Nuclear Stewardship
The Custodianship Mandate
You may own a knowledge-preservation mandate: deep expertise is retiring out the door faster than it can be handed down. Most solutions I have seen focus on:
- Knowledge capture, such as with Confluence or a Wiki
- Knowledge transfer, such as video interviews, or formal presentations with slide decks
However, the route to mastery is not linear. The route from beginner through intermediate to expert is clear. It is based on both knowledge and experience. But the master of the craft who "makes it look easy" operates on a different plane entirely. A master inhabits the invariants, the timeless skills.
This was the situation within Cray Research. We were forced to inhabit the invariants because often the technology did not yet exist. We were creating it. We focused on, and used, those timeless skills. We rarely used analysis tools because they got in the way of understanding the problem. A tool is limited to what its creator knew at the time of creating it, which is useless for working with new technology that has outrun the tool. But the invariants still apply.
The solution is not knowledge transfer. It is formation of thinking. That is, learn to think like the master of the craft. This is why apprenticeship traditions exist.
I take my readers/students through the following stages. Chapter 3, "Apprenticeship Transmission," provides more detail and citations to the theory;
- Adequate motivation. Establish (with the learner) why the action matters to the learner.
- A complete and correct Orienting Basis of Action. What to pay attention to, and in what order, to guarantee correct performance before the learner acts. This step is generally skipped with knowledge transfer sessions, however, this is the primary designed instruction.
- Systematic passage through the stages. Guide the learner in moving the action deliberately from material form, through speech (or reading), to the mental plane rather than jumping straight to the abstract instruction. Perform the action before naming and internalizing it.
- Step-by-step control and feedback (correction) at every stage. In the written situation with the master not present, I walk my readers through the master's thought process. Thus the process is expertise transfer rather than knowledge transfer.
Chapter 1. Incorrect By Design
"Hallucination" is an unfortunate and misleading term for Artificial Intelligence. It obscures the fact that most people think about AI incorrectly. AI is not deterministic like a spreadsheet or word processor. It chooses the "most probably correct" response with no concept of "actually correct." We need to look past the surface to understand the structure.
Chapter 2. When Performance is Everything
My role is to transmit our tradecraft as practiced within Cray Research. We accomplished, repeatedly, what nobody else on the planet accomplished. I have written down how we did it, as an apprenticeship. Apprenticeships are not easy.
Chapter 3. Apprenticeship Transmission
Soviet tradition describes a methodology that matches internal Cray Research practice. Soviet psychologist Dr. Piotr Galperin described the conditions and stages for formation of thinking, that is, how to acquire expertise in a field rather than facts and other knowledge.
Chapter 4. Teaching and Creating Mastery
Separate concept from implementation. Extract the timeless invariants. Demonstrate and teach them. Model your thought process out loud for your students, thus transferring your expertise rather than sharing knowledge.
Chapter 5. Transcendent Patterns: Teaching the Process of High-Tech Mastery in Student-Accessible Fashion
This chapter teaches the skill of identifying the timeless patterns and skills, the invariants. Spotting a pattern in the wild requires deliberate practice.
Chapter 6. If You Want More
At this point you will have successfully shifted your thinking. This chapter completes the first cycle. You will not have mastered anything by merely reading about it, but you will now have names for skills you already use. By the end of this chapter, you will know whether or not you care to invest time and effort into the two larger books. This book teaches the teaching method; the two larger books use this method to teach the tradecraft.
Appendix A. Analytical Machine Employment (1952)
We currently (in 2026) use AI as a labor-saver. We should be using AI to solve previously unsolvable problems. The NSA addressed this same issue in 1952.
Appendix B. Cryptanalytic Machines in NSA (May 1953)
This chapter describes the Top Secret machines according to their potentiality.
Appendix C. HPC Tradecraft Road Map
This is the road map to teaching High-Performance Computing tradecraft as practiced within Cray Research.