Ask HN: The next evolutionary step in LLM usage?

[tomaytotomato]

I'll keep this post short and sweet, we have seen several steps in the evolution of LLM (large language model) usage.

1. Chat

2. Autocomplete

3. Embedding knowledge using RAG

4. Tool calling by LLMs (CLI or MCP)

5. Agentic LLMs executing task(s)

What do you see the next step or iteration?

My theory is that we will get more quantization and efficient models by the end of 2026 and my hope is that we will have mini models that wrap around tools (I call them domain agents) that just give answers without bloating context.

i.e. the Domain agent gives the calling agent the sausage but doesn't explain how the sausage was made.

Curious what your theories are, but I think we might need a whole rethink of the architecture of LLMs being combined with tools etc.

[bohdanstefaniuk]

I would like to see more and more local LLMs. Also it would be great to have some very simple tooling to teach LLMs based on my data. For example I have huge documentation or lots of data and I don't want to manually build RAG. I want to have an option to upload all my files, wait few hours-days and then we able to work with my data across all conversation or at least all project scoped conversations

[wall-eee]

I expect to see domain agents/mini models running on older machines so RAM, and resources are not an obstacle anymore.