Blogmark

My advice about using AI is simple: use AI as an assistant, not an expert, and use it judiciously. Some people will object, “but AI can be wrong!” Yes, and so can the internet in general, but no one now recommends avoiding online resources because they can be wrong. They recommend taking it all with a grain of salt and being careful. That’s what you should do with AI help as well.

Skeptics of the uses of LLMs typically point to strawman arguments and gotchas to wholesale discredit LLMs*. They're either clinging too tightly to their bias against these tools or completely missing the point. These tools are immensely useful. They aren't magic boxes though, despite what the hypemen might want you to believe. If you take a little extra effort to use them well, they are a versatile swiss army knife to add to your software tool belt."Use [these LLMs] as an assistant, not an expert."

*There are soooo many criticisms and concerns we can and should raise about LLMs. Let's have those conversations. But if all you're doing is saying, "Look, I asked it an obvious question that I know the answer to and it got it WRONG! Ha, see, this AI stuff is a joke." The only thing you're adding to the dialogue is a concern for your critical thinking skills.

If you approach AI thinking that it will hallucinate and be wrong, and then discard it as soon as it does, you are falling victim to confirmation bias. Yes, AI will be wrong sometimes. That doesn’t mean it is useless. It means you have to use it carefully.

Use it carefully, and use it for what it is good at.

For instance, it can scaffold a 98% solution to a bash script that does something really useful for me. Something that I might have taken an hour to write myself or something I would have spent just as much time doing manually.

Another instance, I'm about to write a couple lines of code to do X. I've done something like X before, I have an idea of how to approach it. A habit I've developed that enriches my programming life is to prompt Claude or Cursor for a couple approaches to X. I see how those compare to what I was about to do.

There are typically a few valuable things I get from doing this:

1. The effort of putting thoughts into words to make the prompt clarifies my thinking about the problem itself. Maybe I notice an aspect of it I hadn't before. Maybe I have a few sentences that I can repurpose as part of a PR description later.
2. The LLM suggests approaches I hadn't considered. Maybe it suggests a command, function, or language feature I don't know much about. I go look those things up and learn something I wouldn't have otherwise encountered.
3. The LLM often draws my attention to edge cases and other considerations I hadn't thought of when initially thinking through a solution. This leads me to develop more complete and robust solutions.

I’ve used AI to help me write code when I didn’t know how to get started because it needed more research than I could afford at the moment. The AI didn’t produce finished code, but it got me going in the right direction, and iterating with it got me to working code.

Sometimes you're staring at a blank page. The LLM can be the first one to toss out an idea to get things moving. It can get you to that first prototype that you throw away once you've wrapped your head around the problem space.

Your workflow probably has steps where AI can help you. It’s not a magic bullet, it’s a tool that you have to learn how to use.

This reiterates points I made above. Approach LLMs with an open and critical mind, give it a chance to see where they can fit into your workflow.