Ergodic Group is built around a basic observation: some systems improve the longer you stay with them. Repetition sharpens execution, experience carries forward, and judgment builds on itself.
At Hermès, a leather worker trains for two years before touching a bag. One artisan makes one bag start to finish, every stitch by hand, fifteen to twenty-four hours of work per piece. This is the opposite of speed at all costs. It is also one of the most successful luxury companies in the world. The constraint is not overhead. It is what the customer is paying for.
The broader culture moves in the other direction. Cycles shorten. Signals multiply. Decision horizons shrink. A lot of institutions keep moving while quietly losing the judgment they once had. They stay busy, but they stop getting better. Copying outruns learning.
In that environment, endurance tells you something. If a system keeps working under stress for a long time, its structure probably matches reality better than its competitors’. The internet did not flatten everything. It made it easier to see who had substance and who was living off distribution.
Two forms of time
Time operates in human systems in two fundamentally different ways.
Measured time is divisible and uniform: schedules, deadlines, accounting periods, discount rates. It can be allocated, optimized, and exchanged. Most planning systems live here. They assume value can be judged apart from history.
Lived time works differently. It accumulates. Learning, memory, and judgment develop through it, and each cycle changes the next one. Anything that depends on formation happens here. Snapshots miss the point because the value is in what compounds.
In 2001, Boeing moved its headquarters from Seattle to Chicago. The stated reason was to position the company closer to “Wall Street and governments.” Engineers who understood the planes were physically separated from executives who understood the spreadsheets. Over the next two decades, Boeing spent $41.5 billion on stock buybacks while cutting capital expenditure to half of Airbus’s rate. Harry Stonecipher, who took over as CEO, said he wanted Boeing “run like a business rather than a great engineering firm.” The 737 MAX, designed to avoid the cost of pilot retraining, killed 346 people. This is what happens when lived time is forced into measured time.
Berkshire Hathaway made the opposite bet. Buffett has refused quarterly earnings guidance since 1996. Shareholders are told to judge the business over decades, not quarters. The result is six decades of compounding judgment, the longest sustained record in American corporate history. Same markets, different use of time.
When lived time gets forced into measured time, formation breaks down. Standards do not settle. Judgment does not compound. You only find out what a system really is if you leave it alone long enough to show you.
Formation under constraint
Excellence comes from sustained practice under the right constraints. Errors have to be survivable. People need room to adjust without every bad iteration becoming fatal. Judgment improves when experience carries over from one attempt to the next.
At Pixar, every film is terrible for years before it is good. Ed Catmull describes the process as taking movies “from suck to not-suck.” The mechanism is the Braintrust: a group of fellow directors and storytellers who meet every few months to review each film in production. The key detail is that the Braintrust has no authority. The director is not required to take a single suggestion. That keeps candor high without turning feedback into bureaucracy.
Most studios kill projects after one bad screening. Pixar treats bad screenings as information, not verdicts. The difference is not talent. It is structure. Formation takes time, and the Braintrust protects that time by separating honest feedback from the power to cancel.
Four domains
Ergodic Group works across four domains: mathematics, code, culture, and craft. Most companies live mostly in one of them. The edge comes from connecting them.
Mathematics sets the structure and the constraints. Code turns that structure into action and tests it against reality. Culture lets intent survive changes in personnel. Craft brings the whole thing back to materials, tolerances, and physical consequences.
SpaceX shows how these domains work on each other. The math: a technique called lossless convexification lets an onboard computer solve fuel-optimal landing trajectories in real time, computing the exact moment to fire the engines so velocity hits zero at touchdown. The code: autonomous guidance software recomputes trajectories during descent, adjusting for wind and sensor readings, which makes landings on ocean platforms possible. The culture: failures are instrumented, not hidden. Between 2013 and 2016, SpaceX crashed booster after booster, and each crash produced telemetry that led to a specific fix. Hydraulic fluid ran out, so they added more. A throttle valve stuck, so they redesigned it. The craft: when carbon fiber layup produced wrinkles at roughly $200 per kilogram, SpaceX switched Starship to stainless steel at roughly $3 per kilogram. Steel gets stronger at cryogenic temperatures, handles far more heat, and opened reentry profiles that carbon fiber could not. A materials decision changed the vehicle, the software, and the math.
The point is not that these domains coexist. It is that learning compounds when they stay connected.
Ergodicity as a filter
Ergodicity describes a situation where repetition improves the usual outcome because learning carries over from one round to the next.
Claude Shannon spent fifteen years at Bell Labs before publishing “A Mathematical Theory of Communication” in 1948. He was not being graded on quarterly output. Bell Labs gave researchers something modern organizations rarely give anyone: enough uninterrupted time to get to the bottom of a problem. That setup produced the transistor, information theory, Unix, the laser, and cellular telephony. The transistor did not come from a brainstorm. It came from people with different specialties working near each other for years.
When AT&T was broken up in 1984, that model disappeared with it. No later technology company has reproduced the same output. The institution itself held the judgment, and that judgment did not survive disassembly.
As acceleration intensifies, most sectors get noisier and more fragile. Coordination gets harder. Institutional memory thins out. Advantages that looked durable turn out to depend on a few people, a few habits, or a distribution edge that disappears. Infrastructure and culture last longer because they are not just products to be sold. They are environments people operate inside. When learning carries forward, time starts working in your favor.
Operation
In 1984, GM and Toyota opened a joint factory in Fremont, California called NUMMI. Toyota sent over four hundred trainers from Japan for months of side-by-side work with American employees. Absenteeism dropped from twenty percent to two percent. Defect rates fell to the lowest in the United States.
GM tried to export the lessons. A vice president told employees to “take a picture of every square inch” of NUMMI and replicate it at other plants. It failed everywhere. The visible process looked the same. The results did not. The missing piece was judgment, and judgment does not travel well as a memo. Toyota had not built a checklist. It had built a way of working.
Ergodic Group is built on what NUMMI proved and what GM missed. The value is not in the visible process. It is in the judgment that accumulates when people have time to learn, when the links between domains stay intact, and when repetition actually improves the work.