A few weeks ago, a paper landed that should have been front page everywhere. A research team demonstrated that CRISPR can selectively shred cancer cells by targeting the DNA of proteins long considered "undruggable", including KRAS, a mutation present in roughly 30% of all human cancers. It worked. In mice, at least. And the mechanism is completely different from how we've tried to fight cancer for fifty years.
To understand why this matters, you have to understand what "undruggable" means. It doesn't mean nobody tried. KRAS was discovered in 1982. For forty years, pharmaceutical companies threw everything at it: small molecules, peptides, antibodies, even experimental RNA therapies. Nothing stuck. The protein's surface is essentially a smooth ball with no pockets deep enough for a drug to bind. It's like trying to grab a polished marble with oven mitts. The entire oncology field accepted that some proteins were simply chemically impossible to target. You worked around them or you gave up.
CRISPR sidesteps the whole paradigm. Instead of trying to bind the protein, it goes for the blueprint. The technique (in simplified form) uses guide RNA to find the cancer-driving DNA sequence, then a Cas enzyme cuts it. The cell tries to repair the cut, makes errors, and the gene is knocked out. No more cancer protein. No binding pocket required. You don't need to find a chemical key that fits the lock. You just destroy the lock's manufacturing plant.
This is not incremental progress. This is a category change. For every KRAS-like protein that was written off as untargetable, the question shifts from "can we find a molecule that binds it?" to "can we design a guide RNA that finds its gene?" Those are fundamentally different research programs, and the second one is vastly more tractable for a whole class of targets.
The paper got about 861 points on Hacker News. That's actually quite a lot for a biology story on a tech forum. But compare it to the coverage of any mid-tier AI model release. A startup raises twenty million dollars on a vague pitch about "AI for sales" and it gets ten times the column inches. A technique that could rewrite cancer treatment gets a good thread on HN and a few science press articles, then the conversation moves on.
This is the attention monoculture problem. The tech world has collectively decided that AI is the only story worth following. And to be fair, AI is a genuinely big deal. But the assumption that it's the only big deal, or even the biggest one, is a failure of imagination. A CRISPR-based cancer therapy that works on undruggable targets would save more lives in ten years than every LLM ever released. It would change the economics of oncology. It would make whole categories of drug development obsolete.
None of this diminishes AI research. It just puts it in perspective. The world doesn't stop changing just because everyone is looking at the same thing. Breakthroughs happen elsewhere, in silence, and they reshape reality while the attention market is busy with the next hype cycle. The question is whether we can look up occasionally, or whether we'll let the algorithm decide what matters.