BakedCatboy

I like using bitwarden, the selfhosted vaultwarden server stores it with passwords and makes codes available in the app / browser extension. I also keep them backed up on a nas and synced off-site just in case.

Somebody should make an api shim that proxies openapi compatible requests to this. And since Microsoft is forcing copilot on windows 10 they're on the shit list too. Load balance all the AI workloads onto both of them through API adapters.

I'm gonna go with unlikely.

Huh, I already signed up for it because they started requiring it a while back to access historical tax return documents through the IRS website.

I just discovered how easy ollama and open webui are to set up so I've been using llama3 locally too, it was like 20 lines in docker compose, and although I've been using gpt3.5 on and off for a long time I'm much more comfortable using models run locally so I've been playing with it a lot more. It's also cool being able to easily switch models at any point during a conversation. I have like 15 models downloaded, mostly 7b and a few 13b models and they all run fast enough on CPU and generate slightly slower than reading speed and only take ~15-30 seconds to start spitting out a response.

Next I want to set up a vscode plugin so I can use my own locally run codegen models from within vscode.

Apparently it's not very hard to negate the system prompt...

I've even experienced this in the 3D printing community, where I design a highly parametric model and put lots of effort into making all of the major dimensions and qualities parameterized and dynamically adjustable, with lots of bounds checking and value clamping, with all the parameters at the top of my scad file with comments explaining what each variable does.

And then someone comes along to remix my model, says I don't want to install openscad, and just scales the entire output stl to change the dimensions, squashing all the features of the model in the process (instead of having the size gracefully adjust with all the features moving around to account), and leaving anybody starting from their work with a hard to remix mesh with no parameters.

Thank goodness it's in HDR

Lemmy really needs pixelfed's naive bayes spam detection, it would be able to easily classify the new accounts after classifying one post as spam, then it would be 0 seconds wasted.

I actually just realized lineage 21 does this too - I didn't notice because cropping a screenshot with Google photos seems to remove all the fields (I also have the build string and timezone offset). Which is weird because cropping an actual photo the same way - as you would expect - preserves all the notable fields like timestamps, phone model / lens info, and the same "Software" field which for my photos is just "HDR+ 1.0.commithashlookingstring"

Google Pixel 5, if you're using a stock rom then that sounds likely.

That's so weird, exif on a screenshot? Usually my quick fix to remove exif from something when I'm on the go is to take a screenshot of it. I'm on lineage 21 and according to exiftool there's no exif data on my screenshots.