Biological De-Aging Contests and Why LLMs Still Don't Reason

Tech media loves a shiny object. Between breathless press releases about artificial general intelligence and venture capitalists throwing billions at longevity clinics promising to reverse the calendar. It is easy to get whiplash. But beneath the noise, two recent items from MIT Technology Review caught my eye. One explores a radical biological de-aging contest where researchers race to reset cellular clocks. The other tackles a much quieter, less glamorous truth: large language models still cannot actually reason.

Let us talk about the aging race first. We now have formal competitions designed to pit scientific teams against each other in the quest for biological youth. It sounds like science fiction. Yet, unlike the speculative vaporware dominating software feeds, this kind of hard biology forces a sobering reality check. You cannot fake cellular degradation with a clever prompt. Either the organism responds to the intervention at a molecular level, or the experiment fails.

Meanwhile, the software industry continues to spin entirely in the opposite direction. We are told that the next checkpoint model will finally bridge the gap to human cognition. Except, as real computer scientists keep pointing out, predicting the next likely token is not the same as understanding logic. When an LLM solves a complex math problem, it is usually regurgitating patterns from its training distribution rather than working through first principles. It mimics thought. It does not possess it.

I find this contrast fascinating. Because shortcuts don't work, in hardware and biology, we chase brutal, stubborn physical truths. In software, we constantly try to hallucinate our way past fundamental limitations with more compute and better marketing. Maybe it is time we stopped buying into the hype cycles and start to focus on the unsexy, deliberate work of building things that actually function.