Originally coming from physics the transition into founding startups requires finding a new balance between rigor and empiricism, between abstraction and concretion, between reductionism and contextualism. Physics is a beautiful set of reductionism, abstraction and rigorously planning everything through. It is reductionist by design: strip away noise, isolate variables, derive laws from what is left. During experiments physicists do their best to control the environment. The observer stands outside the system. Experiments are repeatable. For startups it is almost antagonistic: You are in a noisy uncontrolled environment, where timing matters, the context of your environment matters and you are more the player than the observer. You must learn to select and set up the optimal environment for your startup to thrive in.
Here is a non-exhaustive list of lessons I learned.
Lesson 1
In physics you plan a sequence of steps to reach one of a limited amount of outcomes. During each step you know where you are. In a startup you don't. The rules change under your feet, you have many allies and competitors - some you know about, some you do not - and everyone has a different starting point, different plan, different skills, different everything. You have an infinite amount of possible paths and yet the two main options are: your startup dies, or it survives. Don't die.
The shift is from guaranteeing outcomes to maximizing the probability of the one you want. You transition from chess-to-poker: from deterministic planning to expected-value optimization under uncertainty. Actually several great scientist I met over the years were superb chess players, and many of the most savvy entrepreneurs I met turned out to be highly skilled poker players.
Now each startup is a different game, but the currencies of success are the same. There are several games running simultaneously, and your job is to find the Nash equilibrium across all of them:
Your startup does not “solve” a problem. It renegotiates the equilibrium across these six games. A pitch is not a physics talk - it is a credible signal that you have identified where the equilibrium shifts in your favor.
Lesson 2
This one is straight out of Thiel's Zero to One. You are trained on Gaussian errors and ensemble averages. The business world is Pareto-distributed and non-ergodic.
Your first ten customers are not a sample of the next ten thousand. They are a different regime - early adopters obey different psychological laws. Geoffrey Moore's Crossing the Chasm is a model to think about here. Interpolating from early adopters to the mainstream is fatal.
VC returns follow the same law: one bet returns the fund, nine die. You cannot diversify your startup's strategy to reduce variance. What you can do is maximize convexity - take bets with fixed, small downside and unbounded upside. Ship a scrappy MVP to a niche. But do not try to build a perfect product for “the market.”
Talent too is power-law distributed. One engineer may be ten times more productive than another. Your hiring process is a search for outliers.
Practical consequence: replace every dashboard average - average revenue per user, average churn - with a Lorenz curve: Who generates what percentage of revenue? Almost certainly three whales followed by ten thousand comparative free riders. Design for the whales, not the mean - but beware of becoming too dependent on them.
Lesson 3
Donella Meadows' stock-flow dynamics apply directly to companies. The stocks that matter most are not the obvious ones. They are:
Morale is a stock with no natural equilibrium - it decays unless actively replenished. Your real runway is not just cash - morale is converted into hours of productive work.
There are two kinds of feedback loops determining your survival:
Your job is to architect one strong R-loop and cut every B-loop before it compounds. Everything else is noise.
Lesson 4
In physics, the system boundary is a clean methodological choice. In a startup, it is a competitive claim - and the market can revise it without asking you.
You define your market as mid-market SaaS. A competitor launches a free tier and redraws the boundary to include individuals. Your ten-person sales team is now optimized for the wrong problem.
The right frame: define your system boundary not by customer segment but by feedback speed. The company that closes the loop fastest - observe, orient, decide, act - defines the category. If improvement compounds, the fastest loop wins. Everyone else is catching up.
Measure your OODA loop concretely. If it takes you two weeks to ship a user request and a competitor takes two days, your market boundary is already dissolving.
Lesson 5
This is the sharpest break from physics. In a lab, the observer stands outside the system. In a startup, you are inside it, and the system optimizes around you.
Raising a large Series A creates the expectation that you will hire thirty people. Hiring thirty people dilutes culture. Diluted culture slows shipping. Slower shipping means missed milestones. Missed milestones mean a down round. The capital created the failure mode.
Announcing AI features creates the expectation of magic. Overpromising leads to shipping buggy code. Buggy code collapses trust. Your marketing became the product loop.
Every action you take is not a move - it is a signal that rewrites the incentive structure for every other actor in the system. You are not playing a game; you are negotiating the game's rules in real time.
The skill this demands is second-order thinking. Not “what happens if I do X?” but “what happens to the incentive structure after I do X, and how will each actor re-optimize in response?”
Physics gives one powerful tools. Most of them have analogues - but the translation is not obvious.
Before writing any code, draw a single causal loop diagram with three stocks (user trust, team morale, investor confidence), two R-loops (one you want, one you fear), and one B-loop that will kill you in eighteen months.
If you cannot map it, you do not understand the system well enough to survive in it.
Physics teaches the universe's indifference. Startups teach you the universe is adversarially indifferent - it will exploit any model you mistake for reality. The defense is not a better model. It is the discipline to keep updating the one you have.
LFG!