Improve how you plan, delegate, code, and ship with AI.
Build with AI is about developing a deliberate way to plan, delegate, supervise, and verify AI-assisted product work. The tools change quickly; the judgment about what to hand over, what to keep, and how to check the result is what carries from one tool to the next.
The four categories move from a portable way of working, to reusable capabilities, to the two coding agents I use most in real project work.
Repeatable process → reusable capability → tool choice → real-project practice
Start any category from the problem in front of you; only Claude Code: Builder Mode is meant to be followed in order. The results described below are things you build in your own projects with what you learn, not files included with membership. A resource offers downloads only when it says so.
AI Workflow Design
AI Workflow Design is the layer that makes any AI tool useful inside a real project: instructions, retained context, clear handoffs, standards, review, and work you can repeat. It explains why an agent can produce working code and still lose the intent of the work.
Questions it explores
- What does an agent need to know before it starts, and where should that context live?
- How do I hand work to an agent and get back something I can actually review?
- Which part of my current workflow loses the most time or the most intent?
Something you could work toward: A project instruction file and a handoff checklist that let you start a new agent session without re-explaining your app.
Start here if you already use an AI tool on a real project and keep repeating yourself, or keep correcting the same mistakes. Bring the part of your workflow that frustrates you most.
Good to know: It teaches a way of working that carries across tools, rather than one product’s settings.
AI Skills
Skills are reusable capabilities an agent can apply, such as a well-shaped technique for a recurring Apple development task. A good skill carries craft and repeatable judgment into the agent’s work instead of leaving it to a one-off prompt.
Questions it explores
- What makes a skill reusable rather than a prompt that happened to work once?
- How do I install a skill and adapt it to my own project?
- How do I check that the agent applied it well?
Something you could work toward: One skill installed in your setup and adapted to your project, with a quick way to check its output.
Start here if your agent handles a recurring task inconsistently and you want it done the same careful way each time.
Good to know: When a resource offers a skill package, the resource says so. A skill from a resource is separate from the Swift Brain teamspace, which comes with Crafter.
Claude Code: Builder Mode
Claude Code: Builder Mode is an ordered exploration of how to make Claude Code a dependable coding agent in your own studio. Context, permissions, tools, skills, hooks, specialist work, supervision, and verification are treated as one practice rather than a list of features.
Questions it explores
- What context does Claude Code need to work well in my project?
- Which actions should it be allowed to take, and how do I supervise them?
- How do individual capabilities fit together into a way of working I can trust?
Something you could work toward: A Claude Code setup for one of your projects with permissions, context, and checks you understand and chose deliberately.
Start here if you want to rely on Claude Code in real work. Begin with Session 01 and follow the sessions in order; to practise alongside them you need your own access to Claude Code.
Read the Claude Code: Builder Mode guideBrowse Claude Code: Builder Mode in the library
Codex: Workbench Mode
Codex: Workbench Mode looks at Codex as an evolving workbench: agent work, reviews, parallel or longer-running tasks, integrations, and useful studio responsibilities. Real product tasks are followed from scope and delegation through review, tests, and outcome.
Questions it explores
- Which kinds of product work suit a longer-running or parallel agent?
- How do I review agent work so that it holds up in the product?
- Where are the limits and tradeoffs, and when is another tool the better choice?
Something you could work toward: A real task delegated to Codex, reviewed against tests, and merged with confidence, or set aside for a clear reason.
Start here if you are weighing Codex for real product work. Each piece stands on its own, so begin with the task closest to yours; there is no fixed order.
Other paths
Move between paths as your project develops; membership does not come with a required order. The Resources guide explains how the library works.
- Build and operate agents — Construct capable agents, give them real work, and run them reliably.
- Build Apple products — Turn Apple platform capabilities into thoughtful, reliable products.
- Run the solo studio — Build the operating system and business judgment behind a one-person product studio.
- Market your products — Find the right audience, communicate the product clearly, and build a repeatable path from launch to learning.