I originally posted this post on my blog.
I'm not an AI evangelist, and I'm not a hater either.
I've tried AI for coding. And after a week or two, I noticed how dependent I was becoming. Since then, I've used AI for small coding tasks, like generating small functions or finding typos, not to treat English as my main programming language.
Marcus Hutchins, a security researches, puts it boldly in his post Every Reason Why I Hate AI and You Should Too:
I'd make a strong argument that what you shouldn't be doing is "learning" to do everything with AI. What you should be doing is learning regular skills. Being a domain expert prompting an LLM badly is going to give you infinitely better results than a layperson with a "World’s Best Prompt Engineer" mug.
I agree with the core message.
When everybody is relying on AI, it's time to go back to old-school habits:
- Read code
- Write trustworthy tests
- Devour classic textbooks
- Troubleshoot bugs with pen and paper
And outside coding: read books on paper, take notes by hand, and write our own summaries. To develop taste and judgment.
Using AI is like holding a calculator on a math exam. Even the best calculator is useless if you don't know what to compute.
Build skills, then leverage AI.
I used to think coding was only about cracking symbols. That's where AI shines. But coding is also about talking to non-tech managers, negotiating deadlines, and saying no politely.
And that's why I wrote, Street-Smart Coding: 30 Ways to Get Better at Coding, to share the skills I wish I'd learned to become a confident coder.
Get your copy of Street-Smart Coding here. It's the roadmap I wish I had when I was starting out.
Top comments (4)
Can't agree more with this to be honest. I always avoided using AI too much or being too dependent on it, but it seems like these tools are not going anywhere, so knowing how to use them properly and efficiently right now is somewhat similar to knowing to use an IDE properly in my opinion. HOWEVER, that doesn't mean that it's ok to just dump everything to an AI model or an agent and let it venture in the wild with the code, it's definitely the most important that before getting to this level, that you have full understanding of the problem at hand, have already devised solutions on your own that at least can be sufficient to solve the problem, then handing maybe the syntax generation to the AI is an okay thing.
This is the way! And AI is just like showing who has good coding hygiene (or practices). Rushing to code without a good understanding of the problem has always been a bad idea, AI is just amplifying the problem.
That "prompt engineering" thing reminds me a joke about a conversation between a person who called herself a TikTok model, and an "average" person. Adapted to this, it would be:
:-P :-D
🤣🤣🤣 OMG! A good one!