Downsides of using LLMs for coding

Using Claude Code and alike with each agent working in a tab in the terminal feels very productive. You spend very little time waiting for AI since you basically start a task and go look at the next agents output.

But with this workflow for 6 months I can now see the downsides:

Finishing software is much harder

Working on 3-4-5 different features at the same time means you work in parrellel but you move slower on a single feature. It feels incredible moving on that many features, but it is harder to finish each. To finish a feature you need to deploy it and test it in the end, making sure it actually works.

AI can't do the final quality check and it is quite easy to postphone that, because you are getting adrenline for working on another exciting feature.

AI likes new features

But it does not keep track of if you actually finished any of the features. It does think about tracking usage or asking users. You are the only user to the LLM and the applications users is not something it has in mind.

A normal developer will have in mind that he/she makes something that in the end breaks, it did not matter how many cool features you shipped or what cool tech you talked with your coworkers.

Short term memory

It can't remember features it has implemented or ideas that it maybe wanted to. I have it write down docs for features that could be potentially features for the future, but it does not try to see if the feature exist there or have any recollection of them like a normal developer.

A good developer has long term memory of features and bugs that happened in the past or is going to be planned in the future. That is why software developers can't be taking in from the street, you have to remember and learn and predict the future.