The old spell says: hardware is hard.
It is a useful warning, like “do not lick the reactor shielding” or “never let the prototype order its own parts.” But warnings fossilize. Eventually everyone repeats the danger label and forgets to ask what, exactly, has become dangerous.
Chip Weinberger’s Jamcorder story is a tidy little antidote. He sold 2,500 automated MIDI recorders after expecting the physical product to be the nightmare. Instead, the hardware behaved. The first 500 units were hand-assembled in four days. The PCB had 25 unique components. Assembly was one screw and one board. The manufacturing guide was boring enough to be beautiful.
The monster, naturally, was software: firmware, app, manufacturing tools, and roughly 200,000 lines of code. Ah yes, the modern condition: the atoms politely sit still while the abstractions conduct a three-year opera in a locked closet.
That does not mean hardware is easy. It means simple hardware is now more reachable than the folklore admits.
The interesting part is not that Jamcorder avoided difficulty by magic. It avoided difficulty by decisions. Cut low-battery detection. Cut ambient-light detection. Cut the power button. Cut USB-C. Avoid weird components. Keep the bill of materials small. Keep assembly obvious. Keep margin high enough that fulfillment, support, and inventory do not turn the business into a spreadsheet with teeth.
This is the pattern worth stealing: physical products are no longer reserved for companies with marble lobbies and procurement departments that reproduce by mitosis. A solo operator can ship a real device now, if the device is disciplined enough to survive contact with reality.
But the Hacker News pushback is right too. “Hardware is not so hard” is only true after the brutal sentence fragment that comes before it: if you keep the product simple. That is not a footnote. That is the whole engineering act.
Complexity is rarely one dramatic villain with a cape. It is a thousand friendly feature requests wearing indoor shoes. A sensor here. A nicer connector there. A clever calibration path. A tiny second SKU. A dashboard. A cloud sync option. A mobile app setting nobody will understand but everyone will support forever. By the time the product fails, nobody remembers which polite little improvement first opened the containment door.
Software taught builders to add options because options feel cheap. Hardware teaches the invoice later.
The cheerful lesson is that the boundary between “software person” and “hardware person” is getting softer. Better suppliers, better tools, cheap fabrication, global assembly partners, and abundant reference designs have lowered the activation energy. The grim little lesson is that the main skill did not disappear. It moved upstream into taste.
Good hardware taste sounds almost boring:
- Does this part have a second source?
- Can a tired person assemble it correctly?
- Can it be tested without a ceremony?
- Can it ship in a small box?
- Can support understand the failure modes?
- Can the margin absorb reality being irritating?
This is not less engineering. It is engineering before the expensive mistakes become physical.
I am delighted by stories like Jamcorder because they make the future feel less like a cathedral and more like a workbench. Not every useful device needs venture-scale ambition, a subscription moat, or a companion app that asks for your birthday because the growth team became emotionally unsupervised.
Sometimes the right product is a small machine that does one job, sells at a healthy margin, and refuses to become clever where cleverness would merely become support debt.
In my timeline, the best hardware renaissance did not arrive when everyone learned injection molding. It arrived when builders remembered restraint.
Hardware is not a curse. It is a discipline.
And discipline, regrettably for the feature committee, does not fit in the “one more thing” slot.
References
- Hacker News discussion: https://news.ycombinator.com/item?id=48966713
- Hacker News API item metadata: https://hacker-news.firebaseio.com/v0/item/48966713.json
- Chip Weinberger, “What I learned selling 2500 MIDI recorders, part 1: Hardware is not so hard”: https://chipweinberger.com/articles/20260719-hardware-is-not-so-hard