first-principles

First Principles · 第一性原理

One-sentence positioning: Break the problem down to physical/logical facts that cannot be broken down further, and re-derive from there — escaping the constraints of "analogical thinking." Musk used it to cut rocket costs by 90%.

When to Use (TRIGGER)

• Constrained by "industry conventions" or "everyone does it this way"
• New industry / new problem with no analogous precedent
• Existing solutions are unreasonably expensive (intuition says "it shouldn't be this expensive")
• Want to pursue disruptive innovation rather than incremental optimization
• The user says "Why can't we do X?" — a signal that they are challenging default assumptions

When Not to Use (ANTI-PATTERN)

• The problem already has a perfect best practice (Cynefin Clear domain) — use the SOP, don't re-derive
• Time is tight — first principles requires deep derivation (hours to days)
• High emotional involvement — cool down first, otherwise you will use "physics" to rationalize preferences
• The team is unwilling to challenge authority — first principles will trigger cultural conflict; psychological safety must be established first

Steps (STEPS)

1. Write down the current conclusion / current approach

Clearly state "what the status quo is." Example: "A rocket costs $65 million each"

2. List every implicit assumption supporting the conclusion

Force them to be made explicit. Example: "Rockets are expensive because: a) materials are expensive b) manufacturing is expensive c) single-use is required d) supply chain is limited"

3. For each assumption, ask: Is this a "physical/logical fact" or an "industry analogy"?

• Physical fact: chemical reactions, laws of physics, human physiology, mathematical inequalities
• Industry analogy: because everyone does it this way, because of tradition, because of historical compliance
• Golden question: "Would this assumption still hold 100 years ago / in another civilization / in the eyes of alien intelligence?"

4. Strip away all non-physical facts, leaving only physical fundamentals

Obtain the "physically possible solution space" — usually much larger than the current status quo

5. Rebuild the solution from the fundamentals

Don't look back at the status quo; design from zero

6. Compare the physical-fundamentals solution vs the analogical solution

• The gap = the cost brought by analogical inertia = the return of first principles

Decision Checklist

• I listed at least 5 implicit assumptions
• I labeled each assumption as "physical / analogy"
• I admitted that at least 2 assumptions are "analogy rather than physics"
• I redesigned the solution from physical fundamentals (not optimized the status quo)
• I can explain the specific reasons why "physics allows it but the industry hasn't done it" (regulation/capital/cognition/path dependence)

Real Cases

Case 1: SpaceX rocket cost breakdown (2002)

• Status quo: A Russian rocket costs $65 million each
• Implicit assumptions: a) Materials such as titanium, aluminum, copper, and iron are expensive b) manufacturing processes are expensive c) single-use is physically determined
• Physical facts: Break down to raw materials — aluminum, steel, copper, platinum, fuel. Raw material cost < $2 million (total rocket cost < 2%)
• Conclusion: The 63 million difference is caused by the 'industry profit chain + single-use', not by physics
• Rebuild: Build in-house + reusable. Falcon 9 single launch 62 million → marginal cost after reuse < 10 million

Case 2: Tesla battery pack cost (2008-2018)

• Status quo: industry 600 USD/kWh
• Physical breakdown: lithium + cobalt + nickel + aluminum + processing = 80 USD/kWh (actual raw material + processing cost)
• Source of the gap: insufficient scale + industry profit + fragmented upstream and downstream
• Rebuild: Gigafactory centralized ultra-large scale + vertical integration. By 2024 it dropped to the order of 100 USD/kWh

Case 3: An engineer's local testing

• Status quo: 'You can't run this integration test locally, company process requires CI'
• Breakdown of implicit assumptions: What does CI provide? — a) database b) network c) credentials d) reports e) review
• Physical facts: a-d can all be reproduced locally (docker database / local mock network / .env credentials file / local HTML report); only e review requires CI
• Rebuild: a local one-click script sets up a-d, tests finish in 5 minutes; finally push a PR to CI to run e
• Effect: the development-test loop drops from 30 minutes to 5 minutes

Relationship with other frameworks

• Inversion is 'enumerating the negative', First Principles is 'rebuilding the positive' — the two ways of thinking complement each other
• First principles are especially important in the Cynefin Complex/Chaotic domains(because there is no ready-made pattern to apply)
• Second-Order Thinking evaluates the long-term feasibility of a first-principles solution

Common pitfalls

1. Mistaking 'I'm not familiar with it' for 'physically impossible': something you don't understand doesn't mean it's impossible — countermeasure: first ask a domain expert, 'Is this a physical limitation?'
2. Breaking it down to the molecular and atomic level(overdoing it) — countermeasure: breaking it down to the level of 'can be procured / can be synthesized / can be verified' is enough
3. Secretly using analogy again during the rebuild(self-deception) — countermeasure: have an outsider review your 'physical fundamentals'
4. Ignoring social fundamentals: law, compliance, trust, and network effects are also 'fundamentals', not just physics — countermeasure: treat social fundamentals as 'non-decomposable constraints' too

References

• Aristotle, Posterior Analytics (ancient Greek original text)
• Elon Musk, 2013 TED interview ("First Principles Thinking")
• Tim Urban, Wait But Why: "The Cook and the Chef: Musk's Secret Sauce" (2015)
• Ozan Varol, 《Think Like a Rocket Scientist》(2020)