I think it may play out in the opposite direction.

If you're developing a new programming language today, one of the assets you need to prepare is a short (~10,000 token or less) LLM-friendly guide to your language, plus a bunch of examples that coding agents can search through and crib from.

Done well, I expect this could accelerate the adoption of your new language - as users can start prompting their coding agents to build with it before they've even finished reading the tutorial themselves.

Your disadvantage will be that LLMs won't recommend your language when people ask "what could I build this in", but people discovered new languages via word-of-month before LLMs came along and I expect that to continue, especially if your language has something genuinely new and interesting to offer.