Drawing on 2,500 units of real sales, Weinberger argues that modern tooling (KiCad, JLCPCB, LCSC, Shenzhen contract assembly) has collapsed the cost and lead time of small-run hardware to the point where a solo developer can go schematic-to-shelf for under $50,000. He frames the persistent 'hardware is hard' narrative as a decade out of date and actively harmful because it deters capable engineers from shipping products people would pay for.
The editorial synthesis extends Weinberger's argument by framing services like JLCPCB, LCSC, and Shenzhen contract manufacturers as a commoditized hardware platform layer analogous to how Vercel and Stripe commoditized web infrastructure. The claim is that this platformization — populated PCBs in two weeks for a few hundred dollars, in-stock passives, turnkey assembly — has structurally moved hardware into the same risk bucket as an indie SaaS.
Chip Weinberger, an indie hardware maker, published a postmortem this week on selling roughly 2,500 units of a MIDI recorder — a small standalone device that captures MIDI data from synths and controllers without a computer in the loop. The essay isn't a victory lap. It's a methodical breakdown of every stage of the process, priced in dollars and calendar weeks, aimed squarely at the software developers he thinks are needlessly scared of physical products.
His thesis is blunt: the 'hardware is hard' meme is a decade out of date, and it's keeping capable engineers from shipping things people would actually pay for. He walks through schematic capture in KiCad, PCB fabrication at JLCPCB and PCBWay, sourcing on LCSC and Digi-Key, contract assembly in Shenzhen, FCC/CE certification, fulfillment via a US warehouse, and the long tail of email support. Each stage gets a concrete number: what he paid, how long it took, and where he got it wrong the first time.
The headline claim is that a solo developer, in 2026, can go from schematic to a boxed product on a shelf for well under $50,000 in tooling and inventory, with unit economics that work at hundreds of units instead of hundreds of thousands. That threshold matters. It puts hardware in the same risk bucket as building an indie SaaS.
For most of the last twenty years, the received wisdom in Y Combinator batches and Hacker News threads has been that hardware is where startups go to die. Long lead times, tied-up inventory, brutal margins, injection molding minimums in the tens of thousands, distributor politics. All of that was true. Most of it isn't anymore.
The stack Weinberger describes is basically a hardware equivalent of Vercel and Stripe. JLCPCB will fab and assemble a small run of PCBs, with 90% of the components populated for you, in about two weeks for a few hundred dollars. LCSC keeps most of the passives and common ICs in stock. Contract manufacturers in Shenzhen — the same ones that stamp out $30M of white-label AliExpress product — will happily take a 500-unit order for a hobbyist SKU because their marginal cost of running your job is small. FCC pre-scan is a $2-3k afternoon at a lab in the US; full certification is another five figures but only when you're serious.
What's changed isn't the physics. It's that the intermediaries who used to charge you $50k in setup fees have been disintermediated by web forms that accept a Gerber file and a PayPal payment. The tooling on the design side has followed. KiCad is now genuinely competitive with Altium for boards of the complexity most makers care about. ESP32 and RP2040 modules mean you rarely need a custom ARM SoC. Rust is starting to get a real embedded story via `embassy`, and if you don't want that, MicroPython or CircuitPython let you iterate on device behavior at software speeds.
The community response on Hacker News (466 points, hundreds of comments) split predictably. The people cheering are usually software engineers who tried a small hardware project and were surprised how far they got. The dissenters — and they have a point — are people who scaled past the first thousand units and hit the wall Weinberger only glances at: the support surface of a physical product is roughly monotonic and doesn't get overtaken by a new release the way SaaS bugs do. A unit sold in 2022 is still someone's problem in 2026 if the USB stack has a race condition.
That asymmetry — cheap to enter, expensive to maintain — is the real update to the folk wisdom. 'Hardware is hard' used to mean 'you'll die before you ship one.' Now it means 'shipping is easy; the five years afterward are what you underestimated.'
If you've been sitting on a hardware side project, the practical implication is that the excuses have thinned out. A weekend spent in KiCad, a $200 board order, and a couple of ESP32-S3 dev modules will get you to a functional prototype. From there to a boxed product for sale is a matter of iteration, not a capital raise. Weinberger's numbers suggest the total cost of a first production run — 100-500 units with certifications, packaging, and a landing page — sits in the $15-30k range for a modest device. That's a used car, not a Series A.
But the essay's second, quieter lesson is where the actual leverage lives. The engineering work that pays off is firmware, not the PCB. The board is a commodity substrate for the code that runs on it, and the code is where every warranty return, every one-star review, and every 'won't enumerate on macOS 26' bug report originates. If you're a software developer eyeing hardware, your competitive advantage isn't going to be a cleverer schematic — it's that you can actually ship a firmware update that doesn't brick the device, write a driver that works across three operating systems, and treat the embedded C (or Rust) with the same rigor you'd apply to a production backend.
The corollary is a warning: hardware punishes 'move fast and break things' more than any web stack. You cannot ship a hotfix to a device sitting in a drawer with no wifi. Every design decision — connector choice, bootloader design, telemetry — either buys you or costs you optionality over a five-year window. Design for updateability from day one or accept that your v1 units are permanent liabilities.
The interesting question isn't whether more software developers will start shipping hardware — Weinberger's data says they already are, quietly, in numbers that don't hit TechCrunch. The question is what happens to the middle of the market. The $50-300 gadget category has historically been owned by white-label AliExpress product with atrocious firmware and no support. If a growing pool of software-literate makers can hit the same BOM cost and layer real engineering on top, that segment is genuinely up for grabs for the first time in years. The MIDI recorder is a niche example. The pattern generalizes to any category where the incumbents shipped hardware that works and firmware that doesn't. Which, if you've bought a smart plug in the last decade, is most of them.
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.