Thinking like a gardener, slime mold, the adjacent possible: Product advice from Alex Komoroske

So much of the way that we tackle problems and build products is this builder mindset. It's like I have a plan. I then manipulate things to match my plan and make it happen. And this is a way you can create tons of value. Part of the problem though is it can't possibly create more value than the effort that you put into it. What I look for instead are things that can be gardened, things that can grow on their own and that you can direct or maybe give a little bit of extra energy to or curate over, and is a totally different mindset for it.

Anything that is shaped like an ecosystem that has some kind of network effect, and many things have network effects, have some kind of compounding loop. Compounding loops are not rare. They are, 'Look.' It's like truffle hunting. You have to know what you're looking for and find the dynamics of a thing that if it worked would work at an accelerating rate.

You need both. So if you only do incremental, you will follow the shortest, the steepest gradient in front of you... So you need coherence about where you're going and the way you get that is by creating a North Star for yourself.

Komoroske refers to LLMs as "magical duct tape" - a versatile but imperfect tool that can handle many tasks but shouldn't be expected to work flawlessly.

Don't assume it's going to solve all your problems. Don't assume it's going to do autonomously be able to give high quality results of every case. But what can you now build now that you have magical duct tape?

The core dynamic that makes organizations hard to navigate as they get larger, even if you assume everyone is actively good at what they do, actively collaborative and actively hardworking, is this emergent force or coordination of finding the subset of projects to work on when everyone's super busy that everyone agrees and commits to and actually works on together. And finding this coordination cost grows with the square of the number of people who are working on that thing.

When the company is very small, you can drive it, you can steer it like a sports car. As a founder, you are allowed to steer. Everyone acknowledges you are allowed to steer. They're never, 'Why is he steering it that way?' So you, a founder, can help navigate an organization around an obstacle the organization cannot see or comprehend itself. The problem is as you steer, as you grow into the size, your organization goes from a sports car and you grow into the size of a big rig, if you drive a big rig like a sports car, you're going to be a danger to yourself and others on the road and you're going to grind the engine.

LLMs are magical duct tape. They're formed principally by the distilled intuition of all of society into a thing that operates between, a cost structure between human and plain old computing.

If you want to get your team to do good work, there's a million different paths to do that. If you want to get your team to do great work, there's no shortcut other than to have an extremely high-trust environment where people lean into their superpowers in a way that adds up to something greater than some of its parts.

Komoroske is describing the ground rules for his "strategy salons" - informal discussion groups he creates to generate ideas in low-stakes collaborative environments.

This is a collaborative debate environment. This is only 'yes, and'. If somebody says a thing in this group that is optional and secret and completely off the side of anything that matters, if they say something that you think is an actively dumb idea, you are free to not engage. Just leave it. That's fine. Because nothing's going to happen. We're deciding anything interesting or important here.

LLMs allow writing shitty software to be significantly cheaper, not necessarily good software, but good enough in certain contexts. And also it means that there's certain software now that isn't plain old computing that can be run cheaply. It's relatively expensive marginal cost.

We're used to computers doing exactly what we told them to do, which is not necessarily what we meant. And only some people have learned the skill of programming, the arcane magical incantations to make computers do exactly what you meant. Now, LLMs can do all kinds of stuff. And they don't do exactly what you told them, but they do typically do roughly what you meant.

The adjacent possible is a set of actions that you do, that you can do. They are right in front of you that if you do them, they would work, almost certainly work.

If you can slice up your decisions into smaller and smaller decisions, I'm like, 'This next step definitely makes sense.' It will almost certainly pay for itself or the very least won't be too expensive. And then it might allow these other things to happen and you take it.

You want to have a space too small, a time too short. If you have a big cabinet space of a lot of people in it and no one's talking, people go, 'I guess this is the place where we don't talk,' for whatever reason.

A community with no people talking is definitely dead. A community with one person talking is already dead. You don't realize it yet.

Most of the tools that we adopt in the workplace are collaborative where it helps your team be better, helps you collaborate better, and AI is the opposite. It makes you individually better.

Most of it is slop. And so in this cacophony, how do you stand out? You stand out by having good taste. I think taste is the most important thing.

I'm saying I don't have to know the answer to the thing. If on a systemic basis I let these ideas and then you respond to the ones that are working that are viable, it doesn't really matter if you didn't know ahead of time which ones were going to work.

My approach at Google was 70% of my effort and my team's effort should go on things that everybody acknowledges are important and useful and create value. Maybe it's boring, linear value, but some kind of value. You're trying to minimize the chance that any other person in a company will say, 'What does that team do anyway?'

There's tons of stuff you can't know ahead of time. So if you're getting a false precision at the beginner, that's a comfort blanket. That's just helping you feel like there isn't uncertainty. There's uncertainty everywhere all the time.

Do things that give you energy that you are proud of.

"This thing" refers to LLMs (Large Language Models), which Komoroske describes as "squishy computers" that don't do exactly what you tell them.

If it punches in the face, that's not a viable product. And so how do you design your products assuming that this thing will be squishy and not fully accurate and fully work?

When I meet with people, when I mentor them, I try with it within the first session or two, whenever I can get a hypothesis, I say, 'I think your superpower is...' And I describe to them what I think I can see them being truly exceptional at.

I try to see the greatness, the seeds of greatness in everything. Everyone and everything around me, I look for, I try to find and see, man, what is the most compelling part of this? And let me lean into that.

Alex refers to LLMs (large language models like ChatGPT) as "magical duct tape" - a versatile tool that can connect different systems but isn't perfect.

Don't assume it's going to solve all your problems. Don't assume it's going to do autonomously be able to give high quality results of every case. But what can you now build now that you have magical duct tape?

Happiness is reality minus expectations.

Kayfabe is a word that comes, I believe is a carny word that is used and applied to professional wrestling and it means a thing that everybody knows is fake and yet everybody acts like is real.

If you as someone who sees this can see that wait a second, the official strategy is definitely not going to work, you're like, 'I got to tell somebody. We're doing work that's going in a bad direction. It's not going to work.' And you go and say, 'I think that it's actually not going to work for these reasons.' And what someone will say to you, this happens by the way, I'm not substituting any particular organization, this happens literally in all organizations to some degree, is the senior person will say, 'Alex, I agree with you. It's not perfect, but if you hit the ground truth button, if you share that and everybody, the whole thing will shatter and we can't do anything.'

The underlying dynamic that must be true in any organization on a fundamental basis is you can't make your boss look dumb because if you do, they're the person who decides, 'Oh, this person's not performing,' or whatever. And that one little asymmetry, that one little fact, in most cases it does not matter. That one little asymmetry is what leads to the systemic compounding thing where you get these really weird dysfunctional emergent things that everybody hates, nobody wants, and yet nobody is in the position to change per se.

I think in this early stage, we're in the community gardening phase, not the factory farming phase of this technology. And so I think what people need most is curiosity and play.

AI tools being discussed can double individual productivity but are like "magical duct tape" - useful for quick fixes but hard to scale organizationally.

I can now do my job twice as fast. And if the organization sees that, they might go, 'Wait a second. We should pay you half this much.' They're like, 'What if we get rid of some extra people or something?' And so if this stuff is magical duct tape, it's very hard to make scaled, repeatable, large scale things out of it.

Always rules are better than sometimes rules for self-control. And so if you're going to diet, 'I'm going to skip lunch every day.' Like holy, you haven't full thought on that at some point like a day with a big executive review, 'I really need to make sure I'm well-fed before I go into this review,' or something. And now you've broken the streak and now it's over.

"Flip that" means allocate a short amount of time (5 minutes) to do the task, rather than using the full two hours available.

If you have two hours to do a five-minute task, the effort to do that is impossible and instead you should flip that.

The secrets of life is things you've heard a million times already, you just weren't ready to hear them.