FIVE PRINCIPLES
HOW TO BUILD SOLO
WITH AI.
My job is to bring the direction, taste, and judgment. AI helps me turn those into something people can use.
Building solo with AI starts with a useful idea and a manageable next step. These five principles help turn a conversation into a working product without losing sight of the person it is for.
Start with the person and the problem
What it means
- Name who you are helping and what gets easier for them.
- Define a useful result before choosing tools or features.
What it’s not
- Starting with a tool and hunting for a reason to use it.
- Assuming an impressive demo is a useful product.
- Complete this sentence before opening a tool: “I’m building this for ____ so they can ____ without ____.”
- Show that description to someone who fits. Ask how they handle the problem today and what makes it difficult.
Become the director
What it means
- Set the vision, break the work into clear tasks, and give AI a concrete brief.
- Decide what happens next, give each tool a clear job, and guide the work with examples and feedback.
What it’s not
- Expecting AI to read your mind.
- Handing AI the whole project and hoping it figures out the direction.
- Give AI a brief with the audience, desired result, constraints, and an example of what good looks like. Ask it to explain its plan before building.
- Break the project into clear jobs. Give each tool or agent a specific task, a definition of done, and a point where you’ll review the work.
Build small, test, iterate
What it means
- Build a small demo that makes the idea tangible and gives someone something to try.
- Test it, learn what works, and improve one piece at a time.
What it’s not
- Asking for the whole business in one giant prompt.
- Adding features before the core experience works.
- Build a demo that lets someone complete one useful task from beginning to end. Put everything else on a “later” list.
- Watch someone try it without coaching them. Fix the first place they get confused, then test again.
Guide with your taste and judgment
What it means
- Use the product yourself and guide how it looks, feels, and works.
- Check important claims, test the core flow, and get help when needed.
What it’s not
- Accepting output because it looks polished or sounds confident.
- Letting speed excuse a confusing or unreliable experience.
- Collect three examples you like. Explain what you like about each: the spacing, wording, behavior, simplicity, or attention to detail.
- Review AI’s output as a user. Replace “Make it better” with specific direction: “This heading is vague,” “This takes too many clicks,” or “The main action should be easier to find.”
Turn each build into a head start
What it means
- Save decisions, useful instructions, and working versions.
- Reuse what worked and let customer feedback guide the next change.
What it’s not
- Starting from scratch every time a chat disappears.
- Automating a process you have not yet made useful.
- After a build, save the useful instructions, decisions, reusable pieces, and mistakes you don’t want to repeat.
- Before starting the next project, review what you already have. Choose one thing to reuse and one lesson to apply.
KEEP EXPLORINGStart with AI
Back to top