026-08-05T17:00:17.342000+00:00 yea looks great ## shibo — 2026-08-05T17:00:36.969000+00:00 does that mean now we pass DRC for 1/4 of all boards in the world? 😂 ## shibo — 2026-08-05T17:01:19.766000+00:00 now I want to add a kicad dataset that has 10 of the hardest boards from kicad, just to humble myself ## shibo — 2026-08-05T17:03:19.556000+00:00 are kicad boards a good sample of boards used in the real world? or are the hard boards for huge companies that do advanced stuff always private? ## shibo — 2026-08-05T18:01:40.696000+00:00 <@757706909351411845> wdyt? ## Seve — 2026-08-05T19:09:09.162000+00:00 🤷 ## Seve — 2026-08-05T19:09:19.694000+00:00 kicad isn't used as much by professional companies ## Seve — 2026-08-05T19:09:22.765000+00:00 Altium is more used ## Seve — 2026-08-05T19:09:28.490000+00:00 the TI reference designs are all altium ## shibo — 2026-08-05T20:40:29.397000+00:00 good to know ## shibo — 2026-08-06T04:25:55.326000+00:00 <@757706909351411845> repair 03 fix, new sample passes DRC: https://github.com/tscircuit/tscircuit-autorouter/pull/1937 ## shibo — 2026-08-06T09:42:12.444000+00:00 <@757706909351411845> you said how could we know it's not just from increasing iterations, i kept only the iteration increase and nothing was fix, we also have a repro in the repair 03 pr that proves the fix ## shibo — 2026-08-06T16:18:09.925000+00:00 <@757706909351411845> new sample passes DRC: https://github.com/tscircuit/tscircuit-autorouter/pull/1941 ## shibo — 2026-08-07T14:39:01.568000+00:00 <@757706909351411845> no new samples pass but tens of DRC issues are fixed: https://github.com/tscircuit/tscircuit-autorouter/pull/1968 ## shibo — 2026-08-07T16:39:28.932000+00:00 <@757706909351411845> fixed another issue, again no new samples pass: https://github.com/tscircuit/tscircuit-autorouter/pull/1950 ## Seve — 2026-08-07T16:49:03.571000+00:00 Nice