introduction

I write here about the strange loop we’re in: humans building minds; minds changing humans. Less futurism, more practice—what holds up when you ship.

The Strange Loop

Something has shifted. We are building systems that can surprise their creators—not only by failing in new ways, but by producing useful lines of thought we didn’t see coming. The tools are starting to feel like collaborators.

It shows up in mundane places: refactors, incident reviews, product decisions, code written faster than you can fully justify. The future arrived quietly.

What I keep coming back to isn’t the technology in isolation, but the feedback. We build AI, AI changes how we work, work changes what we build, and the cycle tightens. I want to understand the shape of that loop—and the cost of getting it wrong.

The Winding Path

My path looks discontinuous on paper: economics, bartending through grad school, then building AI systems.

Economics taught me to look for incentives and for the gap between a model and the world it claims to describe. Bartending taught me that people rarely say what they mean, but they always communicate what they need. Graduate school taught me that expertise is knowing which corners you’re cutting, and why.

Reading people and reading data are more similar than I expected. Both punish overconfidence. Both reward attention to what’s happening instead of what should be happening.

The strangest career advice I ever received: “Your job is to be confused in the right direction.” I think about that constantly.

Philosophy of Work

I believe in doing less, better. Shipping early. Building systems that fail predictably and recover gracefully. Staying close to real problems instead of their filtered versions.

There’s a particular kind of productivity that’s actually procrastination dressed up in busy clothes. Meetings about meetings. Documentation no one reads. Processes designed to distribute blame rather than create value. I’ve learned to recognize it by the way it makes me feel productive without producing anything.

The opposite is also true: the most impactful work often looks like nothing. A small change that prevents a future incident. A conversation that removes a misunderstanding. A constraint that makes the wrong thing hard to do. It compounds quietly.

The Quiet Revolution

The best AI isn’t a demo. It’s quiet infrastructure that keeps working.

I’m suspicious of technology that needs to announce itself. The most transformative tools disappear into the workflow. You stop noticing them the same way you stop noticing electricity. They become assumed, ambient, load-bearing.

That’s what I’m trying to build: systems that earn the right to be invisible. Not because they’re trivial, but because they’re reliable enough that attention can go elsewhere. The goal isn’t to impress engineers with cleverness—it’s to return human attention to problems that require it.

What You’ll Find Here

Some posts will be technical—gradient descent, attention, the geometry of embeddings. Some will be philosophical—what it means to build minds, how tools change work, the ethics of automation. Some will be small observations that don’t fit elsewhere.

I write to understand. If it helps you too, good.

---