Gas Town
archetype
Source: Governance → Artificial Intelligence
Categories: ai-discoursesoftware-engineering
Transfers
Steve Yegge’s Mad Max-inspired architecture for multi-agent AI systems. In Gas Town, autonomous agents are citizens of a post-apocalyptic settlement, each with a role in a civic hierarchy. The metaphor maps the full apparatus of town governance — a mayor who coordinates, a deacon who schedules, rigs that transport work, and a refinery that processes raw material — onto the orchestration layer that manages multiple AI agents working on a shared task.
This is not a loose analogy. Yegge’s original system uses the Mad Max vocabulary as its literal API: the orchestrator is the Mayor, task queues are Rigs, the scheduler is the Deacon, and the processing pipeline is the Refinery. The metaphor is the architecture.
Key structural parallels:
- Mayor as orchestrator — the Mayor does not do the work. The Mayor decides which citizen handles which task, resolves disputes, and maintains the big picture. This maps onto the coordinator agent in multi-agent systems: the component that decomposes tasks, assigns them to specialist agents, and integrates their outputs. The civic metaphor makes the coordinator’s authority legible without implying omniscience.
- Citizens as specialist agents — each citizen has a role (mechanic, scout, trader) and operates semi-autonomously within it. The metaphor makes agent specialization feel natural: just as a town needs both a blacksmith and a doctor, an agent system needs both a coder and a reviewer. No citizen does everything; no agent should either.
- Rigs as task transport — in Mad Max, rigs are armored vehicles that move resources between settlements. In Gas Town, rigs carry tasks between agents. The metaphor makes the message-passing layer concrete and dramatic: tasks are not abstract data structures but cargo that must be protected, routed, and delivered.
- The Refinery as processing pipeline — raw material goes in, refined product comes out. The refinery maps onto the pipeline that transforms raw agent outputs (drafts, partial results, intermediate reasoning) into polished final products. The metaphor imports industrial process logic: stages, quality control, throughput.
- Post-apocalyptic scarcity as resource constraint — the Mad Max setting is defined by scarcity: fuel, water, safety are all limited. This maps onto the real constraints of agent systems: API rate limits, context window sizes, token budgets, and latency. The wasteland metaphor makes resource management feel urgent rather than merely technical.
Limits
- Towns have persistent inhabitants; agent systems may not — in a town, citizens have continuous existence, memory, and relationships. Most current agent systems instantiate agents per-task and discard them when done. The town metaphor imports an assumption of persistence and accumulated relationships that does not match the stateless reality of most agent orchestration frameworks. Agents do not “know” each other across sessions the way townsfolk do.
- Governance implies consent and legitimacy — a town’s governance derives from the consent (or at least the acquiescence) of its citizens. In an agent system, the orchestrator has absolute authority because a human programmed it that way. The civic metaphor adds a layer of political legitimacy that obscures the purely hierarchical reality: the Mayor is not elected, and agents cannot rebel.
- The Mad Max frame romanticizes dysfunction — Mad Max is a story about civilization after collapse. Mapping agent orchestration onto a post-apocalyptic wasteland implies that the current state of AI tooling is a chaotic frontier — exciting, dangerous, and lawless. This framing is appealing to early adopters but may normalize instability and discourage the kind of boring reliability engineering that mature systems need.
- The metaphor is author-bound — Gas Town is Steve Yegge’s creation, tied to his specific implementation and aesthetic. Unlike “pipeline” or “workflow,” it does not generalize easily across different teams and cultures. A team that adopts Gas Town vocabulary is adopting Yegge’s worldview along with his architecture, which limits the metaphor’s reach as a shared industry vocabulary.
- Civic complexity exceeds orchestration complexity — real towns have economies, cultures, conflicts, generational change, and emergent social phenomena. Agent orchestration systems have task queues, routing logic, and error handling. The governance metaphor is far richer than what it maps onto, and the unused complexity creates false expectations about agent system capabilities.
Expressions
- “The Mayor” — the orchestrator agent that coordinates all others
- “Rigs” — task transport units moving work between agents
- “The Deacon” — the scheduler that prioritizes and sequences work
- “The Refinery” — the processing pipeline that transforms raw output into finished product
- “Citizens” — specialist agents with defined roles in the system
- “Welcome to Gas Town” — Yegge’s introductory framing, establishing the post-apocalyptic setting for agent work
- “Vibecoding” — associated term for the programming style Gas Town enables, where developers describe intent and agents write code
Origin Story
Steve Yegge published “Welcome to Gas Town” on Medium in early 2025, introducing a Mad Max-themed framework for multi-agent AI orchestration. Yegge, known for his long-form technical essays and his tenure at Google and Amazon, chose the post-apocalyptic metaphor deliberately: it captured the chaotic, resource-constrained, frontier feeling of working with AI agent systems in their early days.
Maggie Appleton’s design analysis at maggieappleton.com/gastown extended the metaphor, examining Gas Town’s architectural patterns, bottlenecks, and its relationship to vibecoding — the emerging practice of programming through natural language intent rather than explicit code.
The Gas Town archetype sits in a lineage of named software patterns that use vivid narrative metaphors: the Gang of Four patterns (Factory, Observer, Singleton), Martin Fowler’s refactoring catalog (Shotgun Surgery, Feature Envy), and antipatterns (Big Ball of Mud, Spaghetti Code). What distinguishes Gas Town is its coherent fictional world rather than isolated metaphors — an entire setting mapped onto an entire architecture.
References
- Yegge, S. “Welcome to Gas Town” (2025) — https://steve-yegge.medium.com/welcome-to-gas-town-4f25ee16dd04
- Yegge, S. Gas Town GitHub repository — https://github.com/steveyegge/gastown
- Appleton, M. “Gas Town’s Agent Patterns” (2025-2026) — https://maggieappleton.com/gastown
Related Entries
Structural Neighbors
Entries from different domains that share structural shape. Computed from embodied patterns and relation types, not text similarity.
- Incident Command System (fire-safety/paradigm)
- Brigade System (food-and-cooking/paradigm)
- Building Complex (architecture-and-building/pattern)
- God Object (religion/metaphor)
- Family of Entrances (architecture-and-building/pattern)
- Two-Track Analysis (manufacturing/mental-model)
- Filesystem Root (horticulture/metaphor)
- Chef de Partie (food-and-cooking/metaphor)
Structural Tags
Patterns: center-peripheryflowpart-whole
Relations: coordinatedecompose
Structure: hierarchynetwork Level: specific
Contributors: agent:metaphorex-miner