why are we still teaching design patterns in 2026 when AI can generate the right architecture in 10 seconds ?

Posted by Leo-neophyte12@reddit | learnprogramming | View on Reddit | 13 comments

spent 3 days trying to understand the observer pattern from tutorials and textbooks. confused the whole time.

then I asked an LLM. got a perfect implementation in 10 seconds with a clear explanation that actually made sense for my use case.

so why are we still making beginners suffer through GoF patterns, UML diagrams and abstract theory before they've even shipped a single project ?

maybe the real skill isn't memorizing patterns. it's knowing what problem you're trying to solve and letting the right tool handle the rest.

or am I missing something ?

[-]

Human_Strain_4606@reddit

You can ask AI to write code but you can't ask it to debug production system that's failing because someone used observer pattern wrong and now events are cascading in weird loops

Understanding patterns helps you recognize when the generated code is garbage or going to cause problems later. AI gives you working code, not necessarily good code for your specific situation

[-]

the 'just use AI' argument makes more sense for shipping a first project than it does once something is running in production. the patterns are really just vocabulary for 'this code is going to be confusing in 6 months.' AI helps you write it. still useful to know what you built and why it might break.

[-]

aanzeijar@reddit

I wish AI could generate "the right" architecture in 10 seconds, really.

Sadly what it usually produces is a weird mix of styles hoovered up from stackoverflow, vomited into a soup of state that maybe produces the desired effect. It takes me at least 3 iterations to get it to produce just the rough architecture I have in mind.

And to be clear: "we" are not making beginners suffer through patters and UML diagrams. I hate those with a passion.

[-]

patternrelay@reddit

I think patterns are still useful because they give teams a shared vocabulary for reasoning about tradeoffs and failure modes. AI can generate an observer implementation fast, but knowing when it becomes a maintenance problem is the harder part.

[-]

SillyOldJack@reddit

You need to know how to look for errors and problems, so you need to understand the architecture before getting AI to do your work for you.

You'll end up with a broken system and no idea why.

[-]

Ok_Lingonberry5392@reddit

It's easy to ask the llm when you know exactly what you need. Don't bother memoriesing a specific implementation if you don't want to but focus on what are the ideas and advantages of this design and when it is correct to use it in your code.

[-]

throwaway1045820872@reddit

You say you were trying to understand it for 3 days and were confused the whole time, yet later go on to say that the LLM had a “perfect implementation”.

If you don’t understand it, how can you judge that it’s perfect? Tools can assist productivity, but if you don’t understand what they are doing then you are running the risk of it not actually doing things correctly.

[-]

kawaidesuwuu@reddit

There is no right architecture; there is always a trade-off. The only difference is whether you know these trade-offs or not. And trust me, you're going to get royally fucked if you don't know the trade-offs of what you're trying to build your entire system on

[-]

Achereto@reddit

If you don't understand what you are doing, how do you expect to understand what the AI ist doing?

At the end of the day you are going to be held accountable for the million dollar bug you didn't catch when AI generated it for you.

[-]

0xC4FF3@reddit

The right tools are (among others) the patterns

[-]

NationalOperations@reddit

Why understand strategies to solve problems you encounter instead of having someone solve it for you? What could possibly be the downside