TL;DR: Everything I’ve worked on in the last decade comes back to one thing: incentives. I didn’t plan it that way. I just keep ending up there.

The Thread

Charlie Munger said it best:

“Show me the incentive, and I will show you the outcome.”

I’ve quoted this for years. But recently, I realized it’s not just a quote I like. It’s the lens through which I see everything.

Products, protocols, teams, communities, markets, labeling pipelines, compute allocation, payments, decisions, etc. It all comes down to incentives.

2016-2017 and 2022-2025: Crypto

In 2017, I wrote Cryptocurrencies: It’s all about incentive.

The core idea was simple: the innovation of blockchain isn’t decentralization for its own sake; it’s designing systems where self-interested actors produce collectively useful outcomes.

I went on to spend years building at Pocket Network between 2022 and 2025, where adversarial incentives were the core primitive. Every protocol decision was an incentive design question. Stake, slash, reward, repeat.

Over time, I kept coming back to this insight:

Decentralization is never the goal. It’s a byproduct of a permissionless and incentivized system.

2020-2023: Labeling

In 2023, I wrote An Incentive to Label.

The thesis: the bottleneck in AI isn’t compute or model architecture. It’s getting high-quality human labels. You get quality by designing the right incentives for labelers: deposits, slashing, skin in the game.

This came from spending a couple of years at Waymo on the Planner Evaluation team. My focus was on AI evals and labeling pipelines.

Same pattern. Different domain.

My core takeaway from this era was:

High quality ground truth in qualitative environments is subjective, and must be incentivized.

2025-2026: Micropayments: Earning & Tipping

Most recently, I’ve been working on micro-tipping.

The question: how do you fund original content on the open web when AI agents are extracting value without compensating creators?

The answer: It’s not only paywalls. It’s lightweight, optional incentive mechanisms. Small tips that feel good to send and receive. Public ledgers that create social incentives. Lottery-like rewards that make participation fun.

They’re an AND with paywalls (preview, payments, subscriptions), not an OR.

It’s incentive design again. Just applied to the attention economy.

The Pattern

I didn’t set out to become an “incentive guy.” But when I look back at every role, every project, every problem I’ve found compelling, the thread is obvious:

  • Twitter, Google - Tradeoffs in user products: convenience vs. privacy, engagement vs. quality.
  • Magic Leap - Cross-functional alignment in a hardware-first environment. How do you get researchers, engineers, and product to move in the same direction? More importantly, how do you align incentives between R&D and the business when funding, bureaucracy, and market pressures are all pulling in different directions?
  • Waymo - Labeling pipeline quality. How do you build reliable ground truth and evals when humans are subjective and fatigued?
  • Crypto - Economic incentives as the core primitive. Adversarial game theory by design.
  • Micro-tipping - Funding the open web. Making it rational for both humans and AI agents to pay for what they consume.

Incentives and tradeoffs span convenience, time, trust, cognitive effort, and money.

So What?

I don’t think this is unique to me. I think most people have a core lens they apply to everything but never name it.

For some people, it’s systems thinking. For others, it’s storytelling, empathy, or pattern recognition.

For me, it’s incentives.

If you tell me the problem, I’ll ask about the incentives. If you show me the incentives, I’ll predict the outcome. And if the outcome is broken, I’ll look at the incentives first.

Munger was right. He was always right. RIP.

It was all about incentives in 2017. It’s still all about incentives in 2026. I suspect it will be in 2035 too.