AI-Native Product Development

Resources

• Mermaid flowcharts — Official documentation for text-defined graphs, subgraphs, edges, and interactions. Gives agents an editable artifact for mapping user journeys, architectural dependencies, risk areas, and proposed implementation paths that humans can inspect. https://mermaid.js.org/syntax/flowchart.html

• XState graph and path generation — Stately documents traversing state machines and generating event paths for analysis and model-based tests. Useful for synthetic user flows, including alternate and failure paths, provided the teacher makes clear that coverage of a model does not prove the model matches real users. https://stately.ai/docs/graph

• Git log — Git’s reference covers history traversal, filtering, patch output, and graph views. Grounds AI-assisted historical exploration in actual commits so implementation strategies can account for previous changes rather than infer intent solely from today’s code. https://git-scm.com/docs/git-log

• Map the scopes — Shape Up organizes work around integrated pieces of the product rather than separate lists for each role. Helps identify dependency boundaries and meaningful merge points before assigning direct work or parallel agent tasks. https://basecamp.com/shapeup/3.3-chapter-12

• Decide when to stop — Shape Up discusses scope growth, baseline comparisons, and scope hammering without lowering quality. Provides a practical counterweight to agents expanding a plan: simplify or cut optional work while preserving the useful outcome and making remaining risks explicit. https://basecamp.com/shapeup/3.5-chapter-14

• How Anthropic built its multi-agent research system — A first-party account of delegation, parallel exploration, coordination, and evaluation in a research system. Use its task-boundary and handoff lessons for product exploration, while distinguishing research parallelism from coding work that shares files and integration constraints. https://www.anthropic.com/engineering/multi-agent-research-system

• Git worktrees — Official documentation for multiple working trees attached to one repository. Supports isolated parallel implementation, but also makes a useful teaching distinction: filesystem isolation does not remove semantic conflicts or the need for an owned integration checkpoint. https://git-scm.com/docs/git-worktree

• Playwright Trace Viewer — Playwright’s viewer connects browser actions with DOM snapshots, network activity, console output, and timing. Creates a concrete feedback artifact for an agent-human development loop instead of relying on the agent’s account of what happened. https://playwright.dev/docs/trace-viewer

• Storybook stories — Storybook documents reproducible component states through stories and arguments. Useful for rapid visual iteration, synthetic loading and error states, and human feedback tied to a stable artifact rather than an ephemeral screenshot alone. https://storybook.js.org/docs/writing-stories

• AHA Programming — Kent C. Dodds argues for avoiding hasty abstractions and preferring duplication over the wrong abstraction. Anchors decisions about reusable primitives, simplification, and technical debt when AI can generate abstractions faster than a team can judge their long-term fit. https://kentcdodds.com/blog/aha-programming

• Continuous Integration — Martin Fowler explains frequent integration, automated builds, and rapid feedback on shared changes. Establishes why independently successful agent branches still need integration checks before anyone claims the product change is complete. https://martinfowler.com/articles/continuousIntegration.html

• OpenTelemetry signals — Official concepts for traces, metrics, and logs explain complementary views of runtime behavior. Helps make system feedback accessible during development and release monitoring, with enough context to connect a symptom to the change being built. https://opentelemetry.io/docs/concepts/signals/

• Feature toggles — Pete Hodgson’s article distinguishes toggle categories and their operational and maintenance costs. Useful for separating deployment from exposure, designing staged delivery, and assigning ownership for unresolved risks and eventual toggle removal. https://martinfowler.com/articles/feature-toggles.html

• Effective harnesses for long-running agents — Anthropic describes incremental work, progress artifacts, environment setup, and verification across agent sessions. Provides a concrete implementation-loop reference for preserving ownership and conventions through handoffs instead of letting each session restart from incomplete context. https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents