On Robust Technology

2025.09.28

CXIX

[Glasses That Don’t Break, Casios & Apple Watches, Rings & Bomber Jackets, Until The End of the Internet, Software Built Like a Casio]

Thesis: The longer technology works, the more valuable it is.

[Glasses That Don’t Break]

I really like technology that doesn't break.

One of my most favored possessions is a pair of blue light blocking glasses I have been regularly wearing in the evening since early 2020. There aren't many items that I use on a daily or near daily basis that have lasted this long.

I think that the ones that have, though, are way more valuable than the ones that haven't. There's something incredible about being able to buy a thing one time and use it forever.

I want the software I write to be like that, which is a very hard thing to do. That doesn't make it a bad goal, though.

[Casios & Apple Watches]

I recently purchased a Casio watch. Since I am sharing a room with someone and I wake up early, it’s silent, vibration based alarm can wake me up without waking him up.

Not only is this is really cool, but I also would be surprised if it broke anytime soon. The battery life alone on these watches is 10 years, and I have heard stories of the watch itself lasting long enough to be passed down to children.

This watch specifically is fascinating to me because at least part of it is digital technology. Perhaps because we are so inundated with short lived digital technology, a digital Casio watch lasting ‘forever’ is somehow more surprising (but not more impressive) than when an analog Swiss watch lasts ‘forever’.

Contrast a Casio with an iPhone that cannot trigger it's alarm reliably, or an apple watch that will need to be replaced in 4 years, or the whoop I bought that they had to replace after about two years of use.

The Casio has relatively few ways it can break compared to this other technology. The biggest risk is that the battery runs out; but if it does, I can just replace it.

Alternatively, an apple watch could have the battery die or the hardware become 'incompatible' with the new apple software or have an issue with bluetooth or the touch screen or the charging port. And yeah, some of these risks are mirrored with a Casio... but you tell me what you think is more likely to break, and what you think is more likely to be usable in a decade!

More broadly, something like an apple watch is dependent on a company continuing to decide it's worth it to spend money to keep it up and running, and on me to decide I want to keep paying for apple care or get repairs; a Casio is dependent on me being able to find a replacement battery and a screw driver in 10 years.

Also, Casio’s watch making capability is 5x more lindy than apple’s!

But Noah, a Casio doesn't have strava or an odometer to record my runs! It doesn't have biometric scanners! It doesn’t have a microphone! Where’s the touchscreen?! How do I take calls on it!?

If you want those things, yes, a Casio misses the mark. And that being said, for all of it’s functionality, an apple watch really does have an impressive life expectancy. (maybe you’re just paying for it’s functionality with it’s life expectancy?)

At the present moment, I don't want any of that, though. I just want to know how many minutes I ran for, what time it is, and to wake up reliably in a silent way. In this sense, a Casio is really hard to beat.

[Rings & Bomber Jackets]

Here are some items that I use regularly that have had long useful lives, and that I expect to continue being useful for a while:

• 3 Years: PacSafe backpack. Crazy heavy use. Slash resistance must help contribute to robustness

• 4 Years: The Mac I am writing this on. One battery & keyboard replacement. Could very likely get another 4 years out of it, 8 if I boot it to nix.

• 5 Years: Blue light blocking glasses mentioned above from Ra. Got a second pair earlier this year for daytime computer use.

• 6 Years: Alpha Industries Bomber jacket. Looks cool, and is military grade. I don't wear it a lot, though, as it’s a heavy jacket. Expect it to last for a LOT longer.

• 8-9 Years: Titanium ring. On one dimension, it is great (there is no deformation of the ring itself). On another level, it has deteriorated a bit--there is a word engraved on the outside that is now only faintly visible, partly because the black finish it was set off against has come off.

• 9 years: I have a "Sony Sound Tower" in my bedroom at home, a pair of large speakers that I took to college a couple of years. Recently, a cord connecting the two of them broke, and I have yet to repair it; other than that, they have worked wonderfully for nearly a decade!

I value these objects way more than the latest coolest thing I have bought that may or may not be around 3 months from now.

If you have any objects like this that have astounded you with their longevity, please shoot me a note. I’m especially interested in long lasting clothes and pens at the moment.

[Until The End of the Internet]

The company 37Signals has a guarantee that their software will last "until the end of the internet."

If you are not a software person, it might not be immediately obvious how crazy of a commitment this is. To let you in on a little secret, a lot of software is actually dependent on other software that is dependent on other software, on and on. This means that when something else in the very long chain breaks, it can break your software, too!

As an example, if you want to do the equivalent of manipulating spreadsheets in Python, there is a good chance you use Pandas, an open source library that makes this very easy. The catch is, Pandas is just more software that other humans wrote. If they decide to change something, you have to change your code accordingly, as was the case for a lot of people when Pandas deprecated 'append'.

Hypothetically, you could just copy the Pandas code from before they made the ‘breaking change.’ However, since Pandas also has dependencies, you're now in charge of making sure that Pandas itself stays up to date with changes that other humans make!

Most software has a LOT of dependencies, so this is a bigger problem than it seems… a few years ago, a study showed the average project had 500 dependencies; some simple, reasonable math Casey Muratori did suggests that even with 100 dependencies, your code base has only a 40% chance of not breaking after one year… at 1000 dependencies, the code is almost guaranteed to break in under 8 months!

If your code is frequently breaking because of the dependencies, you have a lot less time on your hands to keep making it better.

In this way, a lot of code bases are more like an Apple Watch and less like a Casio. They are dependent on A LOT of things continuing to go right to keep working.

This is without mentioning the dependencies you could have on a company like OpenAi if you use them to get access to LLMs--what if they change the structure of the data they return or the data you need to send them to get the data back? Or, what if your product integrates with a complex product like Salesforce or HubSpot, and they decide that they change how you're supposed to do that?

Subscribe to see my writing every Sunday morning as soon as I finish it!

[Software Built Like a Casio]

I'm a very young developer--only a 3.5 year old. And, the longest project I have worked on, BirdDog, has been for only 14 months, so I would be foolish to think I know how to build truly long lasting software.

But, I’ll be damned if I don’t try to build a product that is more like a Casio and less like an apple watch. My software’s existence shouldn’t depend on a trillion dollar company to continue doing what it does; it should depend on simpler things, like the availability of batteries and screw drivers, or the internet being around.

Asides from being cautious of dependencies, another factor here is keeping the software as simple as possible. A Casio doesn’t have an odometer or a touchscreen or biometric scanners, but what it does have works like clockwork—because it is.

We have a long way to go to make BirdDog like a Casio, but that’s okay. We’re not building something to rapidly take a bunch of market share and hyper scale until someone buys it and the dependencies become their problem; rather, we’re trying to build meaningful software that robustly solves problems that will be around for decades or centuries to come.

I’ll leave you with a conversation I heard last week:

Live Deeply,