In every system — from engineering to economics to life itself — there are three desirable qualities you can never fully achieve at the same time. Pick any two. The third suffers. Always.
The original iron triangle. Want it fast and good? It won't be cheap. Fast and cheap? Quality drops. Good and cheap? It takes forever.
Bitcoin: secure + decentralized but slow. Solana: fast + secure but more centralized. Ethereum tries to balance all three with L2s.
GPT-4 is accurate but slow and huge. A tiny model is fast and small but less accurate. You can't have all three — distillation is the art of choosing where to compromise.
A country can only have two of three: China picks fixed rate + monetary independence (restricts capital). US picks free capital + independence (floating rate).
Brewer's theorem (proven 2002): a distributed system can guarantee at most two of three. MongoDB picks CP. Cassandra picks AP. Network partitions are unavoidable, so you always sacrifice C or A.
A Tesla Model S is fast + efficient but not affordable. A Corolla is efficient + affordable but not fast. A Lamborghini is fast + premium but not efficient.
Soft compound = more grip but wears fast. Hard compound = durable but less grip. Low resistance = fuel efficient but less grip. F1 teams choose compound per race — there's no universal answer.
US: high quality, high cost, limited access. UK (NHS): universal access, low cost, sometimes lower quality/longer waits. No country has solved all three.
More transistors = more performance but more power and cost. Apple's M-series chips optimize power efficiency by accepting larger die sizes. Intel historically chased raw performance.
Healthy + tasty (organic salmon) isn't cheap. Tasty + cheap (fast food) isn't healthy. Healthy + cheap (beans and rice) isn't always exciting.
High quality + free speech (early Twitter) had low engagement signals. High engagement + free speech breeds toxicity. High quality + engagement requires heavy moderation (restricting speech).
Fast growth usually requires VC funding (losing control and profitability). Profitable + controlled (bootstrapped) grows slower. Fast growth + profitable is extremely rare — those are unicorns.
Solar is clean + increasingly cheap but not reliable (intermittent). Gas is reliable + cheap but not clean. Nuclear is clean + reliable but not cheap (upfront cost).
Highly secure + private (end-to-end encryption) is hard to use. Secure + usable (biometrics) sacrifices privacy. Private + usable (anonymous browsing) is harder to secure.
Deep + broad education (PhD + MBA) takes forever. Deep + fast (bootcamp) is narrow. Broad + fast (survey courses) is shallow. Learning has a fixed budget of time.
Efficient + stable autocracies sacrifice democracy. Democratic + stable systems are often slow. Efficient + democratic systems can be unstable (populist swings).
Wide aperture (bokeh) + fast shutter needs high ISO (noise). Low ISO + fast shutter needs wide aperture (shallow depth). Every photo is a three-way trade-off.
Ambitious career + deep relationships leaves no time for health. Career + health (gym, sleep) means less social time. Health + relationships means career takes a backseat.
The universal college joke — pick two. Study + socialize = no sleep. Study + sleep = no social life. Socialize + sleep = bad grades. Every student knows this.
Central + spacious = expensive. Central + affordable = tiny. Spacious + affordable = far from the city. This is why housing is a crisis everywhere.
Every system operates with limited energy, time, money, or attention. Allocating more to one dimension means less for others. This is conservation of energy applied to design.
Once you're on the Pareto frontier, you can't improve one dimension without worsening another. Most real-world systems are already near their Pareto boundaries — that's why trade-offs feel inevitable.
The second law of thermodynamics: you can't create order in one place without creating disorder elsewhere. Every optimization has an entropy cost. There are no free lunches in physics.
In complex systems, variables are coupled. Optimizing one creates emergent effects on others. The more complex the system, the more trade-offs it contains. Simplicity is itself a trade-off against capability.
New tech expands the frontier. EVs broke the performance/efficiency trade-off — but introduced range/cost.
Carbon fiber is strong + light — but expensive. Graphene promises to break constraints but isn't scalable yet.
Transformers "broke" the accuracy/speed trade-off for NLP. But they introduced the compute/cost constraint. New triangle, same law.
You can have all three — just not at the same time. Build quality first (slow + expensive), then optimize cost (economies of scale), then speed (process maturity).
Sometimes the triangle is wrong. AirPods aren't cheap or hi-fi — they redefined "convenience" as the dimension that matters. Change the axes, change the game.
The deepest wisdom: stop trying to break the triangle. Choose your two dimensions deliberately. Make the sacrifice explicit. That's strategy.
"战略不在于你选择做什么，而在于你选择不做什么。"
— Michael Porter