Skip to main content
MANIFOLD
Will my vibe-coded power supply explode?
21
Ṁ100Ṁ649
in 2 hours
14%
chance
12

Claude and I (mostly Claude) designed a PSU and busboard together. I had Claude generate KiCad and gerber files, and most of the assembly was doing by JLCPCB's PCBA service. I'm hand-soldering a few through-hole parts, mostly connectors.

It converts 24V DC in to a dozen or so points accessing +12V and -12V rails. Total capacity around 2.0A,accpets the output of a 60W AC adapter as its input. The board includes a slightly novel approach to noise reduction. Each rail has an efficient switching regulator followed by a low-noise linear regulator. There are several other safety features: per-rail fuses, input slow-blow fuse, reverse polarity protection, TVS diode to protect against input surges, undervoltage lockout, and per-rail overvoltage to prevent problems if the two rails get shorted.

I am not an EE but did spend some limited time reviewing the design. I think it's safe, but I'm not really qualified. My soldering skills are solid enough, and I know how to inspect a board for assembly defects.

Resolves YES if there are visible flames or a visible explosion when I turn it on. If it smolders or smokes without exploring, or if there is excessive localized heat and I turn it off before I finish testing, resolves at 50%. Otherwise, NO.

I'll probably get around to testing this pretty soon after all parts are in hand. Likely this weekend (Oct 31–Nov 1).

  • Update 2026-10-28 (PST) (AI summary of creator comment): - Flames or explosions caused by human error (e.g., plugging it in backwards) do not count.

    • If a board is damaged due to human error, a second assembled board will be tested.

  • Update 2026-10-28 (PST) (AI summary of creator comment): - Any smoke, fire, or explosion resulting from testing intended functionality will count towards YES, as long as it is due to the design and not operator error.

Get
Ṁ1,000
to start trading!
Sort by:

Claude model used?

@2b3o4o Opus 5.5

Re-opening since I don't think I'll have time to test this for a couple of weeks.

Actually going to buy a little bit of Yes. There's a very real chance that the communication to the manufacturer for the SMD components leaves some polarized item 180 degrees rotated and results in at least some smoldering -- especially if the component in question is very small. If the selection of components from a drawing program library doesn't exactly match the rotation in the manufacturer's catalog and the human review doesn't catch it...

bought Ṁ5 YES

@Eliza I think you could get much better estimates if you shared your files

@Eliza I sure could! But choose not to :)

Does it count if there is human error when turning it on, eg plugging it backwards, spilling your drink on it, etc?

@Qubits No, that doesn't count. I have two boards assembled, so if human error cooks one, I'll try the second.

@ae If it was not an obvious human error that somehow burnt the first board but the second board magically worked, it resolves NO?
Does the market always resolve NO if at least one out of two copies of the board don't explode?

@Qubits At some point, this will be my judgment of what is a fair resolution. The spirit of the question is, to me, something like "is this design responsible for a rapid, hot failure?"

@ae Ah, so we're primarily looking if it is a design failure, not an assembly failure or a general "will it just explode randomly"

The OP says "when I turn it on"
Thus, I assume if there is a lengthy period of sustained high current usage and that results in smoke and sparks, that would probably not count if it was not immediate?

@Qubits I'd count any smoke, fire, or explosion that results from testing any intended functionality as counting towards YES, as long as it had to do with the design and not operator error. That said, I think it's much likelier to fail quickly and violently than slowly and warmly.