Imagine an emergent alternative to GasTown in the sharper, oppositional voice of Jane Jacobs, with Robert Moses as her foil.
ChatGPT Prompt (condensed)
1. The Kind of Order We Keep Trying to Build
There is a persistent habit among planners—whether of cities or systems—to assume that order must be imposed because reality is too messy to be trusted.
They look at the apparent disorder of:
- conversations
- exceptions
- complaints
- local adjustments
- human improvisation
and conclude:
this must be replaced with something cleaner, more controlled, more rational.
This is the mistake.
2. Moses’s Error, Repeated
Robert Moses did not fail because he lacked intelligence or ambition.
He failed because he mistrusted the very thing that makes systems work:
the accumulated, distributed, and living context of those inside them.
He replaced it with:
- plans
- authorities
- structures
- top-down coherence
And in doing so, he achieved a certain kind of clarity.
But at the cost of:
- local knowledge
- adaptability
- and the subtle, continuous adjustments that make life possible
3. The Same Error in Digital Systems
We repeat this error constantly.
We see:
- Slack threads full of confusion
- support tickets revealing contradictions
- calls where reality diverges from the plan
- code that cannot be changed without breaking something unseen
And we say:
“This is noise. This must be cleaned up.”
So we build:
- stricter workflows
- clearer ownership
- better abstractions
- tighter control
And each time, we lose something essential.
4. What We Keep Throwing Away
What we call “messy” is precisely the thing that holds the system together.
Those Slack threads, those tickets, those conversations—they are not peripheral.
They are:
the only place where the system explains itself
They contain:
- why something cannot change
- why something was done that way
- what failed before
- what must not happen again
Remove them, or ignore them, and you are left with a system that looks orderly but is in fact blind.
5. The Real Constraint
The reason planners like Moses impose order is not because they are wrong about complexity.
It is because:
in physical systems, context is fragmented beyond recovery.
No one can see:
- all the conversations
- all the local adjustments
- all the reasons
So they replace context with control.
And under those constraints, this is often rational.
6. The New Possibility
But in a digital system, this constraint no longer holds.
We are not limited to:
- memory
- proximity
- rumor
We can, in principle, make the full messy context:
- visible
- indexed
- queryable
Not simplified.
Not cleaned.
But:
accessible
7. WadTown
WadTown is what happens when we refuse to make Moses’s trade.
We do not say:
- “replace context with structure”
We say:
“make context queryable.”
8. The Two Kinds of Order
There are two kinds of order:
8.1 Imposed order (Moses)
- rigid
- planned
- maintained from above
- brittle under change
8.2 Emergent order (Jacobs)
- adaptive
- local
- continuously adjusted
- sustained from within
The first replaces context.
The second depends on it.
9. What WadTown Actually Does
WadTown does not eliminate mess.
It does not resolve contradictions.
It does not simplify the world.
Instead, it ensures that any worker can ask:
- Why is this the way it is?
- Who said so?
- Where does it apply?
- When did it become true?
- How do we know?
And get an answer grounded in the real, messy history of the system.
10. Why This Changes Everything
Once context is queryable:
- workers do not need to be controlled
- plans do not need to be perfect
- coordination does not need to be enforced
Because the reasons for action are no longer hidden.
They are:
available
11. The Collapse of Moses
If context is accessible, the justification for imposed order weakens.
Not because people become better.
But because:
they no longer need to guess.
And when people do not need to guess, they do not need to be tightly controlled.
12. The Manifesto
We reject the assumption that:
- mess must be eliminated
- context must be simplified
- order must be imposed
We assert instead:
- that the mess is where meaning lives
- that context is the system
- that trust comes from provenance, not tidiness
- that workers, given access to real context, will act coherently
13. The Miracle
The miracle is not that the system becomes orderly.
It does not.
The miracle is that:
the full, messy, contradictory, human context
can be made visible enough that people can act within it intelligently.
14. Final Line
Moses built systems that worked despite people.
WadTown builds systems that work because people can finally see.
Core Concepts & Resources
Context as System
The accumulated knowledge, decisions, and reasoning embedded in conversations, tickets, and local adjustments. In WadTown, this context is treated as the primary data structure—more valuable than any imposed schema.
Provenance and Trust
Rather than trusting tidiness or authority, WadTown proposes that trust emerges from being able to trace decisions back to their origins and understand the reasoning that led to them. Provenance is the chain of custody for decisions.
Related Concepts
- Jane Jacobs — Urban theorist who championed emergent order and local knowledge over top-down planning
- Robert Moses — Master builder and archetype of imposed, top-down order
- Emergent behavior — Order arising from local interactions rather than central control
- Local knowledge — The situated understanding of those within a system
- Slack — The collaboration platform where context accumulates
- Queryable context — The central insight: making the messy historical record accessible through search and indexing, rather than imposing a clean schema
Appendix 1: Worker–Agenda–Domain (WAD) as Fractal Theory of Mind
(in the voice and spirit of Jane Jacobs)
1. The Need to See Others Without Controlling Them
One of the enduring justifications for imposed order—of the kind favored by Robert Moses—is the claim that individuals cannot adequately understand one another.
Therefore:
- they must be coordinated
- they must be directed
- they must be simplified
This assumption leads inevitably to:
- roles instead of people
- plans instead of intentions
- control instead of understanding
But this, too, is a false necessity.
2. The Minimum Required for Understanding
In a functioning city, people do not maintain elaborate models of one another.
They do not simulate each other’s inner lives.
Instead, they rely on something far simpler:
- who is present
- what they appear to be doing
- where they are acting
From this, they derive enough understanding to:
- avoid conflict
- cooperate when needed
- adjust their own behavior
This is a practical, lightweight theory of mind.
3. WAD as That Minimum
WadTown formalizes this minimal requirement as:
Worker – Agenda – Domain
Not as a rigid structure, but as a visible, queryable fact:
- Worker — who is acting
- Agenda — what they are trying to make true
- Domain — where that effort is applied
Nothing more is required.
Nothing more is desirable.
4. Why It Is Fractal
This structure holds at every scale.
A single worker:
- has an agenda
- acts in a domain
A team:
- is itself a worker
- with a shared agenda
- across a broader domain
A system:
- behaves as a worker
- pursuing goals
- across still larger domains
Thus:
WAD composes without distortion
It does not require translation or reinterpretation as it scales.
This is what makes it fractal.
5. The Difference from Roles
Roles, as used in imposed systems, attempt to simplify people:
- by narrowing responsibility
- by prescribing behavior
- by abstracting away intention
WAD does the opposite.
It exposes:
- intention (agenda)
- agency (worker)
- location of effect (domain)
Without prescribing action.
6. What WAD Enables
With WAD visible, any participant can ask:
- Who else is acting here?
- What are they trying to accomplish?
- Where might our efforts overlap or conflict?
This is sufficient for:
- coordination
- conflict avoidance
- cooperation
Not perfectly, but reliably.
7. Why This Matters
Without such visibility, participants must rely on:
- instructions
- hierarchy
- mediation
With WAD, they can instead rely on:
direct perception of others’ intent
This is the essence of a functioning public realm.
8. WAD and Context
WAD alone does not explain the system.
It tells us:
who is trying to do what, where
But not:
why those efforts matter or are constrained
That is the role of queryable context.
Together:
- WAD provides social orientation
- Context provides meaning
9. The Combined Effect
When both are present:
- participants can see each other
- participants can understand the constraints they share
This allows coordination to emerge without being imposed.
10. Final Note
The lesson is simple:
People do not need to be simplified to be coordinated.
They need to be visible to one another in terms that matter.
WAD provides those terms.
And because it holds at every scale, it allows a system to grow without losing its capacity for understanding.
Key Concepts
Worker–Agenda–Domain (WAD)
A minimal, fractal structure for representing intentional action: who is acting, what they are trying to accomplish, and where their effort applies. Scalable without distortion—the same structure applies to individuals, teams, and systems.
Imposed Order
Top-down control structures that replace context and intention with roles and plans. Associated with Robert Moses and criticized by Jane Jacobs as brittle and self-defeating.
Theory of Mind
The cognitive capacity to understand that other beings have mental states different from one’s own. WAD operationalizes a lightweight version sufficient for coordination.
Fractal
A structure that holds self-similar form at different scales. WAD’s fractal property means it does not require reinterpretation or translation as it scales from individual workers to teams to systems.
Queryable Context
The WadTown principle that makes historical reasoning, constraints, and decisions accessible through search and indexing, rather than imposing a simplified schema.
Public Realm
The shared space where coordination can emerge from direct perception of others’ intentions and constraints, rather than through hierarchy or mediation. The foundation of functioning cities and systems.
Appendix Note
This appendix operationalizes one core claim of the WadTown Manifesto: that when context is queryable and intentions are visible, people can act coherently without imposed order.
WAD provides the structure for visibility; WadTown provides the infrastructure for queryable context.
Together, they enable systems that scale through understanding rather than control.
Appendix II: Decomposable Content vs Ambient Context
(in the voice and spirit of Jane Jacobs, with Robert Moses still very much in view)
1. The Planner’s Habit of Confusion
One of the planner’s oldest mistakes is to confuse content with context.
Because content can be divided, copied, reassigned, routed, and rearranged, the planner assumes context can be treated the same way.
It cannot.
This is the kind of error that comes naturally to systems built in the spirit of Moses: what can be broken apart ought to be broken apart, and what is broken apart can then be centrally governed.
But this only works for one half of reality.
2. Content Decomposes
Content is what is made, changed, routed, edited, executed.
It includes:
- code
- tickets
- branches
- documents
- plans
- tasks
- outputs
This sort of thing can be decomposed almost indefinitely.
One may divide a large feature into smaller features, those into subtasks, and those into still smaller units of effort. One may do this sensibly or foolishly, but one may do it.
Content is, in this respect, obedient to decomposition.
3. Context Does Not
Context is something altogether different.
Context is not what we are doing.
It is what makes what we are doing mean something.
It lives in:
- old decisions
- customer pain
- prior failures
- standing promises
- regulatory burdens
- unwritten exceptions
- support history
- Slack arguments
- call transcripts
- the accumulated reasons things are the way they are
This cannot be neatly subdivided along the same lines as content.
When content is split, context does not politely split with it.
It remains:
ambient
shared
entangled
inconveniently whole
4. Moses’s Answer
Moses confronted a world whose relevant context was too vast, too fragmented, and too disorderly to be carried in the heads of local actors.
His response was understandable.
He replaced ambient context with:
- structure
- authority
- imposed route
- institutional memory
- formal control
This is what planners do when they do not believe context can be consulted reliably:
they attempt to make it unnecessary.
The result is not nonsense. It is often powerful. It is often effective. It is often, under the constraints, entirely rational.
But it comes at a terrible price, because what is removed is not confusion alone. What is removed is also local intelligence.
5. The Digital Difference
In an ordinary city, context is fragmented by nature.
It is trapped in:
- neighborhoods
- habits
- personal memory
- informal networks
- custom
- rumor
No one can simply ask the city, in a trustworthy way, why a thing is the way it is.
A digital city is not bound by this limitation.
Its context can remain ambient and yet become queryable.
That is the miracle.
Not that the mess disappears.
Not that the world becomes simpler.
But that the living, inherited, scattered meaning of the place can be made accessible without first being reduced to a planner’s diagram.
6. Content Must Be Decomposable
This is not a defect.
A living system must permit decomposition of content, because work must be splittable if many workers are to participate.
Without decomposable content, one gets bottlenecks, dependency on heroic individuals, and a system too brittle to grow.
So content must be divisible.
That is simply a practical necessity.
7. Context Must Remain Ambient
But context must not be treated as if it were merely another kind of content.
It is not something to be chopped up, prepackaged, and shipped along with tasks like a planner’s lunch pail.
It must remain:
- globally available
- ambient to the system
- queryable when needed
- trustworthy in provenance
This is because context is not a work packet. It is the surrounding world.
And the surrounding world does not travel in fragments simply because the task does.
8. The Proper Relation Between Them
The right relation is therefore very simple:
- content is decomposed
- context is consulted
Or, to put it more strongly:
Content may be broken apart.
Context must remain available as a whole.
This does not mean every worker sees all context at once. That would be absurd.
It means every worker, however local the task, can query into the same inherited world of meaning.
9. Why This Matters for WadTown
WadTown depends on two different miracles, not one.
The first is that action can be decomposed through Worker–Agenda–Domain without losing agency.
The second is that meaning need not fragment as action fragments.
That second miracle depends on ambient context.
If context fragments along with content, WadTown collapses back into the old familiar pattern:
- hierarchy
- mediation
- thick roles
- imposed coordination
If context remains ambient and queryable, then even very small local efforts can remain tied to the larger truth of the system.
10. The Planner’s Fear
The planner fears ambient context because it appears messy.
And it is messy.
But the mess is not the disease. It is the life.
The disease is when only the planner has access to meaning, and everyone else must execute content stripped of context.
That is the road back to Moses every time.
11. The Principle
So the principle is this:
Decompose content as aggressively as you like.
Never require context to decompose with it.
That is the distinction on which the entire possibility of WadTown rests.
12. Final Note
A good city does not ask every citizen to carry the whole city in their head.
It asks only that the city remain legible enough that they may consult it.
So too with digital systems.
Content must travel.
Context must remain.
Appendix III: Context as Linkable Snapshots
(in the voice and spirit of Jane Jacobs, with Robert Moses still waiting impatiently with a plan in his hand)
1. The Planner’s Demand for One View
The planner has always preferred a single, official view.
This is understandable.
A single view is tidy.
A single view is governable.
A single view can be placed on a map, or a dashboard, or a presentation, and pronounced coherent.
But living systems do not proceed by means of a single view.
A city does not exist as one view.
It exists as innumerable partial views:
- from the grocer
- from the tenant
- from the mother watching the stoop
- from the child cutting through the alley
- from the shopkeeper who knows which corner is safe after dark
Each sees something true.
None sees the whole.
The same is true in a digital city.
2. Why a Single Snapshot Fails
If one insists on a single snapshot of the whole, one immediately recreates the old problem of imposed order.
Someone must decide:
- what belongs in it
- what is omitted
- what is current
- what is authoritative
- when all must wait for it to be updated
This is the road back to Moses: one central picture, one central interpretation, one central permission to proceed.
It is not that such a picture is useless. It is that it is never sufficient, and very often tyrannical.
For a living system, one snapshot is too little and too much at once:
- too little, because it cannot hold all the needed truth
- too much, because it demands one view fit every participant
3. The Necessity of Snapshots, in the Plural
What a city actually requires is not one view, but many.
Each worker must be able to act from a view that is:
- partial
- coherent
- trustworthy
- sufficient for action
That is a snapshot.
A snapshot is not the world.
It is a grasp of the world, enough for a participant to move without blindness.
And because the city has many participants, there must be many such snapshots.
This plural is not an inconvenience.
It is the very condition of life.
4. Why They Must Be Linkable
Yet if these snapshots remain isolated, they become merely fragments.
A fragment is not a city.
A fragment is a neighborhood severed from its street network, a district cut off by an expressway, a place condemned to local ignorance.
Snapshots must therefore be linkable.
This means that no snapshot stands alone as a sealed box of truth.
It must be possible to move from one to another by means of the claims they share, inherit, dispute, or explain.
A participant must be able to ask:
- What larger truth does this depend on?
- What other view bears on this one?
- What neighboring reality makes this intelligible?
- What prior claim gives this its force?
- What related snapshot would change how I should act?
Without such links, snapshots become another planner’s filing cabinet.
With such links, they become a city.
5. What Makes a Snapshot Trustworthy
A trustworthy snapshot does not depend on elegance.
It depends on provenance.
Every claim within it must be answerable in the old and necessary form:
Who said what, when, where, and how?
This is what allows a participant not merely to receive context, but to trust it enough to act.
A snapshot need not be complete.
It need only be defensible.
That is a great and practical difference.
6. The Difference Between Content and Context
Content may be decomposed freely.
One may split:
- a feature
- a task
- a document
- a branch
- a line of implementation
Context is different.
Context cannot be decomposed in the same obedient fashion, because its significance is cross-cutting, inherited, and shared.
Therefore context must appear not as packets, but as snapshots:
- bounded enough to use
- linked enough to remain meaningful
This is the proper digital answer to a problem that physical cities could only suffer through.
7. The City Moses Could Not Build
Moses governed by replacing local, partial, conflicting views with a singular imposed one.
Again, this was not stupidity. It was the logical answer available to a planner who could not make the city’s dispersed context linkable.
If local views cannot be connected, then they must be overridden.
That was his world.
But a digital city is not so constrained.
- It can preserve multiplicity without surrendering coherence.
- It can allow many snapshots without requiring one sovereign picture.
- It can let participants move among linked truths rather than obey a single authorized simplification.
That is why WadTown is not merely a better engineering pattern.
It is a different political order.
8. Snapshots and Workers
A worker in WadTown does not need the whole city at once.
That would be absurd.
A worker needs:
- a local Worker–Agenda–Domain position
- a coherent snapshot adequate to that position
- links outward into the wider inherited world
In this way, the worker is neither blind nor overburdened.
The worker can proceed locally while remaining answerable to the whole.
That is a rare achievement.
9. The Public Character of Linkable Context
The most important streets in a city are public, not because everyone uses them identically, but because everyone can orient themselves by them.
So too with context.
If the reasons things are the way they are remain trapped in:
- private recollection
- proprietary dashboards
- scattered messages
- inaccessible expertise
then the city is not public in any meaningful sense.
Linkable snapshots make context public.
Not simplified.
Not unanimous.
Public.
That is the condition under which many actors may coordinate without first being domesticated into obedience.
10. The Principle
So the principle is this:
Context must appear as many trustworthy snapshots, each partial, each sufficient, and each linkable to the larger world of meaning.
Not one official picture.
Not a thousand isolated fragments.
A city lives between these errors.
11. Final Note
A good city does not require every citizen to see everything.
It requires only that what they can see be:
- real
- connected
- and enough to let them move intelligently among others
So too with WadTown.
The miracle is not one perfect view.
The miracle is many truthful views that can still find one another.
Appendix IV: Reconcilability Precedes Coherence
(in the voice and spirit of Jane Jacobs, with Robert Moses still very much in the background)
1. The Planner’s Demand for Coherence
The planner is always tempted by coherence.
This is understandable. Coherence is reassuring. It appears to promise that everyone is working from the same picture, toward the same end, under the same explanation. It offers the comfort of a city already resolved.
But living systems do not begin in coherence.
They begin in:
- difference
- partial vision
- competing reasons
- local urgencies
- incomplete understanding
To demand coherence first is to demand that life stop until the planner is satisfied.
That is the old error.
2. Moses and the Burden of Prior Agreement
Moses required coherence in advance because he could imagine no other reliable source of order.
If many actors are to move, then:
- they must first be aligned
- their purposes must first be reconciled
- their understanding must first be simplified
- their differences must first be subordinated
Only then, under such a view, may action proceed safely.
This is not madness. It is the natural conclusion of a system that distrusts plurality.
But it is also the death of a living city.
3. The Reality of Cities
A city does not wait to become coherent before it begins.
Its people do not first gather to produce one unified explanation for:
- why the street matters
- why the corner is changing
- why the shop must stay
- why the block feels safe
- why one use belongs and another does not
They act from partial reasons.
They move from:
- local knowledge
- inherited memory
- immediate need
- practical judgment
And yet the city does not therefore dissolve.
Why?
Not because it is coherent in advance.
But because it remains reconcilable.
4. Reconcilability
This is the more primitive condition.
A city need not require that all its actors agree before acting.
It requires only that their differences do not make a shared world impossible afterward.
That is reconcilability.
It is not:
- agreement
- unity
- consistency
- one official “why”
It is:
the continued possibility that divergent action may still be brought back into relation.
This is a humbler virtue than coherence, but a more necessary one.
5. Why Coherence Comes Later
Coherence is not the natural starting point of a living order.
It is the occasional achievement of many local acts that have not foreclosed one another.
This means that coherence, when it appears, must be understood properly.
It is not:
- a prerequisite
- a command
- a condition for entry
It is:
an emergent outcome of actors whose moves remain reconcilable.
This distinction is of the greatest importance.
6. The Meaning for WadTown
WadTown cannot be built on the fantasy of prior coherence.
Its workers will not begin with:
- one authoritative explanation
- one agreed interpretation
- one complete view of the whole
They will begin with:
- decomposed Worker–Agenda–Domain positions
- local purposes
- partial, linkable snapshots
- plural and sometimes competing reasons
This is not a defect.
It is the very condition under which a living digital city becomes possible.
7. What Must Be Preserved
If coherence is not required in advance, something else must be preserved in its place.
That thing is reconcilability.
For a system to remain reconcilable, its actions must not:
- erase the possibility of understanding what happened
- destroy the shared ground on which others still stand
- sever all links back to the larger inherited context
- make return impossible
A good city permits divergence, but not estrangement without end.
So too with WadTown.
8. The Role of Queryable Context
This is where queryable global context becomes indispensable.
If context remains available, then even after local divergence, workers may still ask:
- what larger truth bears on this?
- what prior reason have I missed?
- whose view must now be brought into relation with mine?
- what have I endangered?
- what remains possible?
Without such context, difference hardens into fragmentation.
With it, difference may yet be reconciled.
The context does not guarantee harmony.
It preserves the path back.
9. The Political Difference
This is the deepest difference between imposed and living order.
The imposed system says:
coherence first, then motion.
The living system says:
motion first, provided reconcilability remains.
The first fears plurality and tries to master it before it appears.
The second permits plurality because it trusts that a shared world can still be recovered.
That is a very different politics.
10. Why This Matters More Than Coherence
Too much admiration for coherence produces brittle systems.
They may look orderly, but only because they have forbidden so much of life.
Reconcilability is less tidy and more merciful.
It allows:
- experiment
- disagreement
- local judgment
- incomplete understanding
- revision after the fact
It does not ask workers to be perfect.
It asks only that they not make return impossible.
This is a much more humane standard, and a much more realistic one.
11. The Principle
So the principle is this:
Do not require coherence before action.
Require only that action remain reconcilable.
From such a condition, coherence may sometimes emerge.
From the opposite condition, life rarely does.
12. Final Note
A good city is not one in which everyone agrees before anything happens.
It is one in which many different actors may proceed from different reasons and still remain capable of living together afterward.
So too with WadTown.
Coherence is a blessing when it appears.
Reconcilability is the condition that makes it possible.
Appendix V: When GasTown Is (and Isn’t) Necessary
(in the voice and spirit of Steve Yegge, channeling hard-won pragmatism from Amazon and Google)
1. The Honest Start
I’ve watched enough systems fail that I’m obligated to say this: WadTown is correct, and GasTown is sometimes necessary.
They’re not opponents. GasTown is what you do when WadTown’s assumptions stop holding.
The question isn’t which is right. The question is: when do the assumptions break?
Most people get this backwards. They implement GasTown first out of fear, then spend a decade trying to escape it.
What you should do is implement WadTown first, then recognize the exact moment you need GasTown—and only then.
The hard part is knowing the difference.
2. WadTown Works When Order Is Already Local
WadTown assumes something crucial: that local order exists.
That teams know how to work. That conventions have evolved. That people understand their domain well enough to explain it.
This is true more often than planners admit.
A pharma lab has learned, over years, how to run experiments. They have conventions. They know what metadata matters. They argue about methodology in ways that are grounded in accumulated practice.
Quilt’s bet is that you can capture these local orders, make them explicit, link them together, and scale from there.
This works when the local order is real.
If you have 50 teams, each with genuine competence in their domain, WadTown lets you scale without flattening them into a single imposed schema.
Each team keeps their local convention. The system makes those conventions queryable and composable. You get coherence without central control.
That’s the miracle.
3. GasTown Becomes Necessary When Local Order Breaks Down
But there’s a moment when this stops working. I’ve seen it happen at Google, Amazon, Microsoft.
The moment is when local order becomes unreliable as a foundation.
This happens in several ways:
3.1 The Chaos Inflection
You scale too fast. You hire too many people. You inherit incompetent teams or teams with conflicting local orders.
Now you have 500 teams, and 50 of them are doing things that actively break the other 450.
At this point, querying context doesn’t help. You can’t ask “why did this team decide to do it that way?” when the answer is “we didn’t know there was a better way” or worse, “we did it for reasons that no longer apply.”
Local order has become historical accident.
In this state, WadTown collapses. Reconcilability becomes impossible when there are too many incompatible local orders.
This is when you need GasTown.
You need to say: “These three things are non-negotiable. Everyone does it this way. No exceptions.”
3.2 The Critical Path Problem
At smaller scale, context can remain ambient. Every engineer knows enough to navigate by it.
But once you have 5,000 engineers, “ambient context” is actually “scattered across thousands of Slack channels, dozens of wikis, and the heads of people who left six months ago.”
In this state, you can’t rely on queryable context. You can’t ask every decision-maker “is this compatible with what you’re doing?” because you have 5,000 of them and decisions happen every second.
You need GasTown: explicit constraints that don’t require consensus. Rules that can be verified mechanically instead of contextually.
3.3 The Inherited Debt Problem
Sometimes the breakdown isn’t your fault. You inherited a system where someone made terrible decisions that everyone else now depends on.
The local order is: “We can’t change this because it would break everything.”
Now you’re stuck with a convention that everyone agrees is bad but nobody can change.
WadTown says: query the context, understand why, reconcile. But if the answer is “we’re trapped,” reconcilability doesn’t help.
You need GasTown: the authority to say “we’re breaking this, here’s how we’re fixing it, everyone moves or falls off.”
Sometimes you need to overwrite local order, not recognize it.
4. The Three Signs You Need GasTown
If you notice any of these, you’re past the point where WadTown alone will work:
Sign 1: Context is no longer queryable; it’s just massive.
You can technically ask “why did we design it this way?” but the answer requires interviewing five people in three timezones who might not remember. The cost of querying has exceeded the value of the answer.
Sign 2: Local orders actively conflict.
You have two teams whose “best practices” are incompatible. They can’t both be right. Reconcilability fails because there’s no common ground to reconcile to.
Sign 3: You’re shipping slower because of coordination overhead.
You spend so much time consulting context, aligning with other teams, and resolving conflicts that nothing gets built. The system has become coordinative rather than productive.
When any of these is true, you need imposed structure.
5. What GasTown Actually Looks Like at Scale
Here’s the thing nobody says: GasTown doesn’t mean tyranny.
It means:
- Clear, written rules
- Mechanical enforcement where possible
- A decision-making process that’s visible, if not democratic
- Permission to opt out (at cost)
Google’s approach: “Here are three ways we do this. Pick one. If you want a fourth way, get Architecture Review Board approval.” That’s not Moses. That’s constraint with escape hatches.
Amazon’s approach: “All teams must expose their data via an API. Here’s the standard. Deviations require VP approval.” That’s not tyranny. That’s a decision made once, at the right level, so 10,000 teams don’t each make it 10,000 times.
The key: You’re not trying to prevent local order. You’re trying to prevent incompatible local orders.
6. The Decision Point
Here’s how to know if you actually need GasTown:
Ask this question: “If I let this decision be made locally, will it break someone else’s work?”
If the answer is “maybe, but probably not,” keep it local. WadTown.
If the answer is “yes, guaranteed,” move it up. GasTown.
If the answer is “I don’t know, but I’m scared it might,” that’s when you’re guessing. Don’t implement GasTown yet. Build better context first. Make the impact visible. Then decide.
Most organizations implement GasTown here—out of fear, not necessity. They lose a decade of velocity trying to escape it.
7. The Reconcilability Bridge
Here’s Appendix IV applied to this problem: GasTown doesn’t mean making reconcilability impossible.
When you impose a standard (GasTown), you should make it queryable (WadTown):
- Why does this standard exist?
- What problem was it solving?
- When can it be overridden?
- How do we change it if it stops working?
Google did this badly. They had five-layer approval processes but nobody knew why they existed. Teams just assumed they were capricious.
Amazon did it better. “Here’s the standard. Here’s why. Here’s how you challenge it.”
The right hybrid: Impose the minimum GasTown structure necessary to prevent coordination chaos. Then make that structure itself queryable and revisable. That’s not tyranny. That’s governance with guardrails instead of walls.
8. Quilt’s Position
This is why Quilt is interesting: they’re building WadTown (local conventions, queryable context, linked snapshots) with enough GasTown (required metadata facets, workflow gates) to stay coherent without becoming brittle.
They’re not trying to run life sciences as a pure WadTown system. That would break.
They’re not trying to impose a unified schema either. That would kill innovation.
They’re trying to capture local order (what labs actually do), make it explicit (required metadata), and link it (workflows, lineage). Then make it queryable so other labs can discover and adopt conventions without reinventing them.
That’s the balance.
They’ll know it’s working when:
- Scientists accept the metadata capture naturally (it reflects what they’re already doing)
- The workflow gates feel like support, not friction
- Teams discover and adopt each other’s conventions without central mandate
They’ll know it’s failing when:
- Scientists are gaming the metadata system to get through gates
- The workflow becomes the bottleneck, not the enabler
- Teams retreat into shadow systems
9. The Inflection Point You Actually Watch For
Most scaling problems aren’t about hitting some magic number (1,000 engineers, 10 services, etc.).
They’re about hitting an inflection in communication topology.
- When you could previously rely on “everyone knows this because we talked about it,” but now you can’t.
- When the cost of aligning decisions exceeds the cost of standardizing them.
- When local knowledge no longer propagates naturally through the organization.
That’s your sign.
Up to that point, invest in WadTown: better context, better visibility, better linkage. Make the order visible rather than replacing it.
Past that point, you need GasTown: explicit standards, clear constraints, mechanical enforcement where it matters.
The mistake: implementing GasTown before you hit that point.
The other mistake: refusing to implement it even after you pass it.
10. The Real Leadership Question
Here’s what separates good engineering leaders from mediocre ones:
Good ones ask: “What is the minimum GasTown we need right now?”
Mediocre ones ask: “How can we make sure everything is safe?”
The first question leads to: constraints at the right level, freedom everywhere else.
The second question leads to: constraints everywhere, in case something breaks.
The first keeps velocity. The second kills it.
If you find yourself implementing GasTown, ask yourself: “Is this preventing real chaos, or am I just reducing uncertainty?”
Because reducing uncertainty is different from preventing chaos. One is necessary. The other is just fear.
11. The Principle
So the principle is this:
- Start with WadTown. Make order visible. Make it queryable. Make it reconcilable.
- Move to GasTown only when WadTown’s assumptions actually break—when local order becomes unreliable, when communication topology breaks down, when coordination overhead exceeds benefit.
- When you do move to GasTown, keep it minimal, keep it visible, and keep it revisable.
- And measure constantly whether you’re actually in GasTown territory or just scared.
Most organizations are scared, not in trouble.
12. Final Note
Robert Moses built GasTown systems before there was any chaos to prevent. He was prophylactic. Cautious to the point of paralysis.
The opposite error is implementing WadTown and pretending it scales forever. It doesn’t.
The skill is knowing which moment you’re in, and being honest about it.
Quilt seems to understand this. They’re recognizing local order (WadTown), making it explicit and queryable (WadTown), but with just enough governance structure to keep it coherent (GasTown, minimally applied).
That’s not dogmatism. That’s craft.
Do that, and you get systems that actually work at scale without sacrificing the intelligence of the people inside them.
That’s rare. Worth paying attention to.

Leave a comment