Comment by ehnto | Hacker News Reader

Comment by ehnto | Hacker News Reader

Parent Full thread

> If you're completely new to the problem then ... yes, it does.

Of course, because I am not new to the problem, whereas an LLM is new to it every new prompt. I am not really trying to find a fair comparison because I believe humans have an unfair advantage in this instance, and am trying to make that point, rather than compare like for like abilities. I think we'll find even with all the context clues from MCPs and history etc. they might still fail to have the insight to recall the right data into the context, but that's just a feeling I have from working with Claude Code for a while. Because I instruct it to do those things, like look through git log, check the documentation etc, and it sometimes finds a path through to an insight but it's just as likely to get lost.

I alluded to it somewhere else but my experience with massive context windows so far has just been that it distracts the LLM. We are usually guiding it down a path with each new prompt and have a specific subset of information to give it, and so pumping the context full of unrelated code at the start seems to derail it from that path. That's anecdotal, though I encourage you to try messing around with it.

As always, there's a good chance I will eat my hat some day.

> Of course, because I am not new to the problem, whereas an LLM is new to it every new prompt.

That is true for the LLMs you have access to now. Now imagine if the LLM had been trained on your entire code base. And not just the code, but the entire commit history, commit messages and also all of your external design docs. And code and docs from all relevant projects. That LLM would not be new to the problem every prompt. Basically, imagine that you fine-tuned an LLM for your specific project. You will eventually have access to such an LLM.

Why haven’t the bug AI companies been pursuing that approach, vs just ramping up context window size?

Because training one family of models with very large context windows can be offered to the entire world as an online service. That is a very different business model from training or fine-tuning individual models specifically for individual customers. Someone will figure out how to do that at scale, eventually. It might require the cost of training to reduce significantly. But large companies with the resources to do this for themselves will do it, and many are doing it.

Well, we don't really know if they aren't doing exactly that for their internal code repos, right?

Conceptually, there is no difference between fine-tuning the LLM for being a law expert of specific country and fine-tuning the LLM for being an expert for given codebase. Former is already happening and is public. Latter is not yet public but I believe it is happening.

Reason why big co are pursuing generic LLMs is because they serve as a foundation for basically any other derivative and domain-specific work.

AI training doesn't work like that. you don't train it on context, you train it on recognition and patterns.

You train on data. Context is also data. If you want a model to have certain data, you can bake it into the model during training, or provide it as context during inference. But if the "context" you want the model to have is big enough, you're going to want to train (or fine-tune) on it.

Consider that you're coding a Linux device driver. If you ask for help from an LLM that has never seen the Linux kernel code, has never seen a Linux device driver and has never seen all of the documentation from the Linux kernel, you're going to need to provide all of this as context. And that's both going to be onerous on you, and it might not be feasible. But if the LLM has already seen all of that during training, you don't need to provide it as context. Your context may be as simple as "I am coding a Linux device driver" and show it some of your code.

Reply on news.ycombinator.com