The part I’m calling out as untrue is the „magic 8 ball” comment, because it directly contradicts my own personal lived experience. Yes it’s a lying, noisy, plagiarism machine, but its accuracy for certain kinds of questions is better than a coin flip and the wrong answers can be useful as well.

Some recent examples

• I had it write an excel formula that I didn’t know how to write, but could sanity check and test.
• Worked through some simple, testable questions about setting up project references in a typescript project
• I want to implement URL previews in a web project but I didn’t know what the standard for that is called. Every web search I could think of related to „url previews” is full of SEO garbage I don’t care about, but ChatGPT immediately gave me the correct answer (Open Graph meta tags), easily verified by searching for that and reading the public documentation.
• Naming things is a famously hard problem in programming and LLMs are pretty good at „what’s another way to say” and „what’s it called when” type questions.

Just because you don’t have the problems that LLMs solve doesn’t mean that nobody else does. And also, dude, don’t scold people on the internet. The fediverse has a reputation and it’s not entirely a good one.

Well that’s just blatantly false. They’re extremely useful for the initial stage of research when you’re not really sure where to begin or what to even look for. When you don’t know what you should read or even what the correct terminology is surrounding your problem. They’re “Language models”, which mean they’re halfway decent at working with language.

They’re noisy, lying plaigarism machines that have created a whole pandora’s box full of problems and are being shoved in many places where they don’t belong. That doesn’t make them useless in all circumstances.

Sure, but you at least have something to work with rather than whatever you know off the top of your head

Because it’s like a search box you can explain a problem to and get a bunch of words related to it without having to wade through blogspam, 10 year old Reddit posts, and snippy stackoverflow replies. You don’t have to post on discord and wait a day or two hoping someone will maybe come and help. Sure it is frequently wrong, but it’s often a good first step.

And no I’m not an AI bro at all, I frequently have coworkers dump AI slop in my inbox and ask me to take it seriously and I fucking hate it.

Charybdis, hippo, appa, Momo, pabu

If you want to self-host, I recommend a used business thin client, docker + docker-compose, and Tailscale for access away from home if needed. Don’t forget to dump & back up nightly.

Edit: thin client because it beats a pi in every respect and doesn’t run on an SD card. Tailscale because you don’t have to open ports in your firewall and point a public domain at your house.

Or you could use hosted services, neon.tech and turso both offer really generous free tiers for SQL databases.

Or you could use a notebook and pen. Sometimes simplicity is king.

The trickier part here his connecting your domain to your raspberry pi and allowing the big internet to access it. You have a few options:

• Set up dynamic DNS to direct your domain name to your (presumably dynamic) home IP address. Assign the rpi a static IP address on your home network. Forward ports 80 and 443 to that address. The world knows your home IP address, and you’re dependent on your router for security. No spam or DDOS protection.
• Use a service such as cloudflare tunnel. You’re dependent on cloudflare or whoever, but it’s an easier config, you don’t need to open ports in your firewall, and your home IP address is not public. (I recommend this option.)

Either way, don’t forget to set up HTTPS. If you aren’t dead-set on using nginx, caddyserver does this entirely automatically.

I’ve been using caddyserver for awhile and love it. Config is nicely readable and the defaults are very good.

If you already have a public facing server for them to connect to then sure.

Set up Tailscale and an SSH key for remote tech support

I have 113k images going back two decades. The screenshot above doesn’t include RAW files, with those included I’m around 2 terabytes of total storage.

• Immich is in fact a photo album, and a damn good one at that.
• Immich keeps google’s grubby paws off my photos. I don’t need or want anyone datamining every precious memory I have in order to modify my behavior to their benefit.
• Immich shares photos between my wife and my phones.
• Immich ensures that if I lose my phone, my photos aren’t lost.
• Immich lets me easily re edit and re-export RAW files without creating duplicates or losing metadata
• Immich lets me conveniently share photos with friends and family without requiring them to have an account anywhere.

Mostly I self-host things when I want data synchronized between multiple devices, or I don’t want to lose it in the event I lose the device it was created on.

Also, like, phone screens are tiny and typing on them is terrible? Why would you want to do everything on your phone?

If encryption is enabled, don’t worry about it. Otherwise ‘dd if=/dev/zero of=definitely-the-bad-drive-do-not-fuck-this-up bs=4M’

Agreed, that should be many tens of pages not one. Also the mobile layout isn’t very good. I think it’s important to remember that normies use their phones for almost everything.

Got a 3 year old kid with another on the way. I just need it to be reliable so the kid can watch Sesame Street and the lights keep working.

I definitely see your point, but the difference is that it’s one thing to learn. Once you know docker, you can deploy and manage anything.

Cons of containers are slightly worse disk and memory consumption.

Pros:

• ease of installation
• declarative, consistent configuration
• security some degree of sandboxing
• dependency management is solved

Stick with the containers

but there is a reason i just explained it to you

Ok but is there room for the idea that your intuitions are incorrect? Plenty of things in the world are counter-intuitive. ‘docker-compose up -d’ works the same whether it’s one container or fifty.

Computer resources are measured in bits and clock cycles, not the number of containers and volumes. It’s entirely possible (even likely) that an all-in-one container will be more resource-heavy than the same services split across multiple containers. Logging from an all-in-one will be a jumbled mess, troubleshooting issues or making changes will be annoying, it’s worse in every way except the length of output from ‘docker ps’

I can see why editing config files is annoying, but why exactly are two services and volumes in a docker-compose file any more difficult to manage than one?

I disagree with pretty much all of this, you are trading maintainability and security for easy setup. Providing a docker-compose file accomplishes the same thing without the sacrifice

• separate volumes for configuration, data, and cache because I might want to put them in different places and use different backup strategies. Config and db on SSD, large data on spinning rust, for example.
• separate container for the database because the official database images are guaranteed to be better maintained than whatever every random project includes in their image
• separate networks because putting your reverse proxy on a different network from your database is just prudent