AI miracle breakthroughs are here, but nobody seems impressed

In a LessWrong post, author Expertium argues that AI is rapidly approaching capabilities that would have been considered science fiction a decade ago—solving decades-old math conjectures, drafting novel bioweapons, and hacking software. They cite recent milestones like the Jacobian conjecture counterexample, OpenAI's cycle double cover conjecture paper, and covert message boards created by OpenAI agents. Yet, people remain unimpressed. The author recounts sharing vibemathed.com (a math visualization tool) with friends, parents, and Discord communities, receiving zero 'holy shit' reactions. Even a mathematician acquaintance said there's 'no reason' to believe LLMs will ever solve Millennium problems, dismissing the current evidence.

The post also observes that even daily users of frontier coding tools like Codex and Claude Code—five of whom the author knows—don't expect to lose their jobs or see ASI in their lifetime. Only one of the five believes AI safety funding needs to increase. One user unironically compared LLMs to NFTs, saying they'll vanish in 1-3 years. The author's mother (a translator) is unimpressed by LLM translation; his chemist friend is unimpressed by AI-driven science. The essay suggests a strange dissonance: AI is producing sci-fi-level results, but the public's reaction is 'a shrug.' The author admits this is anecdotal but argues it's a pattern worth examining. The piece ends mid-thought, but the thesis is clear: exponential AI progress is being met with widespread indifference and skepticism—a phenomenon that could have profound implications for adoption, policy, and how we prepare for transformative change.