I use them every day, multiple times a day, for work. The “context window” limits are largely a non-issue now for everyday work (though not for truly long-form content, like a novel-length book).
For everyday things like code snippets, their context windows are already pretty large by default (especially Gemini’s). Even the smaller ones, like ChatGPT’s, can be made significantly more useful with personalization prompts (in the options, you can give it a prompt that it applies to all your queries, like a preferred coding/writing style, length and tone of reply, etc.)
And specifically for coding, there has been an explosion of development in agentic coding IDEs and CLIs (Claude Code, Codex, Cursor, Windsurf, etc.) where they can look at an entire codebase and semi-autonomously make changes, add features, fix bugs, etc. They handle all the chunking and contextualization on their own, calling iterations of sub-agents as necessary. Even if you don’t use them agentically, those same systems are also pretty good at being able to ingest your entire codebase for the purposes of being able to answer questions or write documentation or an API against it, etc.
Beyond that, more generally, we have other threads like Is AI overhyped? - #258 by Voyager or the Topics tagged ai full of similar discussions.