The Part of Building with AI I Want to Show You

By William Paul “Bill” Herald

An AI answer can arrive before you know whether it is useful. The work I want to show is what happens next: checking it, changing it and deciding whether it solves the task.

That includes choosing the task, working out what information belongs in it, examining the result and deciding what to change. Those steps can tell us more than a polished demonstration on its own.

My own project is the Personal Life Operating System. Privacy, control and my involvement in decisions are central to the direction I want to take. I also want the public lessons to make sense to people who have no interest in the software underneath.

The question I want a reader or viewer to be able to answer is straightforward: what could I learn from this and use in my own work?

Show the purpose first

A useful account of building something should explain the job before introducing the technology. Who would use the result? What would make it worth their time? How would they know it helped?

I am exploring those questions while considering whether a focused part of my work could become a product. That remains an exploration. A product needs a result someone values, a practical way to deliver it and a support commitment I can keep.

I want to establish those things through evidence. A proposed feature is a reason to investigate. It becomes a credible promise only after the relevant work has been built and tested. I want the useful details of that process to be visible: what the task required, what was checked and what remains uncertain.

Make the person's contribution visible

AI can be part of preparing ideas, drafts and technical work. The person still brings the purpose and the judgment needed to decide what to keep.

I want to show that contribution clearly. If a direction becomes too complicated for the intended audience, changing it is part of the work. If an answer needs checking, that checking belongs in the account. If a tool cannot complete the task within a useful amount of time, the limitation deserves an explanation.

This also affects what I mean by productivity. I want to include the effort needed to understand the output, correct it and reach an acceptable result. The finished task matters more to me than the moment an answer first appears.

Share the lesson without exposing the private material

There is a practical boundary between explaining a method and publishing everything that went into a private system.

I can describe a question, a design choice or an uncertainty without disclosing family information, employer material or someone else's confidential work. When a demonstration needs an example, I want its source and status to be clear. A proposed test should be called a proposed test. An actual result should have evidence behind it.

That boundary helps me keep sharing. It gives public work a clear purpose while respecting the information that does not belong in it.

Keep learning and earning distinct

I would like this work to contribute to a sustainable source of supplemental income. I also want to keep sharing useful lessons freely.

A paid offer would need a defined result and a price that makes sense for the effort involved. Free educational material should be useful on its own. Neither goal is served by promising more than I have demonstrated.

That is the approach I want to develop here: explain what I am trying, show what can be supported, acknowledge the remaining work and invite useful discussion.

You can find the broader project through wphc.us.

When you read about someone building with AI, what would help you judge whether their approach could be useful in your own work?