Takeaways: Lessons learned during the session

The Planner, the Street, and the Missing Signal

     In 1955, Robert Moses looked at the West Village (irregular, dense, alive with corner shops and mixed-income households) and saw a problem. He proposed a clean superblock in its place. Jane Jacobs lived there, watched what his planned environments actually produced, and spent the next decade explaining why he was wrong. Moses had more power, more resources, more certainty. Jacobs was right.

     Her insight was deceptively simple: the West Village wasn't working despite its complexity, it was working because of it. Safe streets weren't patrolled - they were watched, by the baker who opened at six, the bar owner who knew every regular, the resident who simply happened to be present. No blueprint had specified this. It emerged from mixed uses, short blocks, and overlapping human rhythms. When Moses replaced that texture with single-use superblocks, the informal network of attention dissolved. Pruitt-Igoe, St. Louis: completed 1956, demolished 1976. It was a human system collapse.

     During our discussion a framework was offered, which made the mechanism precise: every system can be understood through four elements: its parameters, its feedback mechanisms, its design, and its underlying assumptions. The parameters were defined. The design was meticulous. The assumptions were sincere. What was missing was feedback - no mechanism for the people inside to signal what was working. A system without that signal cannot learn. It executes its initial assumptions indefinitely, regardless of what reality is doing. The failure wasn't bad design. It was design that could not adapt. The same pattern appears in any organisation that runs annual surveys instead of live signals - by the time the data arrives, the damage is already structural.

Power, Constraints, and What Cannot Be Moved

     This opened a deeper question: which constraints are actually changeable? Every system exists inside other systems, and some of those outer layers are simply not moveable. Gravity is not negotiable. Certain laws shift only over decades. And even where change seems rationally justified, it can run into something that no blueprint accounts for, which is power.

     The French ban on a particular method of execution was raised as a striking case: a law that reflected clear intent, passed through formal process, yet met resistance that had little to do with majority opinion and everything to do with where structural power actually sat. This connects to what Vilfredo Pareto observed, that in most systems, a small minority holds disproportionate influence over outcomes, regardless of what the majority wants. Kenneth Arrow formalised the deeper problem: in any system with multiple competing preferences, there is no reliable method of aggregating them into a consistent collective decision. Majority rule sounds clean. In practice, who defines the options, controls the timing, and holds the veto shapes outcomes far more than the vote itself. Any founder who has tried to change a company's direction by consensus has felt exactly this.

     The implication for design is uncomfortable: building a system around what seems rational or widely supported does not guarantee it will work as intended. The outer systems will shape what is actually possible, often invisibly.

The Uncarved Block, Simplicity, and the Trap of Precision

     Complexity compounds further when systems don't just nest but cross. A change in one propagates into others in ways that were never modelled. This is why simplicity can be its own failure mode. Reducing a system to its cleanest form often strips out the redundancy and overlap that made it resilient. What looks inefficient from the outside is frequently what absorbs shocks, carries informal knowledge, and allows recovery when something breaks. The company that eliminates its informal communication channels in favour of a clean taxonomy often discovers, too late, that the noise was doing work the structure never could.

     Taoist philosophy has a concept that maps onto this precisely. P'u uncarved block holds all possible forms before the first cut. Every cut removes potential. The craftsman's skill is not boldness but restraint: understanding what the material wants to become, and removing only what prevents it. Applied to human systems, over-specification is not just inefficient, it is destructive. When a structure makes every decision in advance, the people inside stop making decisions. Compliance grows. Judgment atrophies. What gets produced is exactly what was specified, which is rarely what was actually needed.

     A distinction that emerged sharpened this further: the difference between holding a system to a range and holding it to a fixed point. A system designed to maintain a value between five and one hundred can self-correct naturally. A system required to hold exactly seventy-five demands continuous intervention just to stay in place. The tighter the specification, the more maintenance it requires indefinitely. What looks like control is often just deferred instability, accumulating quietly until the system can no longer hold its own shape.

Stable Core, Open Edges

     As an engineer in the room put it: a well-designed system is stable at the core: closed to the changes that would destabilise what everything depends on, and open at the edges, decoupled enough that growth can happen without permission from the centre. It is an engineering principle derived from the same logic as Jacobs's streets: specify the minimum stable foundation, and leave the rest genuinely open.

     The critical caveat raised: this does not transfer cleanly between systems. What works in a distributed codebase does not map 1:1 onto a team, a city, or a policy. Decoupling in software means components can change independently. In an organisation it might mean protecting team autonomy while keeping strategic alignment tight - similar diagram, entirely different human dynamics. The principle travels but the implementation must be reinvented each time.

You can design the garden. You cannot design what grows.

     What every domain in the room eventually arrived at was the same distinction: between designing for a specific outcome, and designing the conditions in which the right outcomes become possible. The first treats the system as a machine with a predictable output. The second treats it as a living environment, one that needs structure to function, but space to become. The designer's job is not to fill that space. It is to protect it.

On Balanced Theory and Why It Keeps Slipping

     This session brought together a remarkable domain variety in the same room. The frameworks landed well and generated genuine engagement and valuable insights across backgrounds. And yet something kept not quite happening: the conversation stayed in the domain of theories, systems, and structures. The more personal register, the kind where someone describes a specific moment from their own life, remained mostly out of reach.

     This is a pattern that has now repeated across multiple sessions, and it deserves an honest name. When opening examples are structural or organisational, participants calibrate to that register and stay there. The group takes its cue from what feels acceptable to share, and if the room has established an analytical tone early, vulnerability reads as a departure from the implicit contract. Amy Edmondson's research on psychological safety is precise here: people do not take interpersonal risks unless they have clear evidence it is safe to do so. An intellectually safe room is not the same as a personally safe room. What gets lost when it stays abstract is specific: personal examples introduce texture, contradiction, and emotional specificity that theory cannot carry alone. They make the pattern feel true rather than just logical.

     The desired state is to have both present simultaneously, rigorous frameworks and genuine personal grounding, because that is where the most original and resonant insights tend to emerge. Achieving that balance has proven repeatedly difficult. Edgar Schein's work on psychological safety in groups suggests the facilitator must model vulnerability first, not perform it, but offer a genuine moment of personal uncertainty before asking others to do the same. The opening, not the middle, is where the balance gets set. That is the hypothesis to test next time.