I suspect the xAI merge was a manoeuvre to pump SpaceX whilst actually quietly beginning to scale down xAI. It’s a money losing venture.

I really like this. Often I just want to watch something but YouTube insidiously steers me towards doom videos, even after clearing cookies. I like that this bypasses the algorithm and lets me just watch stuff, and if there's nothing interesting playing, I can just go do something more productive.

I think it's an LLM wrapper.

The only potential benefit I can think of is getting results faster, but it's really not important enough to optimise for.

Maybe a dual system of paper ballots and e-voting could be good so that they cross check each other. Can't stuff paper ballots without manipulating the digital counter, can't manipulate the digital counter without stuffing ballots. A digital counter also enables meta analysis which could identify suspicious patterns, like a wave of votes for a particular candidate.

I enjoyed this post very much. I think it’s a matter of time until there’s a horrific LLM-powered attack on the masses. Companies are giving LLMs access to all their data and letting employees with poor understanding of the risks use LLMs to query over that data. This is going to end badly.

AI is just too tempting. Even when the risks are understood, they’re still too hypothetical for anybody to care, or volunteer to pour water on the CEO’s whole “efficiency” push.