If you've been putting off learning AI as a developer, you're behind the curve. There have been plenty of people who have groused about how they won't use AI. Honestly, it comes off as whiny. AI is here to stay whether you like it or not.
You don't have to know AI to use it
Honestly, you just need to jump in. Get a free account. Actually, get several. Just start playing around with it. It's really great at summarizing long bits of text. You can feed it docs and search them. It's kinda like an ice bath: you kinda just have to get in there and spend some time in it.
It doesn't have to write code
Even with something like Claude, it can do a lot of stuff for you without even writing code:
- It can write tests to give you coverage over classes that don't already have it.
- It can write documentation and generate diagrams in Mermaid.
- You can have it write "plans" for how to do changes.
- It can review plans and suggest changes and improvements.
- You can ask it questions about the code base.
- You can have it analyze the code base to look for areas of improvement.
All of these things can help give you better understanding of your existing code base, especially if it's big!
Clarity is key!
Unlike a human being, giving an AI a task like this:
Add a column to the user listing on the Coaching page in the admin
May not give you the results you are looking for. You have to be specific with AI. You'll need to tell it where you want the column, how to format it, if it's sortable, etc.
Which... really, you should be doing this already, right? If you were lazy before, you won't get the "bump" from AI that you're expecting because you'll need to take the time to be write better expectations and understandings of your work.
Different tools for different tasks
We use a number of different "AI tools":
- CodeRabbit for PR reviews
- ChatGPT for a little bit of everything
- Claude Code for things like SQL, research, and agentic coding
- JetBrains AI for commit messages and code suggestions
- Junie for helping to "plan" work and doing some light documentation.
- OpenAI for our user-facing Agentic functionality.
- Ollama for local testing and experimentation.
I've experimented personally with half a dozen other ones. It pays to play test and try things out. It's amazing how the quality is so different between various platforms. Claude is excellent at Laravel stuff. ChatGPT is pretty good with SQL.
It still requires a human at the helm
AI is getting better. But it can't understand intent. It can't read your mind. It can't know that you did something one way because you found a bug in an upstream library. It doesn't "know" that this particular approach works because timeouts are issue. It'll change tests rather than fix the bugs. It'll outright ignore it what you tell it sometimes!
But like any tool, you must learn it. You must understand its strengths and its weaknesses.
There are strengths aplenty, but there are also nuanced drawbacks and differences and changes to your own workflows that will take some time to learn and adapt to.
Top comments (0)