How I keep my critical thinking skills sharp in the AI age

I’ve seen many posts online by other developers saying they feel their critical thinking skills slipping as they rely on AI more. This is a trap I’ve found myself falling into as well, and I’m slightly shifting how I approach work with AI.

You are the decider

At the core of all my work, I am still the ultimate decider. For example, let’s say I’m working on a small problem. Before I even send a message to an LLM, I think through the problem myself first and come up with a high-level idea of what I think will solve it. Once I feel good about it, I’ll send a message to an LLM saying something like, “Here’s the problem I’m trying to solve, and I think XYZ is a solid solution. However, I want to hear alternative approaches.”

This allows me to flex my critical thinking skills while still leveraging the power of AI to explore alternatives. Sometimes my solution ends up being the best idea; sometimes the LLM comes up with a better one. That’s where the power lies: being able to think through the problem myself, compare multiple options, and weigh the tradeoffs so I can make an informed decision.

How this applies to larger problems

Repeat the above section. Thank you…

In all seriousness, though, the process itself doesn’t change all that much for larger problems. Break it down into smaller pieces, do some research, weigh the trade-offs, and make a decision.

What gets amplified in larger problems is the speed at which you can make these decisions. AI is great at information synthesis, and I’ve found it profoundly helpful when researching complex problems.

This is where being the “decider” becomes even more important. Sure, you could say, “Hey AI, I need to design an ingestion pipeline for this data source and toss it into a RAG store so I can ground some AI responses,” and walk away from the computer, but that’s not going to serve you or your application in the long run.

With a problem of that scale, the details are a bit fuzzier, so I need to conduct some research. I would tweak the above prompt to be something like, “I need to design an ingestion pipeline that pulls from this data source. I know pgvector is a common approach for a RAG store, but I’m open to other options. What are your thoughts?”

That one simple question sends me down a rabbit hole of research and exploration. I can then document my findings through the LLM via specs, research docs, etc., so I can track my thinking and decisions along the way.

Conclusion

The key takeaway here is that AI is indeed a tool we can use to amplify our work. I’ve found it tremendously powerful for research and exploration, but the critical thinking and decision-making still lie with us. I’ve found that being the decider better serves me and the applications I build. You can still accelerate with AI while keeping your skills sharp.

Go forth and use that big brain of yours!