There is a great post on the website The Street with Peter Semmelhack of Bug Labs, on why hardware is hard. I’ll wait here while you go read it…. Disclaimer: I worked for Bug Labs on ’09-’10. I dig Peter’s vision and experience.

Bug Labs Hardware Engineering an adventure for 2013 that played out around 2007 to 2011, part of which I was there for the last two years of. At the time, their software system was growing well and with the industry, while the Hardware/Firmware development was, frankly, a black-hole of engineering time suck.

For the time I was there, things were a bit rocky, and the realization the hardware vision was nigh impossible was a tough vision change to make. Pluggable complex engineering is absurdly hard, for engineering reasons I’ll get into below. Peter did a great job of herding the company onto a new path, It’s great to see Bug has pivoted to a place where it vision is technically possible, and is having great success.

Given that, It was a surprise to me this month to read about Project Aria and PhoneBlox. Two admirable projects that sound great as YouTube videos, but projects that are none the less destine to a die via death-by-a-million-solder-burns. It was pretty sad to see that hubris (Motorola) and lack of domain knowledge (PhoneBlox) leading well intentioned people down a rocky road.

A road to a black-hole that other companies, like Bug Labs and Open Moko, had already mapped out with giant warning signs and ‘Black Hole Ahead’ articles sharing their traumatic experience.

I could write a whole detailed post on each of the red-flags on these projects, and how it overlaps with the insurmountable challenges of Bug Labs Hardware. but I’m going to stick to bullet-points unless I get some reader-request to elaborate on some of these.

  • To ‘plug together’ a device of phone complexity, you need decades of engineering, or 2-4 years of education. These won’t be accessible outside of trained engineers, or a huge engineering NRE.
  • Ruggidizing core components while maintaining speed is impossible.The faster, hotter, and smarter you make electronics, the more finicky they are.
  • If you do ruggidize them, you need slower plug-friendly protocols which are 2 orders of magnitude too slow for modern expectations.
  • Delay: The open phone tech will always be 6 to 24 months behind the advertised *buy now, we ship next month* phones.
  • People pay a lot *not* to assemble their own PC’s. DIY PC kits and purchases are a niche of a niche, even amoung nerds and hackers. Seriously, people want to change their wallpaper and font-defaults, not their RAM or front-side-bus.

We’ve been down this road, in projects by Open Moko. Bug Labs, and more. The ‘fully configurable’ system is going to be crash-tastic. It’s great PR, it’s shit engineering.
Imagine the phone as a car-level complexity devices. Changing out the dashboard, or some plastics. or adding a spoiler to a car? Sure, a great idea. Changing out the engine, or suspension, or powertrain really takes a mechanic, or a car so de-tuned it’s worthless on a highway (which is why I loved my Vespa ET3). A few hours of thinking would make it clear the fully-pluggable system is a daydream.

I have more hope for Jolla, built by thoughtful engineers. The ‘plugable’ modules are nice-to-have add-ons and external to the core system. They are a bit isolated, and they are not core components of operation. It’s much more likely to succeed, and they can use isolation circuits and ‘firewall’ the plug-system to a few external pins/ports, rather than ruggidizing a lot of internal connections and parts.

If you want me to expand on any bullet-points, just drop a note in the comments, and I’d be glad to rant on in detail why those are such problems.

Tags: , , , ,