Mostly I'm seeing breakthroughs in the ways we use LLMs. Agentic harnesses, MCPs, etc - its the wild west still but we've come a long way from a basic chatbot. Gains are now coming from tools that make better use of the LLMs existing capabilities, and put guardrails on their worst tendencies.

I personally feel like we've plateaued in raw model intelligence (even regressed, I find sonnet 4.6 to perform better than Opus 5) but we've gifted them new skills that allow them to run for longer and explore the search space more thoroughly, making them more effective at the same level of "intelligence".

Take two smart people and a problem to solve. Give one person the tools, the other person nothing. The one with the best tools wins. At some point its more about abilities than raw intelligence. Watching a "frontier" model fumble with basic syntax is still common, but not if you give it treesitter.

Real breakthrough happens with custom trained models

Having your grandpa create a todo app over the weekend is great, but churning out code faster does not really make human science progress.

Custom models trained on niches could perform better scientifically, for example solving protein folding problems.