Thinking Worth Multiplying

2026-1029

As I reflect on the past year, I’ve realized that much of my work has been driven by a fear of becoming replaceable. Our economy rewards intelligence and labor, but as AGI approaches, what happens to our value?

Most recently, I spent time collaborating with Sam at The Sentience Company (Sentience), fine-tuning models to speak and think like me. Sentience is creating digital twins that allow you to multiply yourself: your memories, your judgement, and your voice. This narcissistic view of AI stood out as a way to push back against a fear of being replaced: my thinking can scale beyond where I could personally be.

But in a world where intelligence becomes replicable and abundant, how can humans continue to grow their own value?

I keep coming back to these two challenges:

Scale your intelligence: How can your own thinking scale through agents, robots, organizations, and content?
Increase your own intelligence: How can you spawn novel ideas, perspectives, approaches, and conviction by being a curious learner every day?

Scale your intelligence

Scaling thinking is already a ubiquitous practice. We scale ourselves through writing, content, and organizations. Articles and videos let others encounter you without your flesh being present. Your team lets your principles encounter decisions without needing you there.

Agents extend this further by applying your context and instructions to new tasks.

Sentience takes an even more personal approach to scaling your intelligence. The product ingests as much of what you see and do as data. It ingests your screen recordings, audio recordings, and iMessage to triangulate your conversations, memories and thoughts.

While this makes Sentience a powerful recall machine, there is still a significant gap between recalling my life and replicating my judgement.

Even if Sentience could perfectly replicate my judgement, would my thinking be worth multiplying?

Increasing your own intelligence

Much of what we call AI slop is familiar thinking produced at scale. Humans are culprits of producing slop: repeating accepted ideas, applying familiar frameworks, and reaching predictable conclusions.

As intelligence becomes easier to replicate, humans should focus on developing perspectives, ideas, and conviction that lead to value creation in the non-obvious.

This is what I believe to be our advantage over silicon intelligence: the ability to think “out-of-distribution.” Making a coherent series of unconventional bets, guided by unique and strong conviction.

White balls fall through a Plinko board and cluster around familiar outcomes. An orange conviction ball follows a series of unconventional bets to an outcome beyond the crowd. — White balls cluster around familiar outcomes. An orange ball follows a series of unconventional bets beyond the crowd.

Musk’s ambition to support civilization on Mars is a recognizable example. That conviction sustains a willingness to make successive bets that look unreasonable in isolation. Reusable rockets are one expression of that thinking. The deep conviction allows these bets to compound, supplying them a higher likelihood for an outsized outcome — even when seemingly every step feels like it departs from conventional wisdom.

These bets still have to produce something meaningful. Conviction needs to interface with reality through action: take action, receive feedback, adapt, repeat. For humans, building, selling, and encountering rejection create the opportunities to test and develop our thinking. Conviction guides the direction.

What this means tangibly is that the already imperative tasks of choosing worthwhile bets while learning from their consequences become even more important.

Onwards: Brains, Voice, and the Pursuit of Asymmetry

Human-computer interaction is central to this loop. Interfaces influence what information we consume, how quickly we test an intuition through an experiment, and most importantly, how effectively we learn.

My work has explored this process through voices and agents. What does an interface look like that takes people’s half-baked thoughts, manifests them into a prototype, and helps people learn from them?

As we build agents that scale intelligence, we ought to build tools that help humans develop durable intelligence.

This is my life mission; with limited time and energy to read, talk, build, sell, and learn, how can we optimize our tools and interactions to help us develop thinking worth multiplying?