There are two writers that always make me feel a little smarter after I read anything they write. One is Nassim Nicholas Taleb. The other is Naval Ravikant.

Taleb often talks about something called the Lindy Effect. You can look it up but in my words, it means that if something has endured for a long time, chances are that it will continue to endure for at least a similar amount of time.

Certain kinds of food are great examples: Wine. Coffee. Cheese. Bread.

The reverse also holds true. This means that newer things have to prove themselves over time. There is no guarantee that they will endure.

Writing is an old technology, for example, that has been around for thousands of years. But email is not. Email is very new.

Following the Lindy Effect, written forms of communication are likely to endure. But email will likely get replaced by something better. Hopefully soon.

The Lindy Effect explains why we no longer use stone tablets or papyrus but we still use writing.

Now let’s talk about education.

Bricks and mortar institutions of learning have been around a long time. Universities, for example, have been around at least since medieval times in Europe, but possibly earlier elsewhere in the world.

Other types of physically located places of learning have also been around a long time as well.

Like the idea of a traditional Māori Wānanga in Aotearoa New Zealand.

But some of the modern manifestations of public and privately funded education are much newer.

Like Polytechnics and Institutes of Technology for example. These have been around since at least the 18th century in Europe.

The oldest polytechnic in New Zealand started out as the Auckland Technical School in 1895. It’s now a university. Let that sink in.

Apprenticeships, in contrast to polytechs, are a much older technology. They go back at least to medieval guilds in the formal sense. And in an informal sense, they are probably older still.

Also newer are the bureaucracies that fund, monitor and quality-assure public and private education.

Again, for contrast, much older is the idea of private wealthy patrons who act as protectors, sponsors, and benefactors.

This is where Naval Ravikant comes in. On Twitter recently, @naval said:

Scaling a company via code and contracts instead of via people mitigates the Principal-Agent problem.

I had to look this up. You can read about the principal-agent problem here.

In my words, the principal-agent problem is when one group (the agent) acts on behalf of another group (the principal). The result is eventually chaos or some kind of disaster because the agent usually acts in their own interests rather than in the interests of the principal.

The textbook example seems to be voters (principals) versus politicians (agents). But I want to talk about education and funding agencies.

A funding agency is a principal. An education provider receiving the funding is an agent. Now read this quote:

“Too Big To Fail” is another example of the principal-agent problem. The idea behind too big to fail is that some companies become so significant and critical to the economy, that no matter what they do, the government will bail them out.

This situation creates a moral hazard, where the agents have no incentive to do the right thing since they know they won’t be holding the blame at the end.

Taleb’s latest book is called Skin in the Game. You should buy it and read it. Or not. Most people won’t like it.

But this kind of situation where the agents have nothing personal at stake is what he would call a lack of Skin in the Game.

Here’s what’s happening in Aotearoa New Zealand at the moment. I’m just talking about the polytechs and Institutes of Technology:

• $50 million government bailout to Unitec Institute of Technology.
• $15 million government bailout to Whitireia Polytechnic in Porirua.
• $33m bailout for the West Coast’s Tai Poutini Polytechnic in February.

My math is not that great. But that adds up to about $100 million. Give or take a few million.

If you’ve read this far, try and hold on to those thoughts while I introduce a couple more things. @naval also says this:

Scaling via community-generated code and content mitigates Hayek’s Local Knowledge problem.

I had to look this up too. You can read a short blurb on the local knowledge problem here.

The local knowledge problem says (my words again) that on their own, central planning authorities (like economists, funders and government bureaucracies), despite increasing amounts of increasingly sophisticated statistical data, don’t know really jack about whatever it is that they’re planning and funding.

If they did, they’d be able to anticipate a $100 million blow up.

Hayek would say that the special knowledge you need to understand and anticipate these kinds of events is distributed among individual actors.

What he means is that outside of centralised silos there exists very important but often unorganised bodies of knowledge held by individuals which give them advantages over others.

You can’t leverage this kind of knowledge out of statistics and big data. At least, it’s not clear to me that you can. Rather the danger is that more data just results in more noise and less signal.

And the only way in which we can make use of this special knowledge (as a collective) is with the active cooperation and collaboration of those who hold it.

And then, only if their self-determination is maintained if I understand Hayek correctly.

This is why I think Taleb says (somewhere) that: “theory works in theory but not in practice”.

Taleb would also call the $100 million blow up an example of the “Turkey Problem” or a “Black Swan”. You can google those later.

Right now, think about Wikipedia.

Wikipedia is code. There are no offices, no bricks and mortar buildings. People volunteer their time to build it. It’s a community project.

Because it’s code, Wikipedia is infinitely scalable. Volunteering your time is a kind of contract. Volunteers are (mostly) interchangeable over time but also scalable.

And Wikipedia is the main thing. There are no celebrity Wikipedia content creators creating entries to serve their own interests.

Or when people do try this it gets caught quickly and edited out by the community including a decentralised network of experts with incredibly specialised knowledge.

This community code approach means that Wikipedia solves both the principal-agent problem and the local knowledge problem in its DNA.

Now back to education.

I know that no one is going to give me $100 million dollars. But that is a lot of money.

I could do a lot with $100 million. I believe the appropriate mathematical measure word is “a shit-tonne”.

Of course, if someone gave me a theoretical $100 million then we’re back to the principal-agency problem.

Central funding agencies are a bit like banks and a bit like venture capital (VC) investors. That’s what I’ve been told.

I don’t know about banks but VCs at least talk about Return on Investment (ROI). They also talk about “sunk costs”.

I don’t have a view on what the government should do about the fragile polytech sector. But I do like thought experiments.

This is article is a thought experiment you can try in the safety of your cubicle. Or kitchen table if you work from home like me.

There is the main question and then a series of other (possibly leading) questions. I like to think about each one deeply and see where it takes me.

If I had $100 million dollars to build a new kind of tertiary education model, what would it look like?

• Would it look more like a business, a polytech, a university, an apprenticeship model or something else? Note to self: these are not mutually exclusive or even the only categories.
• What if I applied the Lindy Effect as a filter to models of education? What existing models have stood the test of time? What newer models are likely to endure and why?
• If I had to build this new model along the lines of either an encyclopedia or Wikipedia, which one would I choose?
• How would I solve the principal-agent problem? Or would I need to?
• How could I build a new education model out of code and contracts rather than bricks and mortar and people?
• How could I design the model so that it’s built out of community-generated code and content?
• Who would the customers be? This on its own is a non-simple question so don’t skip it just because it’s last.

Feel free to vent in the comments.