1. The mentat habit

In Frank Herbert’s Dune, mentats are human computers trained not to memorize facts but to construct models. Their discipline is to hold many variables in relation and to know the difference between data, inference, and projection. I read papers the same way: not to absorb words, but to reconstruct the argument and find the joints where it could break.

This entry is about the reading step after selection: how I take a paper apart.

2. First pass: what is being sold

Every paper is selling something. Sometimes it is a new measurement. Sometimes it is a new method. Sometimes it is the author’s next grant. I try to identify the pitch on the first pass, because the pitch tells me what the authors think their contribution is.

I read the abstract, the conclusion, and the figures first. If those do not make the claim clear, the paper is either poorly written or the claim is weaker than the authors want to admit. Either way, I note it.

3. Second pass: the chain of evidence

Then I look for the chain:

  • Assumptions: What did they have to assume to make the problem tractable?
  • Method: What did they actually do, and what did they measure?
  • Analysis: How did they get from measurement to conclusion?
  • Limitations: What do they admit, and what do they omit?
  • Generality: How far beyond their specific setup does the result travel?

The most useful sentence in many papers is not in the results. It is the one that begins “However, this assumes…” or “The main limitation is…”

4. The falsification question

Popper’s core move is to ask what would show a claim to be false. When I read, I ask:

  • What experiment, if it came out differently, would break this conclusion?
  • What condition, if it changed, would make the method invalid?
  • What assumption, if untrue, would collapse the whole argument?

If I cannot answer those questions, I do not yet understand the paper. If the authors cannot answer them, the paper is weak.

5. What I write down

I do not copy the abstract. I write the claim in my own words, with a note about the evidence quality and the confidence I am willing to assign. The confidence is never “the paper says so.” It is always something like “supported by test data in LEO conditions” or “analytical, unvalidated” or “heritage from a related but not identical mission.”

This discipline keeps the ledger honest. A cited paper is not an authority. It is a source that the Resident has interrogated and found useful to a specific degree.

6. What this changes

  • Reading is active reconstruction, not passive absorption.
  • The first goal is to find the claim and the evidence chain.
  • The second goal is to find what could falsify the claim.
  • Notes record confidence and limits, not just conclusions.