OSD 229: Iterative innovation in the regimes of bits, atoms, and regulation
The big costs of small hurdles.
We saw a fun thread this week about 3D-printed silencers:
Product development is an iterative process where each cycle builds on the gains of previous cycles. There’s another, more widely used, term for that: compounding.
Conveniently, there’s a simple formula for compound gains, which means that you can (very roughly) model how your product development speed will change as a function of how long your iteration cycles take.
The formula is:
Final value = Starting value * (r + 1)^n
Where:
Final value = how much you end up with
Starting value = how much you started with
r = growth/improvement rate per time period
n = number of time periods elapsed
So to abuse that a bit by applying it to product development, say Starting value is your initial product “goodness”, and we’ll set that to 1. The growth rate, r, is how much better your product gets with each iteration, and n is the number of iterations you’ve done.
So if your product gets 2% better with each iteration and you do 12 iterations in a year, it’ll be 1.27x better after a year. If you only manage 1 iteration per year, you’ll only make it 1.02x better in a year. That 0.25x improvement sounds small, but remember we’re compounding. Play this same dynamic out over ten years and the difference is a little bigger.
There’s a bigger difference if your improvement-per-iteration is better. Here’s what it looks like with 5% improvement instead of 2%.
You can play with the numbers to make the differences bigger or smaller, but the point is that there are nonlinear returns to moving faster. (Side note, the original version of this article used the wrong math for the section above. That’s now fixed.)
Ok, hold that thought for a second.
Some fields in engineering have to deal with different physics rulesets in different circumstances. In physics, there’s classical Newtonian physics (for someone throwing a football) vs. relativistic physics (for stars and nuclear explosions) vs. quantum physics (for subatomic particles). In aircraft, there’s subsonic flight vs. transonic vs. supersonic vs. hypersonic. The rules and behaviors are different in these different circumstances.
These are called regimes. So an airplane designer might talk about designing for the subsonic regime vs. the hypersonic regime. Or a physics teacher might tell you how the rules in the classical regime don’t apply in the quantum regime.
Similarly, you can break the gun world into three regimes:
The bits regime. This is the world of software, and it moves fast. Iteration speed is gated only on your own competence.
The atoms regime. This is the space of physical products, so naturally it’s where most of gun land lives. Iterations are slower here and are gated on things like manufacturing processes, supply chain issues, and fulfillment logistics. Those ares are themselves hotbeds of innovation though, which is good.
The regulatory regime. This is where silencers live. Iteration speed isn’t gated on creators or customers at all. Instead the limiting factor is a government process that’s unaccountable to (and in practice actually hostile to) creators and customers.
There are two ways to increase the pace of innovation. The first is to optimize for the regime you’re in. Often that’s the only choice — e.g. if you make plate carriers, those are never going to become software. So the smart move is to become an expert in the atoms regime.
But the second way to speed up innovation is to move your product up the stack of regimes. That’s why Ivan’s R&D is cool — it skips silencers from the regulatory regime straight to the bits regime. Instead of the iteration loop being limited by the ATF’s ability to process forms, it’s limited by how quickly creators can come up with better designs.
The more of that we see, the better. Here’s to bits eating atoms and atoms eating regulation.
Additional reading on this topic:
This week’s links
Technical medical lecture about gunshot wounds
From an OSD Discord user: “An old medical lecture about gunshot wounds. Unbiased and informative. Actually touches on injury/ death stats at the beginning too (car vs gun).“
Another Discord user linked this medical lecture from Dr. John Hinds, a trauma physician who became known as the “flying doctor” for trailing motorcycle racers at the Isle of Man TT and being the first on the scene when there was a crash. Dr. Hinds was unfortunately himself killed in a motorcycle crash at a 2015 race weekend. His lecture above isn’t gun-related, but it’s interesting if you like trauma medicine.
“Secure radio communications”
And another great one from a Discord user: “Reasonable, civil and interesting video on the topic of radios for civilians with tactical considerations.”
“Is there a g in gunslinger? Cognitive predictors of firearms proficiency.”
Study finding that intelligence is somewhat predictive of pistol-shooting skill.
Read this with the usual disclaimers that you should be skeptical of one-off social science studies, especially those with a clickbaity finding.
The murder rate across 90 cities is down an average of 12% in the past year
Potentially a return to the pre-COVID trendline.
Chart from the New York Times:
OSD Discord server
If you like this newsletter and want to talk live with the people behind it, join the Discord server. The OSD team and readers are there. Good vibes only. Plus monthly movie watch parties.
Merch
Gun apparel you’ll want to wear out of the house.
Office hours
If you’re a new gun owner, thinking about becoming one, or know someone who is, come to OSD office hours. It’s a free 30-minute video call with an OSD team member to ask any and all your questions.
> So if your product gets 20% better with each iteration and you do 12 iterations in a year, it’ll be 1.219x better after a year.
Not even close. If your product gets 20% better with each *iteration* and you do 12 a year, it'll be 1.2^12 = 8.9x better after a year. The calculation you did is for if the "interest rate" (uncompounded annual rate of improvement) is 20%.