Your source number 2 involves a hard interdisciplinary research problem: “what is meaningful language use?” My grad school thesis was tangentially related to it, so I’m most familiar with it from the AI perspective. Early AI researchers quickly realized that you can’t just dump a dictionary into a computer and suddenly have it understand language. You can add high-level “scripts” to the computer (Schankian scripts), but then it will just be manipulating symbols (chinese room problem). You can tie the symbols to things in the world (symbol grounding) or to its own processing (embodied meaning) but how do you coordinate those symbols with other agents, be they people or machines?

Think about that last question for a moment. Do you have an answer? I don’t think anyone does yet, so whatever you’re thinking is probably a good start towards further reading. @fubo@lemmy.world’s reply points out some of the issues involved, and these issues suggest the problem’s interdisciplinary nature: psychology, sociology, corpus linguistics, philosophy (both analytic i.e. Wittgenstein and Kripke, and continental as suggested by @AbouBenAdhem@lemmy.world’s reply), cognitive science, neurolinguistics, etc. Literary theory is fun too, they say things like: an interpretation is situated, subjective, and performative. OK, sounds great, but how do you turn that into something that a computer does? It turns out that there are a lot of great ideas, but there’s still a lot of work to do to tie it all together. (and unfortunately way too many people think deep neural networks / LLMs can just solve it all by themselves grumble grumble…)

Given the above, to answer your specific question: “Is [meaningful] language mostly solipsistic?” I think most people would say “probably not”, with the caveat that it depends on how you define your terms. Clearly there are very important processes that work only in your own cognitive system, but it seems likely that external factors also play a necessary role.