Yep, bind mount the data and config directories and back those up. You can test a backup by spinning up a new container with the data/config directories.

This is both easy and generally the recommended thing I’ve seen for many services.

The only thing that could cause issues is breaking changes caused by the docker images themselves, but that’s an issue regardless of backup strategy.

You’re right, but people over a certain tax bracket are also pretty good at not paying taxes.

Just want to expand on this as it’s the most direct explanation.

With two die there are 6 ways to you can roll a seven (each side has one way to add up to seven), and 36 total combinations (6 sides * 6 sides). So the odds are 6 times out of 36 or 6/36.

With one weighted die, you have a set value (say 3 for example). There is only one side on the other die that will equal 7 (4 in our example). So you have 1 out of 6 possibilities, or 1/6 chance.

However, this is only true for 7. If you were targeting 2 for example, the odds can change substantially. Normally you have one way to get 2 (1 and 1) so you’d have 1 out of 36 possible rolls or 1/36. If the weighted die was weighted to 6 though, you’d never be able to get 2, so your odds would be 0.

It’s really great software, and the android app is great (given it supports offline mode). I just wish the folder structure was simpler/flexible. Makes me tempted to try to make a pull request, but haven’t done something like that in ages.

To be honest, I think there is diminishing returns on more money (look at star citizen for example).

I feel like it would end up a live service game (where the extra money results in content over a long period of time) or just become bloated.

I saw Boox called out, but not the Boox Palma². I just got it and it’s been pretty nice. The major draw is the form factor though as it’s phone sized making it pretty portable.

It runs android and I’ve set it up to work with AudioBookshelf and Komga

AudioBookshelf, while designed for audiobooks, allows you to download books for offline reading and seemed the best all in one for books self hosting. It also has a native android app.

Komga seems pretty amazing for manga and comic books (haven’t settled on an app, just using the browser now). The e-ink display isn’t the best for reading this medium, but it’s not terrible for black and white comics.

Since both of those are self-hosted solutions they could integrate with readarr pretty easily (although audiobookshelf’s folder structure can be frustrating).

Yeah, that would be my recommendation too. Anything else will produce a worse experience (laggy and slow) and more complexity to get setup/maintain.

Powerstone is such a great game!

Some of the commands I use a lot for debugging containers, in case you go down that route:

• docker run --rm -it --entrypoint bash <image_name>
• This command let’s you enter a docker image with bash so you can test commands and treat it like a temporary VM. Great way to see how the image is setup
• docker exec -it <container_name>
• This let’s you enter a running container with bash. A great way to inspect why something might not be working or check mounted volumes, etc.

Once you know how to use docker/containerization it’ll be the only way you want to deploy applications. Most popular applications will also have good guides on how to setup/config the container, but sometimes you’ll need to read up on docker and Linux to figure things out.

Have they gotten better? I had the first GDP Win and it was really underwhelming.

I mean, there is a lot wrong with it, but every language has its quirks. Generally I like discussing it’s actual flaws cause it helps me better understand the language.

Maybe I’m missing something, but:

• Python Lambda Functions - https://www.w3schools.com/python/python_lambda.asp
• Python map - https://www.geeksforgeeks.org/python-map-function/
  • They do have list comprehension, which let you basically one line a map operation more intuitively too.

Im not sure I’d call cops expert interrogators, it’s more like they’re playing with a stacked deck.

Yeah, they aren’t trained to make “correct” responses, but reasonably looking responses; they aren’t truth systems. However, I’m not sure what a truth system would even look like. At a certain point truth/fact become subjective, meaning that we probably have a fundamental problem with how we think about and evaluate these systems.

I mean, it’s the whole reason programming languages were created, natural language is ambiguous.

While it isn’t apparent from looking at recent events, senators are actually some of the most powerful people in US politics. That’s because they are the smaller chamber of the US legislative branch.

Why I say they are powerful is that they (along with the house) dictate laws and the budget for the US. While the house, the larger chamber of the legislative, performs a similar role the house has 435 people vs the senate’s 100 (meaning the power of the senate is less diluted).

While looking at the current affairs in the US might seem to imply Trump is all powerful. Right now he is only able to do what he is doing becuase Republicans effectively control all three branches of the US federal government (legislative - makes laws, executive - executes laws, judicial - evaluates/interprets laws).

If they wanted to, the senate could stop a lot of what Trump is doing. If they had support of the house, they could literally kick him out of office. While the judiciary also has a lot of power, it’s more reactive rather than proactive when compared to the legislative.

So In summary a senator is one of the top law makers (as in literally defines the law) in the US. And if senate and house work together, they are the most powerful branch of the US government. The main issue is getting 635 people to agree on anything, which is why over the past couple decades the executive had been able to exert power power/influence in the US.

If you’ve been paying attention to the field, you’d see it’s been a slow steady march. The technology that LLMs are based in were first published in 2016/2017, ChatGPT was the third iteration of the same base model.

Thats not even accounting for all the work done with RNNs and LSTMs prior to that, and even more prior.

Its definitely a major breakthrough, and very similar to what CNNs did for computer vision further back. But like computer vision, advancements have been made in other areas (like the generative space) and haven’t followed a linear path of progress.

I think that’s context relevant though. If we think about when dates are most frequently used (news, business, planning) it’s typically within the year (or month will give context).

That added with the fact it’s not uncommon in some situations to just provide month/day.

That being said, I don’t think either is better or worse. Just a preference kinda thing, unlike the issue between metric and imperial units.

Why is this downvoted?

It’s the right response, the top link is giving creditability through a ranking algorithm and is not guaranteed to have the right info. An LLM is trained on large corpus of (hopefully) quality data, but may not return the right information. Both may lead you to the wrong results and it’s always been the users responsibility to verify information.

The only major difference between search and an LLM is that the LLM believes it knows the answer and search just tells you “this is the most relevant thing I could find”.

Hmm good to know, we have a couple and they seem great so far, but wasn’t sure how others felt.