The timeline could easily be 50 or 100 years. No emerging development of technology is resistant to diminishing returns and it seems highly likely that novel breakthroughs, rather than continuing LLM improvement, are required to reach that next step of reasoning.