Research Insights.
Methodology · domain notes · practice retrospectives — first-hand thinking from the field and the workbench.
murRipple, closed out — a song woven into visible ripples
The project is done. It leaves two documents behind, one for readers who want to understand it and one for people who want to use it.
I had an AI run a capability checkup on me
Standalone reportOne rule only. Nothing I claimed about myself could count, only machine-checkable traces in the repositories.
Two Weeks from Zero to a Customer Demo
Speed does not come from working late — it comes from cutting the work down to a single line.
Verified = Command + Output
The bill for one “it should be fine” usually falls due three weeks later.
One Person Walks into Wind-Tower Monitoring
The hardest step is not the technology — it is getting someone to lend you their asset for an experiment.
Putting Reins on the AI
What sets the ceiling is not how you ask — it is the layer outside the model.
Human + AI: A Multi-Window Operating System
Not hunting for a smarter model — designing an organization that keeps ordinary AIs from drifting.
The City Under Your Feet Is Talking: An Introduction to DAS
How an ordinary telecom fiber becomes a stethoscope several kilometers long.
Signal and Noise
The discipline for deciding whether a waveform is a real event turns out to work on AI answers too.
From Optimal Transport to LLMs
After changing tracks three times, I found that what followed me was never the tools — it was one question.
Why I Write a “Task Book” for Every Project
Once AI made writing code easy, what matters most is saying exactly what you want.
AI as a Partner, Not a Tool
The dividing line is not how strong the model is — it is whether you give it a chance to disagree with you.