Two reviews found eight real defects in a board parity checker that had already merged. A reference pad on a net passed against a board pad with no net, holes were compared without plating, size or net, and some of the checker's own gates and controls could not fail. Four tests failed on the unchanged code before the fix.
Recurring agent failure issue in netsmith (noahdgrant)
What engineers report
A recurring agent failure problem in netsmith (noahdgrant) that mostly hits the electrical engineer.
Failure pattern: agent / false pass
Analyst note
From the Grok pain-radar digests, Nov 11-12 2026 (Nov 11 8:41 AM ET through Nov 12 5:41 AM ET digests plus the Haitch editions). Verified live 2026-11-12: issue/PR read; title and state confirmed.
What it costs a team
Modeled annual cost
$30,000
For a team of 5
Prototype delay
3 days
Modeled schedule slip per occurrence
Time lost
240 min
Per engineer per month, about 1 incidents
How the cost is modeled
240 min/month × 5 engineers × 12 months = 240 hours × $125/hour = $30,000
A modeled estimate with medium confidence. Swap in your own team size and loaded rate to size it for your team.
Engineering tools that address it
No engineering tool profile lists this error yet. Browse tools that connect to netsmith (noahdgrant).
Evidence
More recurring errors in netsmith (noahdgrant)
- On a real 12 V automotive board, four loads on the 3.3 V net had no current value before any part pick, so the regulator's output current was unset and the bead and shunt were judged on nothing, with the tool taking the first catalog row nothing refused.MCAD · impact 74
- A 2N7002 (7 ohm at 5 V Vgs) was picked for a 3.3 V drive because nothing compared the datasheet's test voltage with the design's gate drive. On a real 12 V automotive board two modem driver rows in the BOM changed to AO3400A after the check went in.ECAD · impact 66
- CI failed in this PR because the committed parity overlay was made by the old checker before another PR changed it on main.ECAD · impact 66
- It reads both boards with one reader and reports per item offsets for outline, holes and parts, with must-match tolerances. Tests fail a hole 0.1 mm off, J2 0.2 mm off, SW1 rotated 90 degrees, and two pads on each other's nets. The overlay SVG is hand-written because the project's viewer only draws schematics.ECAD · impact 66
- Review of overnight merges finds design hazards. A fuse's rating is used as the diode fault current and the comment calls that conservative, which is the opposite of what happens. A consigned part's LCSC number lands in bom.csv with nothing saying it is consigned.ECAD · impact 66
- The author wrote five footprints straight from manufacturer drawings, logging what was read, derived and assumed for each. Two differ from KiCad's stock footprints: the HRO TYPE-C-31-M-12 in KiCad cites an older drawing with signal pads 1.45 mm long where Rev A gives a different length, and the Omron B3U-1000P switch differs too. The library footprint and the current datasheet are two sources for one part, with no link telling which revision the library used.ECAD · impact 66
Sourced from public community reports. Costs and delays are modeled estimates, not measurements. Know a better source or a fix? Tell us at hello@productflo.io.