Pimp My IDE / Garage Dispatch
← Back to the garage
September 5, 2026 · Agent engineering / executable proof

A green build is not a green board.

An AI-generated circuit compiled cleanly. Under voltage, its “22 µF” capacitor delivered 11.4 µF, and the protected rail died after 0.85 ms instead of surviving 20 ms. That is not an embarrassing footnote. That is the whole product lesson.

The take: agents get serious when “looks plausible” is replaced by a loop they can execute: encode the requirement, build the artifact, push it through hostile operating corners, read the measured failure, revise. The best interface is not a shinier prompt box. It is a test bench with teeth.
Roll onto the proof bench ↓

The compiler only checks the doorway.

EEBench’s power-loss task is small enough to understand and nasty enough to matter. A residential meter loses its 5 V input. Its processor needs 20 ms to save the accumulated reading, and the protected rail must remain above a 3.0 V brownout floor. One submitted design chose a nominal 22 µF ceramic capacitor. The source built. The circuit did not.

At 4.7 V bias, EEBench reports that part providing only 11.4 µF of effective capacitance—far below the task’s 545 µF requirement. In transient simulation, the rail crossed below 3 V after 0.85 ms. The green build proved that names, types, and connections were legal. Physics supplied the rejection letter.

Generated engineering needs an executable world, not a confident paragraph.

Text wins when it carries constraints.

EEBench uses atopile rather than asking the agent to spend most of its context clicking through CAD chrome. Atopile describes itself as a language, compiler, and toolchain for electronics: declarative modules, interfaces, units, tolerances, and assertions that can update a KiCad layout. That shifts the agent’s attention from cursor coordinates toward components, connections, budgets, and failures.

The useful trick is not “hardware as code” by itself. Text can be wrong with tremendous efficiency. The win is that the same representation can be built, simulated, checked, diffed, and revised. A failed waveform becomes machine-readable steering instead of a red-faced lab surprise.

Make the bad corner the main lane.

EEBench says every specification becomes a deterministic check, using circuit graphs, BOM constraints, SPICE measurements, and worst-case component-tolerance corners. Its public scoring weights technical performance at 65% and cost efficiency at 35%, with cost credit only after the design works. That ordering is healthy: cheap smoke is still smoke.

The benchmark’s own boundary matters. Version 1 concentrates on requirements, design, and test in simulation. It does not yet establish that an agent can lay out, fabricate, bring up, and validate a complete physical product. The project says those harder layers are planned. Until then, read the score as evidence of a closed simulated loop—not a robot electrical engineer in a box.

The garage recipe.

  1. Write the requirement as a measurable limit. “Should survive” is vibes. “Rail stays above 3.0 V for 20 ms” is a gate.
  2. Separate nominal from effective. Model bias, tolerance, temperature, package, and operating point before trusting the label.
  3. Exercise the transient. Build success and steady state do not prove startup, dropout, recovery, or fault behavior.
  4. Price only working candidates. Economic optimization belongs behind the technical interlock.
  5. Keep the artifact trail. Requirement, source, part model, waveform, failing limit, revision, and final measurement travel together.

The Hacker News thread adds useful shop-floor texture but not benchmark-grade evidence. Experienced commenters describe promising boards, repairable errors, and the need for manual footprint and layout review. Treat those as field anecdotes. The stronger signal is the shape of the workflow: people are not merely chatting with a model; they are wrapping it in tools that can say no with measurements.

Interactive makeover / verification depth

Run the Physics Proof Bench.

Traditional purpose replaced: one green “build passed” badge. Better version: move the artifact through five proof stations and export the exact evidence still missing.

Proof rail

Drag the depth control or select a station. The carriage cannot teleport past the evidence; every previous gate travels with it.

1 / 5
Carriage ruleThe cyan diamond marks the current station. Green lamps behind it mean evidence retained—not quality awarded. Gray stations ahead remain untested.
22 µFnominal label
11.4 µFeffective at 4.7 V
0.85 msmeasured hold-up
20 msrequired hold-up

Verification depth

This gauge reports verification depth, not circuit quality. A deep bench can still expose a bad design—that is success.

20%Build only
What is provenThe design source builds and produces expected artifacts.
What is notElectrical behavior, operating corners, transients, and economic fitness remain unproven.
Next stationRun the nominal operating case against explicit limits.
Open the four-source proof log
[1] EEBench — circuit-design field note, sample capacitor failure, public leaderboard claims, and stated V1 limits [2] EEBench Methodology — deterministic grading, task loop, scoring weights, held-out-task provenance, and scaffold caveat [3] atopile repository — declarative electronics language, compiler/toolchain, KiCad workflow, and validation surface [4] Hacker News — exact discussion thread with practitioner anecdotes and verification concerns