Hacker News Reader

Comment by sschueller | Hacker News Reader

I recently needed to create a Dockerfile which I have done many times before. Instead of opening the documentation I used ChatGPT.

It took 3 times longer than if I had done it by hand because I just kept copy and pasting each error back to ChatGPT ending up in a dead end having to start over from scratch. As soon as you have a requirement that is not the 90% of cases you can quickly end up in a dead end especialy if you have more than one such requirement.

Once the LLM starts hallucinating you can easily get into corner where you will need to reset and start from scratch.

View on HN

> As soon as you have a requirement that is not the 90% of cases you can quickly end up in a dead end

You framed that as a criticism of ChatGPT, I think that is indeed the point the OP is making. IMO if this is something you used to do you could get a nice starting point with a single prompt and then go from there, then that would take 3 times less compared to reading the documentation from get-go.

I haven't written any C++ in ages so I used a local Ollama3.2 to do a simple class and struct. The AI wanted me to use the `friend` keyword on class functions that weren't being accessed from anywhere else... I did correct itself after I asked but it took some intuition and googling to get there.

This is basically the modern "google-driven" development experience as well. The copy-pasting from SO trope has existed for some time, but it's much worse in the last 5-7 years with the increase of blogspam/LLM generated content on Medium that's SEO'd to the top and effectively just rewriting the "getting started" page of some language/framework/tool/etc. for resume boosting points.

It seems like a lot of the models in ChatGPT and Copilot were trained on that content and in turn tends to produce a lot of dead end solutions for anything that isn't the 90% cases and often leads to more pain than reading documentation and building a solution through iteration/experimentation.

> I recently needed to create a Dockerfile which I have done many times before.

> It took 3 times longer than if I had done it by hand because I just kept copy and pasting each error back to ChatGPT

You know that if you already know what you're doing you can tweak the output directly instead of mindlessly pasting it back into ChatGPT, right?

Its harder to tweak someone's else code than to tweak your own.

yeah, I’ve had many moments in my AI-driven development cycles looping with the LLM in a dead end, eventually giving up and falling back to regular old programming.

I think LLMs are great for reducing the friction with just starting a task, but they end up being more of a nuisance when you need to dig into more nuanced problems.

I used to be like you. This is why you may want to try Cursor paired with Claude for this.

Set Cursor in "agent" mode (with Claude as a model).

Ask what you need.

It will automatically test your Docker deployment and configuration.

Then you become like me, a monkey clicking "Accept".

But a happy monkey.

Eventually as your project grows in complexity, what you're asking of the model will exceed the intrinsic limits of what the model can do, and you won't be able to troubleshoot your problems yourself because you didn't learn anything along the way.

Reply on news.ycombinator.com