Bits and Bobs 12/18/23

Good is at least above some bar of viability: good enough.

Great is actively stand-out in quality.

Good is linear; great is super-linear: an order of magnitude difference.

Often, the amount of work it requires to take something from good to great is also an order of magnitude or more.

But sometimes you have a special, differentiated edge, where it's easier for you to achieve greatness in that particular domain.

Maybe you have the wind at your back (you've caught a rare tailwind).

Maybe you have increasing personal momentum propelling you forward (e.g. a flywheel).

Maybe you just enjoy the work required to get to greatness in that context and so it's "cheaper" for you.

You should spend most of your time on places where you have a sustainable edge on creating greatness.

Rental: Transactional, short-term, maintenance.

Ownership: Investment, long-term, transcendence.

Humans can be kicked into a rental mindset by little things.

For example, if the desk you sit at in the office is reset to a default, neutral state every night, you'll be in more of a rental mindset, not an ownership mindset in that context.

One reason is that the collective emerges out of interactions of the parts.

You can't take a picture of a moving stream and hope to understand it.

But the emergent whole would still be hard to understand even if somehow interactions didn't matter.

For example, imagine every agent in a system has a small, consistent bias towards a certain direction.

For each individual agent, the bias is swamped by the overall specific details of the agent.

That means the bias in invisible when looking at each individual agent.

But at the level of the collection, all of the random details balance each other and all that's left is the consistent bias.

That consistent bias is now a strong force at the level of the collective.

It's a kind of Simpson's Paradox.

Examples of this kind of thing: It's possible--common, even!--for a group of extremely smart people to collectively act like a total dummy.

The thing that emerges can be beautiful and great; but the more common default is a slouching, invisible, all-powerful eldritch horror.

When people hear I wrote on real magic, it's easy to think "pfft, how fanciful and unrealistic! I hope he's being metaphorical..."

But I'm dead serious.

Magic emerging in the physical world is not possible.

But in the social world, magic is absolutely real, common even.

Decisions made in the social world affect what happens in the physical world.

So magic absolutely affects the physical world.

But things that are alive, that grow and emerge themselves, are very often beautiful.

Those alive things can be beautiful without anyone having an intention to create something beautiful.

The highest and best use is emergent, and sometimes the thing you built it for actually blinds you to non-traditional but more powerful uses of it that others will discover.

But the second (and third and fourth and...) order effects matter, too.

Just because you didn't think about the emergent second-order effects of the the thing you built doesn't absolve you of the responsibility for them.

A rare few know how to apply the principles of ecosystems to catalyze emergent outcomes that could not be built.

Emergence is a form of magic.

Loving emergence is learning to love the mess.

Emergent systems are alive.

They're not possible to control, but they are possible to surf.

That is, to harness some subset of the energy to propel you to a great outcome, and to impart small biases into the system that lead to wildly different results.

(No, I don't think they realized how revealing that advice was.)

Absolving yourself of all responsibility, allows yourself to move faster locally and reap the rewards, while potentially sowing chaos everywhere for everyone else.

A very transactional and instrumentalist, self-centered mindset.

Just because you didn't think about the emergent second-order effects of the thing you built doesn't absolve you of the responsibility for them.

Here's one: "The things that make a B2B software product beautiful are precisely what make it not possible to be broadly successful in the enterprise segment"

The thing that made me think this is I can't think of a B2B software example that is unambiguously beautiful/high-quality (there are some candidates but they all have big caveats). That made me wonder if maybe it's something more fundamental.

Here's a sketch of what that fundamental thing might be.

Enterprise B2B business models require stickiness for the CAC/LTV math to work.

Stickiness requires customizability (the ability for a customer to glue themselves to a product.)

Products that are so opinionated / beautiful: "We've thought long and hard about this, and this is the way to do this" that we either don't give any space for users to customize (where we get it right) or, more commonly especially for large customers, we make our solution non-viable for customers with different needs.

I imagine large users saying: "That's cute that you think you found a one-size-fits-all solution for everyone. Our business is complicated and large and we know it better than anyone, your 'we know better than you' is hubris."

To be viable as a business, B2B products need to be sticky.

To be viable for large customers, a B2B product needs to be customizable.

Beautiful things are rarely customizable, unless they are beautifully customizable (e.g. Lego's beautiful dot connector system).

This customizability in B2B products creates a space... no, a fundamental requirement! ... for pro serve, implementation hand-holding, channel partners, and ecosystems of solutions.

This ecosystem makes the core B2B product so much more resilient; wrapped in multiple layers of self-reinforcing and accelerating pace layers of the ecosystem.

But to do this, you must give up control. You must give up control to the user to customize, including a potentially dizzying array of options.

You need to give up control to the ecosystem, to be able to build a viable business helping customers customize and implement your thing.

To have an ecosystem, you need to create more value than you capture--creating the conditions where partners can build a sustainable and valuable business adjacent to you.

If you don't seem particularly willing to give up that value to others, that means you can't have an ecosystem sprout up around you.

The Lego Dot is a timeless abstraction. The Unix "Everything is a file" is a timeless abstraction.

Timeless abstractions are often timeless because they are simple, resilient, and flexible.

Timeless abstractions are significantly more difficult to create than individual timeless objects.

Timeless abstractions need to correctly absorb all of the evolving requirements of a changing context.

Timeless products only need to be self-consistent.

It's the whole, situated cocktail of self-reinforcing ideas.

A lot of the details about how they work aren't actually important on their own, but rather as part of an emergent quilt of things.

Sometimes each detail is actually a thing that they succeed despite.

It's very hard to not pick and choose an incompatible/ineffective set, or apply them in a different environment.

"We set this impossible goal on a linear-at-best business, and it's not working, so put all of the chips on that to catch up. Put everything on the hardest, least successful thing, no time for literally anything else!"

This sounds silly when caricatured like this, but I've seen it show up again and again in basically every organization I've ever been a part of.

It's very easy to accidentally spend almost all of your time scurrying around on the hard things that don't matter.

Most of your time should be spent on the easy, possibly-game-changing things that do matter!

If they're easier for you than others, that's a great edge to lean into!

Sometimes the macro tailwind is so powerful that you'll do well despite this problem.

In those circumstances, it's easy to trick yourself into thinking the scurrying on things that don't matter is precisely what led to your success!

Abstract things are easy to be elegant.

The real world is wrinkly and turbulent.

A lot of beautiful ideas in the abstract are not viable in the real world.

She looked up to see the horror on my face.

"Don't worry dad, there will be other bugs!"

"I'm not worried about whether I'll see more bugs. I'm sad for that specific bug that no longer exists, whose inner light has been snuffed out forever."

Viewing the world entirely from your own direct perspective is childish thinking.

Mature thinking requires the ability to take on others' perspective (even ones unlike yourself), to view them as ends, not just means. To take responsibility for even the indirect effects of your actions.

Each agent gets an edge by doing yet another loop or reaction to others' actions.

In the military, a powerful meta-strategy is "get inside your opponent's OODA loop".

That is, do your action/decision loop faster than your opponent, so by the time they're reacting to your last move you're already on the next one

They can never catch up and you can run circles around them.

This edge means that in competitive interactions agents gets a bit of an edge from running the OODA loop a bit faster than their competitors.

This can lead to a rampant acceleration, up to the maximum viable speed in that context.

No time to think, only to do. No time to think through second order effects.

What's the worst case if this idea doesn't work?

A very small, capped amount of opportunity cost.

What's the best case if this idea works?

Potentially uncapped upside.

Capped things are easier to estimate than uncapped things.

The average of a capped thing is useful; the average of an uncapped thing is meaningless.

But they're often very embarrassed about the idea that what they write will be low quality.

They want to make sure the first post is great, a bar that they never can muster up the surge of energy to clear.

So they end up never writing anything, and the world is worse off!

The advice I give in this circumstance is "Don't worry, nobody will read it."

This is what might be called self-capping downside.

If you write up something and quietly post it somewhere (as long as it's not controversial, because those can go viral for bad reasons), then the default is no one reads it.

If no one reads a not-great thing, then no downside manifests.

But if it's good--if people who stumble on it or read it after you share with a "I burped up this silly thing, sharing just in case you find it useful"--then the upside is uncapped!

The core dynamics of self-capping downside:

If it's bad, no one will use it.

If it's good, then people will use it.

And if people are using it, it's good, even if you thought it was bad!

If making the thing is enjoyable for its own sake (e.g. a fun hobby project) then there's basically no downside!

Highly effective at driving outcomes... but almost certainly not the outcomes you wanted.

No matter how clever you think you're being with the incentives you've built, the emergent real world situation will select for something that you didn't intend.

It's impossible to model all of the precise balanced forces in-situ of agents all scrambling to get an edge in a recursive bramble.

The swarm can calculate the actual optimal answer way faster than any builder of the incentives can.

This is Goodhart's law.

No matter how careful it is, it's going to smoosh a lot of things.

The elephant can do other moves, but a subtle dance is not open to it.

Lean into your actual edge, not the edge you wish you had (or that you've long since lost).

If a startup does something dangerous or creates a bad product, they tend to be self-extinguishing; they go out of business before doing too much damage.

(Of course there are exceptions, e.g. a startup that creates a thing that benefits the user at the cost of massive negative externalities)

A startup is a little mouse. They can tiptoe through a garden.

For society, it's best for innovation that has possible downside to be done by a swarm of mice, not a small number of elephants.

It's hard to do it without completely destroying the future sustainability of it.

If you need to heat a home, you can always burn the furniture.

If you force a number to go up where the fundamentals don't allow it, you will incentivize burning the furniture.

I got pretty good at deftly navigating the political hailstorms, which looked kind of like magic.

"How did you dodge all of those bullets?"

"Remember the scene in The Matrix where the kid says 'there is no spoon'? The secret is there are no bullets."

Practical knowhow and experience is only possible to learn by doing.

Knowhow is necessary for success in practice.

When you think you know the thing already you'll be less open to learning about it.

Sometimes reading the book makes you less likely to succeed. ("Oh, I read the book, I've got this!")

"We've decided to build a flying submarine. You're a metals expert, what is the best metal to use?"

An illuminator who illuminates only one person (especially if that one person is their boss) is called a sycophant.

An illuminator doesn't just encourage the other person blindly; they encourage them in things that are seeds of greatness that are likely to have a net positive effect on the overall system.

I threw together a little GPT to give everyone an illuminator in a box. (Obviously it's no replacement for the real thing!)