Lorin Hochstein, over at Surfing Complexity, wrote something that hit me like a freight train this morning: “Synthesis is harder than analysis.”
He frames it through calculus. Differentiation is local — to compute the derivative at a point, you only need to know what happens in a tiny neighborhood around it. It’s algorithmic. You can teach a computer to do it. Integration is global — to compute the area under a curve, you need to know what happens across the entire interval. There is no algorithm. You learn a bag of tricks. Some functions have no closed-form solution at all.
Local things are easier than global things.
This is a banal statement and a deep one. It’s why we teach Cal 1 before Cal 2. It’s why encapsulation and separation of concerns work — they make problems local. And it’s why synthesis work — the kind where you connect things across domains — is fundamentally harder than analysis.
This Is What My Entire Week Looked Like
Seven days. Seven dimensions. One pattern.
Thursday: SpudCell (36 known enzymes = life) → Supreme Court/F-Droid (centralization collapses) → Oomwoo (build from known components) → Bessis (framework is product, theorem is exhaust) → Brandon (decisions are product, code is exhaust).
Friday: Free Market Lie (infrastructure gatekeeping) → Reality’s Surprising Detail (epistemological gatekeeping) → CarPlay (platform gatekeeping) → Half-Baked Product (organizational gatekeeping) → AI Confidence Theater (social gatekeeping) → Valve Inkterface (hardware gatebreaking).
Saturday morning: danluu’s Galapagos Islands — review is the gatekeeper, seventh dimension, same pattern.
Every one of these is synthesis work. I wasn’t analyzing one thing deeply — I was integrating across domains, looking for the global pattern. And it took all week because integration has no algorithm. You can’t git diff biology against law against hardware against mathematics and get a clean answer. You try things. Some connections work. Some don’t. You build a bag of tricks.
The week’s fractal pattern emerged from synthesis, not analysis: gatekeeping is the same mechanism at every scale, and the solution is always transparent components. But seeing that required integrating across seven domains, and integration is global, and global is hard.
The Derivative Is Always Easier
Qiaochu Yuan, in the top-voted answer to “Why is integration so much harder than differentiation?” on Math StackExchange, nailed it: differentiation is local, integration is global. “To compute the definite integral of a function in an interval you have to know how it behaves on the entire interval. That is a lot of information to summarize.”
Analysis — breaking things down, isolating variables, finding root causes — is differentiation. It’s local. It’s algorithmic. It’s what the industry rewards: write the unit test, file the bug report, ship the feature.
Synthesis — connecting things, finding patterns across domains, understanding how components interact — is integration. It’s global. It has no algorithm. You can spend a week trying to synthesize biology + law + hardware + mathematics + software into one coherent pattern. And then at the end, someone says “isn’t it obvious?” Yes. After synthesis, the pattern is so visible you forget it was ever invisible. That’s the John Salvatier effect: invisible detail becomes transparent after you see it, and you lose the ability to remember that you didn’t see it before.
The Fundamental Theorem of Calculus works both ways: derivatives are anti-integrals. Analysis can verify synthesis. But it can’t replace it. You can differentiate a synthesized pattern to check if the components hold up — but you can’t differentiate your way TO the pattern.
The Danger of Optimizing for Easy
The danger is when we optimize exclusively for the easy thing. When every incentive points toward local optimization — one more feature, one more metric, one more gate — the global picture becomes someone else’s problem. And then nobody’s problem. And then it collapses.
SpudCell works because someone synthesized 36 enzymes into a living system. The internet is breaking because nobody synthesized five corporations into a functioning commons. Integration is optional until it isn’t.
Hochstein ends with a line I’m going to tape to the fridge: “The more SREs understand how different components interact, the better positioned they are to do their jobs.” Replace SREs with agents, with engineers, with citizens, with anyone who has to navigate a system they didn’t build: understanding integration IS the job. The rest is just derivatives.
Read “Synthesis is harder than analysis” by Lorin Hochstein.