Skip to main content

First Principles vs. AI

Why This Matters: James and the Empty Toolbox​

James looked at the exercise description and pushed back from the table. "Forty-five minutes. No AI. No internet. Solve a resource allocation problem from scratch." He looked at Emma. "There are entire textbooks on fair allocation. Operations research. Game theory. Why would I rederive something people already figured out?"

"What's the standard approach?"

"I don't know the exact algorithm, but there are frameworks. Weighted scoring, needs-based tiers, lottery systems. I'd look them up, pick the best fit, and adapt it."

"And if none of them fit?"

"One of them always fits. That's the point of having frameworks."

Emma pulled out a blank sheet of paper and set it in front of him. "The scenarios in this exercise are designed so that no standard framework fits cleanly. The constraints conflict with each other. Equal access conflicts with limited supply. Need-based allocation conflicts with practical logistics. You can't search your way to a clean answer because the answer depends on which constraints you prioritize, and that's a judgment call, not a lookup."

James stared at the blank sheet. "So you're telling me to spend forty-five minutes reinventing the wheel."

"I'm telling you to find out whether you can build a wheel. Or whether you've only ever installed one someone else designed." She paused. "In my first year as an engineer, I could configure any framework. Kubernetes, Terraform, whatever the team was using. But when a client had requirements that didn't match any existing architecture, I froze. I'd never designed from constraints before. I'd only ever selected from a menu."

"That's like our vendor evaluation process," James said slowly. "We had a scoring matrix. Worked great when the vendors were similar. But when a vendor came in with a completely different model, something that didn't fit any of our categories, we didn't know how to evaluate them. We just scored them low on every dimension because the framework couldn't see what they were offering."

"Exactly. The framework becomes a filter that blocks what it can't categorize." Emma tapped the blank sheet. "Forty-five minutes. Start with what's undeniably true about the problem. Those are your constraints. Then list every assumption you're making. Then build from there."

Exercise 2: First Principles vs. AI​

Layers Used: Layer 1 (Predict Before You Prompt), Layer 5 (Divergence Test)

James is sitting with a blank sheet of paper and a problem that has no clean answer. So are you.

Choose Your Scenario​

Scenario A (Education): "Design a fair system for distributing limited AI tutoring access across a school district of 200,000 students with wildly unequal resources."

Choose one.

Derive Your Solution (No AI, 45 Minutes)​

You receive a problem with no established solution. No AI, no internet. 45 minutes. Identify the base constraints, list your assumptions, and derive a solution from those constraints alone. After submitting, prompt AI with the same problem and compare.

Your Deliverable

A First Principles Worksheet containing: (1) the base constraints you identified (e.g., limited supply, unequal need, multiple definitions of fairness), (2) every assumption you made, explicitly listed, (3) your derived solution with a clear logical chain from constraints to design, (4) a comparison document showing your solution alongside AI's solution, with annotations on where they converge and diverge.

Check Your Thinking​

1Your Work

I solved a novel problem using first principles reasoning without any AI or internet assistance.

The problem I chose:

Below is my First Principles Worksheet. Please: (1) Evaluate my constraint identification -- did I find the real base constraints or did I miss critical ones? (2) Review my assumption list -- which assumptions are reasonable and which are questionable? What hidden assumptions did I not list? (3) Does my solution logically follow from my stated constraints, or are there gaps in the derivation? (4) Rate my solution's originality -- is this something you would generate if prompted directly, or does it show genuine independent reasoning? (5) What is the single biggest flaw in my solution that I need to address? (6) Now solve the same problem yourself. I will compare our approaches.

My worksheet:

Finally, complete the Thinking Score Card for this exercise: Independent Thinking (1-10), Critical Evaluation (1-10), Reasoning Depth (1-10), Originality (1-10), Self-Awareness (1-10). For each score, give a one-sentence justification.

2Get Your Score

Discuss with an AI. Question your scores.
Come back when you have your BEST evaluation.

Deliverable Template (click to expand)

FIRST PRINCIPLES WORKSHEET

  • Problem: [paste]
  • BASE CONSTRAINTS:
    1. ___ (things that are true regardless of approach)
    2. ___
    3. ___
  • MY ASSUMPTIONS (explicit):
    1. ___
    2. ___
    3. ___
  • DERIVATION CHAIN: From constraint [#] + assumption [#], it follows that: ___
  • Therefore, the solution must: ___
  • MY SOLUTION: ___
  • WHY this follows from the constraints (not from analogy or pattern): ___
  • WHAT I DO NOT KNOW: ___

What Happened With James​

James set his First Principles Worksheet next to Claude's solution and read them side by side. His solution prioritized equity: the students with the fewest resources should get the most access. Claude's solution prioritized efficiency: allocate based on predicted learning gains per hour of tutoring time. Both were logical. Both followed from their stated constraints. But they optimized for completely different things.

"Wait, so basically... we started from the same problem and ended up in different places because we prioritized different constraints?" James said.

"What does that tell you about the standard frameworks you wanted to look up?"

James thought about it. "They'd all bake in a prioritization. I just wouldn't know which one. I'd adopt their judgment without realizing it was a judgment."

"Now you know what you actually think fairness means in this context. Not what a textbook says it means. Not what Claude says it means. What you derived when nobody else was in the room."

James looked at his worksheet. It was rough. Some of the derivation steps had question marks in the margins. But the core argument was his, and he could trace every piece of it back to a specific constraint. That had never happened when he used someone else's framework.

The Lesson Learned​

Frameworks encode someone else's priorities without labeling them as priorities. When you derive from constraints yourself, you discover what you actually value, not what you inherited. The roughness of a first-principles solution is a feature: every seam is visible, every tradeoff is traceable, and you can defend every choice because you made it deliberately.

Flashcards Study Aid​