How I work with AI
Learn orkai in the 7-day trial by building a local shop: catalog and orders on your machine. Concept first, then exact steps and paste-ready prompts.
You will learn
- How structured context (requirement, architecture, standards) replaces re-explaining the project
- How to onboard an AI the way you onboard a teammate
- How roles and workflows tell the AI how to work, not only what to build
- Why knowledge has to live outside the chat
- How we implement Stall v1 in a fresh session
- How review and git hooks protect the house rules
- How unrelated work runs in parallel
- How to ask the project where things stand
- Why this is still being an engineer, with a teammate that remembers
The idea
I am not going to invent a new way to build software because AI arrived. We already have requirements, architecture, standards, plans, tasks, reviews, and session notes. Those are how engineers work. The AI was trained on all of it.
What changed is who executes. I still decide what we are building and what done looks like. The AI writes and edits inside that process. When I work this way I feel more like an engineer, not less: I spend my time on intent and review, not on re-pasting context into a blank chat.
Stall: catalog and orders
This tutorial has one teaching project: Stall. A stall publishes what it sells and takes orders. Think a café counter, a market stand, a bakery window. Same type of shop.
We will ship a catalog and a way to place an order. Sold-out items cannot be ordered. There are no payments, no accounts, and no delivery. That fence is the point. You finish a real shop in the 7-day trial instead of disappearing into checkout flows.
The 7-day trial is the window: one device, full product. You start it in Create the project.
The path
This page is the first. The rest are hands-on in OpenCode, building Stall.
- Create the project
- Requirement, architecture, standards
- Plan the work
- End the planning session
- Implement in a new session
- Extend the requirement
- Review and git hooks
- Workflows and review
- Workers, monitor, merge
- Ask the project
- Conclusion
Where orkai came from
I built orkai so this process lives on my machine and any coding tool can follow it: requirements, plans, sessions, and standards that outlive a chat, not a vendor's context window. If you want the persuasion, read Why orkai. This path is the doing.
Recap
You now know the method, the shop we will build, and the path. There are no orkai prompts on this page: you need the tool running first. The first hands-on lesson is Create the project: trial, install, Stall folder, MCP.