Version 1.0 — January 2026

I’m trying to build a consistent way to think about a recurring problem in my work:

Given a structure, what signal should exist? Given a signal, what structure can be inferred?

This page documents my current approach. It’s not a prescription for others, it’s a record of what I’m trying to do, so I can check later whether I actually followed it.

Current workflow

When I start a project, I try to:

  1. State what’s actually measured
    Not what I wish I’d measured, but what the instrument actually reports.

  2. Write down the forward physics
    How do I think the signal gets generated? What assumptions am I making?

  3. List alternative explanations
    What else could produce this? Geometry, defects, artifacts, effective parameters?

  4. Make constraints explicit
    Measurement limits, fabrication bounds, modeling assumptions, what I’m taking on prior.

  5. Ask what’s identifiable
    Can I actually separate these effects with this data? Or am I fitting a linear combination?

  6. Stop if I can’t distinguish
    If the data doesn’t support the inference, I try to say so instead of forcing it.

  7. Report bounds, not just conclusions
    What survived, what didn’t, what would help resolve it.

Why I’m doing this

I’ve run into situations where:

  • My model fit the data perfectly but I couldn’t tell which mechanism was real
  • Adding more parameters made fits better but understanding worse
  • Agreement with experiment came from error cancellation, not insight

This workflow is my attempt to catch those issues before they become claims.

It’s built from mistakes I’ve made (or almost made) across 5 undergraduate projects. That’s not much - I expect this to change as I encounter systems it doesn’t handle well.

Where this gets applied

Case Studies: Examples where I tried to follow this
Constraints: Where it actually mattered in each project
Reading Ledger: What I’ve learned about method limits
Notes: Confusions this workflow tries to prevent

Status: This reflects my thinking as of January 2026 (graduation soon). Expect revisions as I learn more.