Post

Giving AI Enough Context to Help

How I use AI during project planning, and what context still can’t do on its own.

Giving AI Enough Context to Help

I’ve found AI most helpful on side projects when I bring it into the planning stage. Sometimes I want a second opinion on how a feature fits together. Sometimes I want help thinking through what someone using the software would actually need. In either case, the more useful conversation starts when I can explain the problem and let the tool ask about the parts I haven’t thought through yet.

Write down what the project needs

A file like AGENTS.md gives me a place to document the setup, constraints, and conventions I want an assistant to follow. It also helps me think through those decisions for myself. Writing them down can save time otherwise spent repeating instructions or reviewing work that started from the wrong assumption.

I’m curious about research comparing how many tokens coding tasks take across languages. That can be an interesting measurement in a particular benchmark. For my projects, I also have to think about the libraries I need and how easily I’ll be able to maintain the code after the initial development is done.

Pay attention to the decisions the system makes

I have a different set of questions when AI influences decisions about people, such as screening job applicants. The data and instructions can carry assumptions that are easy to miss. In that setting I want to know how decisions are checked, who is accountable for them, and how someone can contest an error.

That’s part of why context matters to me as an engineer. It gives the tool a better chance of helping, and it gives me something concrete to check when it does.

Adapted from my original LinkedIn post. Read the original on LinkedIn

This post is licensed under CC BY 4.0 by the author.