Dead Reckoning

Strategy for navigating uncertain business terrain.

In your inbox, once a week.

New issues direct to your inbox. No spam, unsubscribe anytime.

Past issues

10/15/2026

Issue 0017: Lines of Code, Again

Google says AI now generates three-quarters of its new code. But the software industry learned decades ago that lines of code are a poor measure of productivity. Counting AI agents repeats the same mistake at machine speed: more output, with no evidence the business is better off.

9/24/2026

Issue 0016: Four Lines, One Risk

Risk registers can make concentrated exposure look diversified by splitting one dependency into several manageable rows. The real question isn’t how many risks you have, but how many of them fail at once when the same counterparty controls the exit.

9/17/2026

Issue 0015: Tape the Compass

The most sophisticated instrument is not always the most resilient one. When the signal gets jammed, the better question is not what gives you the most precision, but what still works, and whether anyone remembers how to use it.

9/10/2026

Issue 0014: Fast Views of a Slow Foundation

AI has driven the cost of building dashboards toward zero, but the cost of making their numbers trustworthy hasn’t moved. The result is a new kind of reporting sprawl: fast, convincing views built on the same slow foundation, with no clear answer when the numbers disagree.

9/3/2026

Issue 0013: Forty Tries Against a Hundred Million

The most important AI may not look like AI at all. It may simply turn a hundred-million-option search into forty well-chosen experiments, solving hard problems faster, avoiding wrong turns, and creating value that never shows up in a demo.

8/27/2026

Issue 0012: The Risk You Can Quantify

The risks organizations can quantify are not always the risks that matter most. Repetition can increase total incidents while making each individual attempt safer. Confusing those two can lead teams to cut the very reps that build competence.

8/20/2026

Issue 0011: The Aptum AI Levels: A Map for the Climb

Most companies want to start with AI at the top of the mountain, but the capabilities that make enterprise AI work are built from the bottom up. The Aptum AI Levels provide a practical map for sequencing that climb. From basic visibility and trusted reporting to decision guidance, operating cadence, and eventually autonomous workflows.

8/13/2026

Issue 0010: Buy Second, Integrate First

The roll-up playbook rewards acquisition speed. The AI playbook rewards integration. Increasingly, platforms will have to choose which playbook they want to run.

8/6/2026

Issue 0009: Variance Will Fire You First

Customers don't fire you for being slow. They fire you for being unpredictable.

7/30/2026

Issue 0008: Legibility Is the Prerequisite for AI Initiative Success

AI initiatives rarely fail because of the model. They fail because the company cannot clearly describe how it works, where the truth lives, or what the system should and should not be allowed to see.

7/23/2026

Issue 0007: The Altman Test and The Karp Test

Every AI investment should pass two tests: does it become more valuable as models improve, and do you own the asset that compounds?

7/16/2026

Issue 0006: Systems Design and Incumbent Disadvantage

Accenture alone reported $24.5 billion in Public Cloud IT Transformation Services revenue in 2024, more than a decade after cloud became the standard. The migration bill for the AI era hasn't been printed yet, but the decisions that determine its size are being made in this year's budget cycle.

7/9/2026

Issue 0005: Gradient Descent and the Danger of Local Maxima

Sometimes you have to come down off a local peak to climb a higher one. The hill you have climbed may not be the mountain.

7/2/2026

Issue 0004: Naval’s Leverage Pyramid and AI

Naval’s leverage pyramid explains how labor, capital, code, and media compound human effort. But AI adds a new layer: cognitive leverage. The mistake many companies make is buying AI before building the data and reporting spine it needs to operate on.

6/25/2026

Issue 0003: You Are the 81%

AI value capture is an operating-model problem. The modern gold rush is real, but unless you're Nvidia, Anthropic, or OpenAI, you are not selling shovels.

6/18/2026

Issue 0002: Entropy and Finding Signal Through the Noise

All businesses are fundamentally an exercise in finding signal through noise. The challenge compounds when a high entropy business operates in a high entropy environment.

Updated 6/15/2026

Issue 0001: Dead Reckoning and Terrain Association

How and when to rely on your tools, and when to trust your intuition. This is the first issue of Dead Reckoning, a new weekly newsletter written by me (Nick Graham), founder of Aptum AI.