What has my attention

AI tooling for operational businesses

Working on a system that handles the classification, routing, and follow-up of operational exceptions. The kind of work that currently requires constant human attention. Early stage, but the architecture is becoming clear.

Iredale Digital: new client work

Taking on a small number of design and development projects for businesses that care about quality. Selective. Focused on work that is genuinely interesting.

Agent architectures

Spending time understanding how multi-agent systems work in practice. Not in demos, but in production. The gap between the two is significant and interesting.

LLM reliability patterns

How do you build systems on top of models that are probabilistic? What are the patterns that make this work reliably? Reading, experimenting, and taking notes.

Fleet and leasing automation

Mapping the administrative and compliance work in fleet operations that could be handled by well-designed systems. A lot of it is more tractable than it looks.

The Innovator's Dilemma by Clayton Christensen

Returning to this. Still the clearest framework for understanding why good companies fail.

Working in Public by Nadia Eghbal

On open source, but the ideas about how public work compounds over time apply broadly.

Various papers on agent evaluation

Trying to understand how people are measuring whether agents actually work.

The economics of AI for small business

Large enterprises have the resources to build custom AI systems. Small businesses do not. What does the right tooling look like for an operator running a 10-person business?

How to write well in public

Still figuring out the right cadence and format. Trying to write things that are worth reading, not just things that are easy to produce.