AI-Native Way of Working: Using Codex as a Partner at Every Step
This starter helps a team go from a rough idea to a credible demo within one hackathon day. It is designed for technical teams who want to use Codex across the full delivery lifecycle, not just for code generation.
This repository gives teams the "how":
- a simple workflow
- clear planning artifacts
- role guidance
- prompt starters
- quality checkpoints
- demo preparation structure
Teams bring the "what": the product, tool, workflow, or prototype they want to build.
- It works before you choose a stack.
- It keeps the team aligned while scope is still moving.
- It supports fast collaboration between humans and Codex.
- It avoids wasting time on boilerplate too early.
- It works for web apps, agents, dashboards, developer tools, data products, and experiments.
- Start with 00-product-brief.md.
- Assign roles in 01-team-roles.md.
- Break the work down in 02-task-breakdown.md.
- Decide the stack and architecture in 03-architecture.md.
- Plan implementation in 04-implementation-plan.md.
- Define validation in 05-test-plan.md.
- Use Codex to help generate and review implementation work.
- Capture the outcome in 06-pr-summary.md, 07-demo-script.md, and 08-codex-workflow-log.md.
- Pick an idea with a visible, demoable outcome.
- Narrow scope aggressively.
- Choose a stack only after the problem and constraints are clear.
- Split work into parallel streams for product, implementation, quality, and demo.
- Use Codex for planning, architecture, implementation, review, and documentation.
- Keep a running workflow log so you can explain how AI helped.
See guides/ai-native-workflow.md and guides/codex-usage-guide.md.
By the end of the day, teams should ideally have:
- a clear product brief and scope
- a role split and task plan
- an architecture decision
- a working prototype or a credible partial implementation
- tests or validation evidence
- setup notes and limitations
- a PR-style summary
- a short demo script
- a Codex workflow log
- Scope the idea and challenge over-ambition.
- Turn goals into parallel tasks with acceptance criteria.
- Compare stack options and suggest a right-sized architecture.
- Generate implementation plans before generating code.
- Produce code, tests, docs, and review notes when the plan is clear.
- Help summarize progress and prepare the demo narrative.
Use the simplest stack that makes the demo believable.
- If the value is visual and interactive, prefer a small web stack.
- If the value is automation, prefer a CLI or agent workflow.
- If the value is analysis, prefer a notebook, script, or dashboard.
- If setup risk is high, choose fewer services and fewer moving parts.
Start with guides/stack-presets.md.
- Use examples/example-ideas.md if your team needs inspiration.
- Use examples/submitted-ideas-scoping.md to pressure-test ideas against one-day reality.
- Borrow scope cuts, stack choices, and prompt starters directly.
The team is done when the scope is clear, the implementation is credible, the demo is rehearsable, and the role of Codex is visible and explainable.
Use guides/definition-of-done.md as the final checklist.
The final implementation can use any stack. This starter only gives structure so teams can move quickly, stay realistic, and show an AI-native way of working.