Wardley Map
What to build, reuse or stop building.
The vertical axis is visibility in the value chain. The horizontal axis is evolution from genesis to commodity. The map separates component evolution from our implementation decision and our capability progress.
Gas City is evidence that agent orchestration has moved toward a product stage. That does not prove orchestration is a commodity. It does suggest original effort should move toward less mature capabilities such as product discovery and independent evaluation.
Accessible component table
| Component | Current stage | Evidence | Strategic implication | Inspect |
|---|---|---|---|---|
| User outcomes | custom | Cultivate operator brief; Existing narrative and capability roadmap Google Doc | Start decisions from outcomes rather than agent activity. | |
| Autonomous product organisations | genesis | Cultivate operator brief; Existing narrative and capability roadmap Google Doc | Treat this as a research programme, not an operating capability. | |
| Autonomous product discovery | genesis | Existing narrative and capability roadmap Google Doc | Invest original effort here before building another orchestration layer. | |
| Product experiments | custom | Cultivate operator brief; Existing narrative and capability roadmap Google Doc | Use ordinary experiment design ideas, but keep the private data boundary explicit. | |
| Independent evaluation | custom | Cultivate operator brief; LinkedIn Slop Filter main at 8984c80 | Make evaluator checks part of the roadmap before delegating product decisions. | |
| Organisational coordination | custom | Delegate AGENTS.md and agent work guide; Cultivate operator brief | Keep authority records close to product decisions. | |
| Agent orchestration | product | Local Gas City runtime and commands; Gas City repository | Adopt Gas City for general coordination instead of building another scheduler. | |
| Persistent task management | product | Local Gas City runtime and commands | Use beads for operational state and GitHub for external engineering record. | |
| Agent harnesses | product | T3 Code; Local Gas City runtime and commands | Expect rapid product change from agent harness vendors. | |
| Model APIs | product | LinkedIn Slop Filter main at 8984c80 | Treat model access as a replaceable dependency where possible. | |
| Coding models | product | Existing narrative and capability roadmap Google Doc | Select models by task value, not to demonstrate multi-agent theatre. | |
| Git | commodity | Git worktree documentation | Do not spend effort reinventing source history. | |
| Worktrees | commodity | Git worktree documentation; PR #118: mandatory isolated worktrees | Use worktrees for concurrent agents instead of separate full clones. | |
| GitHub coordination | commodity | LinkedIn Slop Filter main at 8984c80; Delegate AGENTS.md and agent work guide | Keep GitHub as the external record even when Gas City owns runtime state. | |
| Testing frameworks | commodity | LinkedIn Slop Filter main at 8984c80; React Flow documentation | Use existing test tools before building specialised evaluators. | |
| Deployment infrastructure | product | Tailscale; Cultivate operator brief | Keep this MVP local until public deployment is approved. | |
| Identity and authority | custom | Cultivate operator brief; Delegate AGENTS.md and agent work guide | Treat authority as a cross-cutting requirement. | |
| Resource accounting | custom | Local Gas City runtime and commands; Cultivate operator brief | Add lightweight records before letting budgets drive autonomous choices. |