an unstructured distillation of my taste and various bouts of inspiration and thought tunnels.

What I’ve been thinking about recently

thoughts on new interfaces

*hacking incomplete/uncertain environments to increase information advantage and decision making — and the human interaction layer to enable this.

Why: Too often, systems assume reliability scales linearly with more data, as if data were discrete and absolute (this is a philosophy carried by many big data companies). But in dynamic systems (anything from battlefields to live sports) data is rarely absolute. This perspective is dangerous in scenarios where we can never get the full operational picture — we are left in an endless hamster wheel of collecting more data.

Tailwinds: In the military, we rarely have a complete picture of the enemy yet strategies are built on unrealistic assumptions and brute-force scenarios. In space, simulation engineers spend dozens of hours curating assumed data about the lunar landscape. In tech at large, we are still stuck hand labeling data. The root node is unnecessary brute-force design of scenarios.

How: This problem is fundamentally an interface challenge. If we can no longer assume we have all the data, we must embed uncertainty as a fixed part of the equation. Current simulation software has a system brute-generate data first, then test the system on a user approach. The next interfaces will evolve more towards co-adaptation: Humans fill in knowledge gaps and AI uses this to evolve the behavior/features over time.

If you work in this space (or adjacent), or are thinking about the same problems, please reach out.*

brands

I enjoy things that are raw and stripped to its essence.

taste