If you have a phone, you can call anyone you like. The recipient doesn’t have to answer, but the network is an open API, available to anyone. The Bell System began as an actual networked system–many companies, using the same protocol, shared calls with each other. When Bell got greedy and stopped interconnecting, phone use became annoying–several phones on your desk, or there were simply people you couldn’t call. AT&T interconnected when it was finally more profitable to own the standard than to block it, and the government mandate then locked that in and led to the phone system we have now.
Email caught on and persisted for the same reason, but more so. There was never “the email company.” Instead, there’s a protocol, and anyone can send and receive. For free.
It’s hard to overstate how profound the idea of permission-less contact was to the flow of information and the growth of commerce and culture. Just as you could mail a letter to a stranger, you could also call or email them. When it works, it’s very powerful.
Friction was the key. Stamps cost money. Phone calls required an account and a human to dial.
The first real challenge for most users wasn’t crank calls or wrong numbers. It was spammers. Computer users who would send thousands or millions of emails or phone calls at a time, taking advantage of asymmetry. It cost them nothing, but it cost each recipient something.
Usenet, the original internet discussion layer, died from this asymmetry. No one had an incentive to create filters or clean it up at scale, so people stopped showing up.
Even with free email, the first generations of spam were mostly uneconomic, and much of it was stopped by filters. No one liked the noise, but email was useful enough that we put up with some spam.
Multiply spam by AI and the cloud and VOIP, though, and it’s obvious that open networks can’t survive. 30 junk SMS notices in one day is more than enough to turn off your notifications. How many voicemails about an approved business loan need to show up before the signal-to-noise ratio becomes simply noise?
We’re going to need to create a cost for the sender (so it can’t scale to ridiculous) as well as an identity layer (so scammers can’t be fully anonymous–reputation is earned, not invented). It might be an open protocol in the spirit of the best parts of the net, or it might be a single monopolist that figures out how to extract value from it.
A thousand years ago, people built walls around their villages because criminal marauders with nothing to lose would destroy open cities. The scale of the new digital marauding is going to be so large (and it’s coming so fast) that we’re about to see a fundamental shift in how we contact each other and who we trust.
I have no idea what it looks like on the other side of this transition, but the assumptions we’ve always made about open communication with strangers are about to change.
We have no idea.
Of course, there’s plenty of behavioral data. Say this phrase, or offer that treat, and this particular dog is likely to act in a certain way.
But our inclinations about what it’s actually like to be a dog are all inventions, reverse-engineered to give us a clue about what they might do next.
“If I were you,” is a pretty useless sentence, particularly for dogs. You’re not them, and you can’t imagine what it’s like.
The same is true for computers and for AI. We make up a story about what the computer wants, expects or thinks. But it’s simply a way to explain our guesses about behavior, not actually a statement about what it’s like to be that device or program.
You’ve probably already guessed (there I am, imagining what it’s like to be you) that the same is true for other humans. We only know for sure what it’s like to be ourselves. Everything else is speculation.
Empathy is essential, but it’s also difficult.
“Why are you taking this class?”
That seems like a fair question. After tenth grade or so, it’s a choice, after all.
One honest answer is, “I have to get a good grade to get to where I want to go.” That means certification, compliance, regurgitation. It means enrollment in the outcome, not the process. When this happens, we’re seeing a failure of the system we call education. Because that’s not learning.
One answer is, “Because I’m curious.” This is a great reason to take a class, and the instructor’s job isn’t merely to satisfy the curiosity; it’s to amplify it and turn it into a habit and the practice of the autodidact.
For many professional settings, the answer might be, “To learn how to use these tools and this insight to make a change in the world after I graduate.”
That sort of enrollment becomes a productive bargain. It gives the student agency–you don’t have to like everything the instructor has to say, you don’t have to use it when you leave, but the standard is: Is it helpful to imagine having this tool in your kit, and does this course prepare you to use the tool effectively?
Teaching is expensive, so is learning. Active enrollment on both sides is part of the bargain. Students are free to reject the pedagogy, the tools, even the aims of the current practitioners of a craft. But they’re on the hook to do that after they’ve absorbed what the instructor has to offer.
Take what you need and leave the rest.
Apophenia is the uniquely human tendency to perceive meaningful patterns or connections in random or unrelated data, events, or objects.
Humans are story telling machines. And one thing we do is turn co-incident events into more than coincidences.
When we see faces and shapes in clouds, apophenia wastes our time in the form of pareidolia. There isn’t actually a teddy bear in that cloud, or a face in that grilled cheese sandwich.
On the other hand, our ability to make out patterns is essential when trying to understand a system. Systems are nothing but non-coordinated conspiracies, individuals following their interests in response to a culture that is shaped by individuals following their interests.
The skill worth developing is the insight to tell them apart. Useful stories when needed, uncorrelated noise when there’s nothing actually going on.
Frederick Taylor taught Henry Ford how to do mass production. Deming brought quality, systems understanding and respect for the worker. And operations research brought insight.
If you have a repeated production process, the method for improving it is almost always the same, regardless of what you and your team produce:
1. Measure before you change. You can’t improve what you haven’t observed. Go to the floor, watch the actual work, time it, and document what’s really happening—not what you assume is happening. Taylor called this time study. Operations research calls it data collection. Either way, you start by looking.
2. Map the flow. Trace the path of materials and information from start to finish. Where does work queue up? Where does it sit idle? Where does it move backward? A simple process flow diagram reveals bottlenecks you’d never see otherwise.
3. Identify the constraint. Your system can only move as fast as its slowest step. Find it. Everything else is secondary until you address that bottleneck. (At a buffet, when you double the number of stations of the slowest item, the entire line runs faster.)
4. Separate value from waste. For every step, ask: does this transform the product in a way the customer would pay for? Anything else—waiting, moving, inspecting, reworking—is waste. You don’t need to eliminate all of it, but you need to see it.
5. Standardize the best-known method. This is Taylor’s core insight: once you find a better way, write it down, teach it, and make it the default. Not to control workers, but to create a floor that everyone can build on. Deming’s insight is that variation is the enemy of quality.
6. Reduce variation before you optimize speed. This is Deming’s most important and surprising lesson. A consistent process running at moderate speed beats an erratic one running fast. Get the process under statistical control first.
7. Build in feedback loops, not inspection gates. Smart managers don’t like end-of-line inspection because it’s too late. Instead, give the people doing the work the information and authority to catch problems as they happen. The goal is to make quality intrinsic to the process, not bolt it on after.
8. Optimize the system, not the parts. This is where operations research and Deming converge. Making one station 30% faster can actually make the whole system worse if it just piles up inventory before the next step. Ask: what does this change do to the entire flow?
9. Involve the people doing the work. Taylor got this wrong—he treated workers as interchangeable parts. Deming fixed this: the people on the floor know things management never will. Create structured ways to capture that knowledge. Invest in reducing fear so people will share what they know.
10. Iterate in small cycles. Plan-Do-Study-Act is Deming’s learning wheel. Don’t redesign everything at once. Make a small change, measure the result, learn from it, adjust. Then do it again. The factory you want isn’t built in a single leap—it emerges from dozens of small, informed improvements compounding over time.
The meta-principle underneath all ten: respect the system and the people in it. Change the system before you blame the people.
And don’t get efficient at doing something you’d rather not be doing at all.
[You're getting this note because you subscribed to Seth Godin's blog.]
Don't want to get this email anymore? Click the link below to unsubscribe.