Everyone is talking about how AI is transforming everything. Firstly, it is not AI, it is LLM.
Yet there are two important metrics in software;
- Code surface area
- Confidence (i.e. tests)
I haven't used correct terminologies, but think of it like this. Decentralisation. Decentralisation is great, it empowers. Yet if one is trying to build a platform or entire system, hell - reporoduce it elsewhere, the more AI is used, the more amnesia occurs. The less centralised the codebase becomes.
The result is a mess. Almost impossible to control. Building small apps or utilities is the way development should be done but if you are a serious player, much of the current AI offering won't work. Building platforms is another thing altogether.
We use LLM all the time as experienced developers, but at the right time.
Yet there are two important metrics in software;
- Code surface area
- Confidence (i.e. tests)
I haven't used correct terminologies, but think of it like this. Decentralisation. Decentralisation is great, it empowers. Yet if one is trying to build a platform or entire system, hell - reporoduce it elsewhere, the more AI is used, the more amnesia occurs. The less centralised the codebase becomes.
The result is a mess. Almost impossible to control. Building small apps or utilities is the way development should be done but if you are a serious player, much of the current AI offering won't work. Building platforms is another thing altogether.
We use LLM all the time as experienced developers, but at the right time.
1