Admission: I find myself increasingly leaning into AI. We've had some *serious shit* come up for our cats, over the past couple years. We've had a couple of them pass during this time. We have three now where two are grappling with serious shit: one with FGESF and another 16 y/o dealing with typical age-related stuff. Claude, in particular, has helped enormously searching for and digging into studies/the underlying medical science to get the best possible treatment. This has helped to aleviate a significant amount of my fear for their quality of care and life. FGESF... JFC. It's an enormously poorly understood condition where the science is slowly evolving. *I* (not Claude, IIRC) found a study that none of the internists we've worked with even knew about! But Claude has helped me keep an eye on the tests to ask for, the questions to ask, the medication protocols, all of it. As someone #actuallyaudhd, because of the extreme overwhelm of life at age 53, it's entirely too easy for me to lose track of extremely important details in their care. Doing this all without AI assistance? There would be unforgivable (to myself) errors in their care. Increasingly, I've been setting up a system at home, using my COVID-purchased gaming rig and its 3090. The power consumption is insane but at least it's only so when it's actively "reasoning". It is increasingly impressive, and a bit frightening too, how capable the models are becoming that can run on commodity hardware. Qwen3.6-35b-a3b runs fairly quickly and gives useful answers when connected up to a search engine. I also use it for "Retrieval-Augmented Generation" (RAG) to help keep track of cat care, given the sheer number of PDFs of instructions and data points that we receive for each cat. Of course, like with all technologies, it can be wielded for evil. It's already happening. And it's going to get worse. The democritization of AI, particularly through self-hosted open weight models, will certainly empower bad actors to affect more malicious outcomes at a likely exponential rate. I don't have data for this; what follows is my pattern-matching nature digesting information. The advent of (currently) increasingly skilled/"intelligent" LLMs seems likely to outstrip humanity's maturity. This is the latestsocietal inflection point where we are at high risk of self-destruction. The difference here: AI is more accessible than the technology to build nuclear weapons. I can't speak to how easy it is to build bio weapons. But that's the crux: AI seems poised to lower the barrier to entry for creation of deadly physical tools. Then, of course, there's the infrastructure-crippling InfoSec possibilities—and this seems the most immediate threat vector as it requires the least resourcing. We seem to have already seen an uptick in InfoSec attacks on infrastructure. This Pandora's box isn't going to be closed. While there's so much potential for good, if used responsibly, the opportunity for evil is enormous.