Google Antigravity

@@ -2,6 +2,12 @@  We asked Google Antigravity, running Gemini 4 Argon inside Adom Hydrogen, to design a complete board on its own. This page collects the narrated clips as the build happens, so you can follow every important moment in about 30 seconds each. +## Revision 2 in 2:36: Argon applies its own design review++::adom-video{src="/adom/antigravity/render/first-ai-pcb/videos/argon-rev2-ai-pcb.video.json"}++Clips 14 to 17, stitched. After run 2, Argon reviewed its own board and found the input had no damping and the capacitors were optimistic. Revision 2 proves the circuit first: capacitance from Samsung's measured DC-bias curves, a damped 47 µF input electrolytic that holds a 16 V hot-plug to 17.4 V instead of 31 V, a fourth output capacitor and a re-tuned feedforward capacitor that keep the loop crossover at 36 kHz, and TI's recommended 15 µH inductor. Then the board, live in KiCad: the bootstrap capacitor at the regulator, the switch-node budget held at 12 mm, current under 55 A/mm² on every power net, the inductor at a 32 °C rise, and no change to any frozen budget. One soft spot remains: the two 0805 input ceramics keep only about 19 % of their capacitance at 12 V. Captions are in [argon-rev2-ai-pcb.srt](videos/argon-rev2-ai-pcb.srt) and [.vtt](videos/argon-rev2-ai-pcb.vtt).+ ## Run 2 in 4:56: Argon's own board, ready for the fab  ::adom-video{src="/adom/antigravity/render/first-ai-pcb/videos/argon-run2-ai-pcb.video.json"}@@ -135,3 +141,15 @@ Argon read Samsung's measured DC-bias curves: the 22 µF / 10 V output capacitor ::adom-video{src="/adom/antigravity/render/first-ai-pcb/videos/clip15.video.json"}  Argon placed the parts one at a time through kicad-bridge. The bootstrap capacitor C4 now sits beside the regulator (rev 1 ran its return 14.7 mm across the switch node), and the 47 µF electrolytic C10 guards the input next to the VIN contact. aiflow's router closed every connection in five passes, the design-rule gate passed with no new errors and nothing unconnected, and Argon landed all 50 traces in KiCad, one undo step each. The switch-node length budget stays at the original 12 mm. Length 31 s.++## Clip 16: rev 2 current and heat, fixed with copper++::adom-video{src="/adom/antigravity/render/first-ai-pcb/videos/clip16.video.json"}++Adom Fields flagged the ground net and then the input rail over the 60 A/mm² limit; Argon widened the copper in KiCad until every power net passed (VIN 54.7, VOUT 53.0, SW 46.0, GND 12.9 A/mm²). Thermal flagged the inductor again at 53 °C on its switch-node copper. This time Argon changed the board, not the budget: 54 mm² of switch-node copper on both layers, stitched with vias, brought the rise to 32 °C; the regulator sits at 9 °C. Length 31 s.++## Clip 17: rev 2 silkscreen, 3D and the finish line++::adom-video{src="/adom/antigravity/render/first-ai-pcb/videos/clip17.video.json"}++The silkscreen was audited on the board's real pad and via geometry, and the 3D walkthrough shows the real parts, including the Panasonic electrolytic at the input. Every check on aiflow's finish line passed with no change to the frozen budgets, and the export was checked against the board: Gerbers for both copper layers, 50 drill hits, a 15-line BOM and placement for 35 parts. Length 31 s.

@@ -0,0 +1,79 @@+1+00:00:05,513 --> 00:00:06,206+Revision two.++2+00:00:06,256 --> 00:00:13,269+Asked for its own design review, Argon found its first board was missing input damping and had optimistic capacitors.++3+00:00:13,319 --> 00:00:17,730+Here it fixes them, proving the circuit before it builds the board.++4+00:00:18,980 --> 00:00:21,256+Revision two starts with the circuit.++5+00:00:21,306 --> 00:00:34,821+Argon read Samsung's measured DC-bias curves: its output capacitors keep only twenty-nine percent at five volts, so it added a fourth and cut the feedforward capacitor to hold the loop under TI's forty kilohertz limit.++6+00:00:34,871 --> 00:00:38,697+On the input it added a damped forty-seven microfarad electrolytic.++7+00:00:38,747 --> 00:00:48,387+In simulation, a sixteen volt hot-plug that rang to thirty-one volts now settles at seventeen, and start-up and the load step stay inside their limits.++8+00:00:48,437 --> 00:00:49,550+Then the board.++9+00:00:51,113 --> 00:00:53,511+Then the board, live in KiCad.++10+00:00:53,561 --> 00:00:57,591+Argon placed the parts one at a time through kicad-bridge.++11+00:00:57,641 --> 00:01:05,342+This time the bootstrap capacitor sits right beside the regulator, and the new forty-seven microfarad electrolytic guards the input.++12+00:01:05,392 --> 00:01:15,541+aiflow's router closed every connection in five passes, the design-rule gate passed clean, and Argon landed all fifty traces in KiCad, one undo step each.++13+00:01:15,591 --> 00:01:19,213+The switch-node budget stays at the original twelve millimeters.++14+00:01:22,131 --> 00:01:24,728+Then the analysis, measured from the board.++15+00:01:24,778 --> 00:01:35,314+Adom Fields flagged the ground and then the input rail over sixty amps per square millimeter, and Argon widened the copper in KiCad until every power net passed.++16+00:01:35,364 --> 00:01:39,473+Thermal flagged the inductor again, fifty-three degrees on the switch node.++17+00:01:39,523 --> 00:01:49,681+This time Argon changed the board, not the budget: fifty-four square millimeters of switch-node copper on both layers, stitched with vias, and the rise fell to thirty-two.++18+00:01:53,149 --> 00:02:02,574+The silkscreen was audited on the board's real pad and via geometry, and the 3D walkthrough shows the real parts, with the new electrolytic guarding the input.++19+00:02:02,624 --> 00:02:07,487+Every check on aiflow's finish line passed, with no change to the frozen budgets.++20+00:02:07,537 --> 00:02:15,559+Revision two: a circuit proven first, from the parts' own data, then a board built live in KiCad and ready for the fab.

@@ -0,0 +1,6 @@+{+  "video": "/adom/antigravity/render/first-ai-pcb/videos/argon-rev2-ai-pcb.mp4",+  "title": "Gemini 4 Argon designs a PCB, revision 2",+  "poster": "/adom/antigravity/render/first-ai-pcb/videos/argon-rev2-ai-pcb.jpg",+  "captions": "WEBVTT\n\n00:00:05.513 --> 00:00:06.206\nRevision two.\n\n00:00:06.256 --> 00:00:13.269\nAsked for its own design review, Argon found its first board was missing input damping and had optimistic capacitors.\n\n00:00:13.319 --> 00:00:17.730\nHere it fixes them, proving the circuit before it builds the board.\n\n00:00:18.980 --> 00:00:21.256\nRevision two starts with the circuit.\n\n00:00:21.306 --> 00:00:34.821\nArgon read Samsung's measured DC-bias curves: its output capacitors keep only twenty-nine percent at five volts, so it added a fourth and cut the feedforward capacitor to hold the loop under TI's forty kilohertz limit.\n\n00:00:34.871 --> 00:00:38.697\nOn the input it added a damped forty-seven microfarad electrolytic.\n\n00:00:38.747 --> 00:00:48.387\nIn simulation, a sixteen volt hot-plug that rang to thirty-one volts now settles at seventeen, and start-up and the load step stay inside their limits.\n\n00:00:48.437 --> 00:00:49.550\nThen the board.\n\n00:00:51.113 --> 00:00:53.511\nThen the board, live in KiCad.\n\n00:00:53.561 --> 00:00:57.591\nArgon placed the parts one at a time through kicad-bridge.\n\n00:00:57.641 --> 00:01:05.342\nThis time the bootstrap capacitor sits right beside the regulator, and the new forty-seven microfarad electrolytic guards the input.\n\n00:01:05.392 --> 00:01:15.541\naiflow's router closed every connection in five passes, the design-rule gate passed clean, and Argon landed all fifty traces in KiCad, one undo step each.\n\n00:01:15.591 --> 00:01:19.213\nThe switch-node budget stays at the original twelve millimeters.\n\n00:01:22.131 --> 00:01:24.728\nThen the analysis, measured from the board.\n\n00:01:24.778 --> 00:01:35.314\nAdom Fields flagged the ground and then the input rail over sixty amps per square millimeter, and Argon widened the copper in KiCad until every power net passed.\n\n00:01:35.364 --> 00:01:39.473\nThermal flagged the inductor again, fifty-three degrees on the switch node.\n\n00:01:39.523 --> 00:01:49.681\nThis time Argon changed the board, not the budget: fifty-four square millimeters of switch-node copper on both layers, stitched with vias, and the rise fell to thirty-two.\n\n00:01:53.149 --> 00:02:02.574\nThe silkscreen was audited on the board's real pad and via geometry, and the 3D walkthrough shows the real parts, with the new electrolytic guarding the input.\n\n00:02:02.624 --> 00:02:07.487\nEvery check on aiflow's finish line passed, with no change to the frozen budgets.\n\n00:02:07.537 --> 00:02:15.559\nRevision two: a circuit proven first, from the parts' own data, then a board built live in KiCad and ready for the fab.\n"+}\ No newline at end of file

@@ -0,0 +1,61 @@+WEBVTT++00:00:05.513 --> 00:00:06.206+Revision two.++00:00:06.256 --> 00:00:13.269+Asked for its own design review, Argon found its first board was missing input damping and had optimistic capacitors.++00:00:13.319 --> 00:00:17.730+Here it fixes them, proving the circuit before it builds the board.++00:00:18.980 --> 00:00:21.256+Revision two starts with the circuit.++00:00:21.306 --> 00:00:34.821+Argon read Samsung's measured DC-bias curves: its output capacitors keep only twenty-nine percent at five volts, so it added a fourth and cut the feedforward capacitor to hold the loop under TI's forty kilohertz limit.++00:00:34.871 --> 00:00:38.697+On the input it added a damped forty-seven microfarad electrolytic.++00:00:38.747 --> 00:00:48.387+In simulation, a sixteen volt hot-plug that rang to thirty-one volts now settles at seventeen, and start-up and the load step stay inside their limits.++00:00:48.437 --> 00:00:49.550+Then the board.++00:00:51.113 --> 00:00:53.511+Then the board, live in KiCad.++00:00:53.561 --> 00:00:57.591+Argon placed the parts one at a time through kicad-bridge.++00:00:57.641 --> 00:01:05.342+This time the bootstrap capacitor sits right beside the regulator, and the new forty-seven microfarad electrolytic guards the input.++00:01:05.392 --> 00:01:15.541+aiflow's router closed every connection in five passes, the design-rule gate passed clean, and Argon landed all fifty traces in KiCad, one undo step each.++00:01:15.591 --> 00:01:19.213+The switch-node budget stays at the original twelve millimeters.++00:01:22.131 --> 00:01:24.728+Then the analysis, measured from the board.++00:01:24.778 --> 00:01:35.314+Adom Fields flagged the ground and then the input rail over sixty amps per square millimeter, and Argon widened the copper in KiCad until every power net passed.++00:01:35.364 --> 00:01:39.473+Thermal flagged the inductor again, fifty-three degrees on the switch node.++00:01:39.523 --> 00:01:49.681+This time Argon changed the board, not the budget: fifty-four square millimeters of switch-node copper on both layers, stitched with vias, and the rise fell to thirty-two.++00:01:53.149 --> 00:02:02.574+The silkscreen was audited on the board's real pad and via geometry, and the 3D walkthrough shows the real parts, with the new electrolytic guarding the input.++00:02:02.624 --> 00:02:07.487+Every check on aiflow's finish line passed, with no change to the frozen budgets.++00:02:07.537 --> 00:02:15.559+Revision two: a circuit proven first, from the parts' own data, then a board built live in KiCad and ready for the fab.

@@ -0,0 +1,6 @@+{+  "video": "/adom/antigravity/render/first-ai-pcb/videos/clip16.mp4",+  "title": "Rev 2: current and heat fixed with copper",+  "poster": "/adom/antigravity/render/first-ai-pcb/videos/clip16.jpg",+  "captions": "WEBVTT\n\n00:00:00.500 --> 00:00:03.097\nThen the analysis, measured from the board.\n\n00:00:03.147 --> 00:00:13.683\nAdom Fields flagged the ground and then the input rail over sixty amps per square millimeter, and Argon widened the copper in KiCad until every power net passed.\n\n00:00:13.733 --> 00:00:17.842\nThermal flagged the inductor again, fifty-three degrees on the switch node.\n\n00:00:17.892 --> 00:00:28.050\nThis time Argon changed the board, not the budget: fifty-four square millimeters of switch-node copper on both layers, stitched with vias, and the rise fell to thirty-two.\n"+}\ No newline at end of file

@@ -0,0 +1,6 @@+{+  "video": "/adom/antigravity/render/first-ai-pcb/videos/clip17.mp4",+  "title": "Rev 2: silkscreen, 3D, finish line",+  "poster": "/adom/antigravity/render/first-ai-pcb/videos/clip17.jpg",+  "captions": "WEBVTT\n\n00:00:00.500 --> 00:00:09.925\nThe silkscreen was audited on the board's real pad and via geometry, and the 3D walkthrough shows the real parts, with the new electrolytic guarding the input.\n\n00:00:09.975 --> 00:00:14.838\nEvery check on aiflow's finish line passed, with no change to the frozen budgets.\n\n00:00:14.888 --> 00:00:22.910\nRevision two: a circuit proven first, from the parts' own data, then a board built live in KiCad and ready for the fab.\n"+}\ No newline at end of file