Magnetically levitating the model in the tunnel and measuring the forces by measuring how the magnets need to be driven to keep the model in place is pretty cool.
> By definition you can't protect yourself from the entity that provides you the software you use, because you have now way to guarantee that they aren't going to backdoor you.
Well that’s just wrong, just make it open source and do reproducible builds
By self-modifying the software. Currently the model harnesses only allow the model to modify its own prompt (which could be considered a really weak kind of learning), but theoretically, a model could design and train its own replacement and run that, continuously improving itself. I’m not sure if LLMs will be able to do that but the static hardware has nothing to do with it (since the bits on the harddrive aren’t static).
Opacity is fixed at 1. and obviously you can just break your own site until you get permission, but then people will probably leave before they give your website which has no business accessing the camera permission.
Not closely related to the app itself, but how does the worker count graph work? Do you count them visually and put in the data? Or how do you know how many ants you have?
I think some browsers display the query params muted compared to the part before (or maybe I’m confusing it with the behavior that the domain is highlighted to prevent phishing?). But technically, you can just parse the request url however you want. I also think it shouldn’t matter with regards to any caching layers inbetween.