At my last job, we sold lots of hobbyist electronics stuff, including microcontrollers.
This turned out to be a little more complicated than selling, like, light bulbs. Oh how I yearned for the simplicity of a product you could plug in and have work.
Background: A microcontroller is the smallest useful computer. An ATtiny10 has a kilobyte of program memory. If you buy a thousand at a time, they cost 44 cents each.
As you'd imagine, the smallest computer has not great specs. The RAM is 32 bytes. Not gigabytes, not megabytes, not kilobytes. Individual bytes. Microcontrollers have the absolute minimum amount of hardware needed to accomplish their task, and nothing more.
This includes programming the thing. Any given MCU is programmed once, at the start of its life, and then spends the next 30 years blinking an LED on a refrigerator. Since they aren’t meant to be reflashed in the field, and modern PCs no longer expose the fast, bit-bangable ports hobbyists once used, MCUs usually need a third-party programming tool.
But you could just use that tool to install a bootloader, which then listens for a magic number on the serial bus. Then you can reprogram the chip as many times as you want without the expensive programming hardware.
There is an immediate bifurcation here. Only hobbyists will use the bootloader version. With 1024 bytes of program memory, there is, even more than usual, nothing to spare.
Consumer electronics development is a funny gig. It, more than many other businesses, requires you to be good at everything. A startup making the next Furby requires a rare omniexpertise. Your company has to write software, design hardware, create a production plan, craft a marketing scheme, and still do the boring logistics tasks of putting products in boxes and mailing them out. If you want to turn a profit, you do this the absolute minimum number of people. Ideally, one.
Proving out a brand new product requires cutting corners. You make the prototype using off the shelf hobbyist electronics. You make the next ten units with the same stuff, because there's no point in rewriting the entire codebase just for low rate initial production. You use the legacy code for the next thousand units because you're desperately busy putting out a hundred fires and hiring dozens of people to handle the tsunami of new customers. For the next ten thousand customers...
Rather by accident, my former employer found itself fulfilling the needs of the missing middle. We were an official distributor of PICAXE chips for North America. Our target market was schools, but as a sideline, we sold individual PICAXE chips, which were literally PIC chips flashed with a bootloader and a BASIC interpreter at a 200% markup. As a gag, we offered volume discounts on the chips up to a thousand units. Shortly after, we found ourselves filling multi-thousand unit orders.
We had blundered into a market niche too stupid for anyone else to fill. Our customers were tiny companies who sold prototypes hacked together from dev boards. And every time I cashed a ten thousand dollar check from these guys, I was consumed with guilt. We were selling to willing buyers at the current fair market price, but they shouldn't have been buying these products at all! Since they were using bootloaders, they had to hand program each chip individually, all while PIC would sell you programmed chips at the volume we were selling them for just ten cents extra per unit! We shouldn't have been involved at all!
But they were stuck. Translating a program from the soft and cuddly memory-managed education-oriented languages to the hardcore embedded byte counting low level languages was a rather esoteric skill. If everyone in-house is just barely keeping their heads above water responding to customer emails, and there's no budget to spend $50,000 on a consultant to rewrite your program, what do you do? Well, you keep buying hobbyist chips, that's what you do.
And I talked to these guys. All the time! They were real, functional, profitable businesses, who were giving thousands of dollars to us for no real reason. And the worst thing. The worst thing was... they didn't really care? Once every few months they would talk to their chip guy, who would make vague noises about "bootloaders" and "programming services", while they were busy solving actual problems. (How to more accurately detect deer using a trail camera with 44 cents of onboard compute) What I considered the scandal of the century was barely even perceived by my customers.
In the end my employer was killed by the pandemic, and my customers seamlessly switched to buying overpriced chips straight from the source. The end! No moral.
