My first thought would be to try and fix it, then you have a nice laptop or get to be the family hero when you return it working to your brother.
Past that there are several services you might use that can take advantage of the video card: Immich image recognition, Paperless OCR, some of the local Whisper servers.
Or you have a nice set-top PC for big screen gaming and media playback.
It heavily depends on how close you are and their hover system and how much work has been spent to make them quiet. You probably heard about the Amazon ones, which are loud and hover low to the ground. There are other ones, such as the ones Zipline makes, which are quieter to begin with and hover very high, using a powered descend module on a cable. Those are basically silent from ground level.
Avoiding traffic and using an electric vehicle instead of a ICE one. In Bay Area traffic an electric drone can do a delivery in five minutes that would take a car thirty. And the Zipline style high hover + tethered descent drone is quieter than a car. I’m not sure if these are using that style or the low hover type like Amazon had been using, which are much louder.
You definitely don’t have to go full LLM for this. There are a wide variety of local spam filtering tools available, depending on what email application you use. If you really want to use modern ML, I’d go with something like Laya as it is tunable and runs locally.
Easily! My Pi runs PiHole and a print server and I’ll could probably run almost my entire stack in it if I wanted. Jellyfin would be the hardest due to transcoding CPU usage, but it’s probably ok for one stream at a time.
The other issue for me would be storage. Since I have a 32TB NAS I run everything that requires large storage on it, but a USB RAID attached to a Pi would probably work just fine.
The other issue is RAM. I rarely hit 8GB, but do use more than 4GB fairly regularly.
I have a git repo with all my docker files, config, notes, etc. I have a main overview file covering the overall system and then a directory for each machine. I edit things on my local machine, then scp and ssh to the various remotes. Once I’m happy with my changes, I commit and push to my Forgejo install on my NAS in case my main computer fails. Secrets, passwords, and keys are in my Vaultwarden.
I don’t use AI for it, but a local model could probably give me the commands to do whatever I wanted based on the repo.
I bought a 1000 foot spool a few years back when wiring my house and have used it for helping out friends and family ever since. I’m not sure how much I have left, but I’ll probably pick up another whenever I do.
Cat 6 is what, $0.15 a foot? Plus maybe another $0.50 for the connectors and boots? That’s a 16 foot snake for $3.00.
I’m not really a KISS fan, but my brother-in-law was and he had two extra tickets. That concert was the best showmanship I’ve ever seen.
Or you can go straight to the source and use the Trash Guides
Like all projects, you should think about needs, budget, and time. You’ve got a good list of what you want to run which is a good start. Budget is next, particularly considering the price of hardware these days.
FWIW, I run Jellyfin, Navidrome, the entire arr stack, Forgejo, Paperless, mealio, and Immich on a NAS with 24GB RAM and have never needed more. 8GB would have been a bit tight but probably doable. 16 will be plenty for you.
From TFA, emphasis mine:
What veans is for#
Coding agents keep their own to-do lists. Claude Code has TodoWrite, other tools have their own scratch files, and either way that list lives inside a single session and disappears when the session ends. You can’t see it while the agent is working, your teammates can’t see it at all, and nothing survives to the next session.
veans replaces that ephemeral list with real tasks in a Vikunja project. The agent claims a task before it starts, moves it along as it works, and leaves it in review when it’s done. The result is that the agent’s work is visible in the same place you already track everything else: you can watch a task move across the board in real time, comment on it, reprioritize it, or pick up where the agent left off. When the session ends, the trail is still there.
veans itself is built for machines, not people . Everything prints JSON, so it’s made to be consumed by an agent or a script. The human side of the workflow happens in Vikunja.
What Taskwarrior lets you do is something like this for people:
task add priority:high due:thursday depends:32 something I need to do
task ready # shows all of my tasks that aren't waiting on something or depending on something
task project:home-repairs export > home-repairs.json # export my entire home repair project to a json file to process into a gantt chart, budget (since I can add a price tag to each task) and whatever else
but then I can also have a script that looks like:
# taskfile.sh
infile="$1"
proj ="$2"
while IFS= read -r line; do
[ -z "$line" ] && continue
[[ "$line" =~ ^[[:space:]]*# ]] && continue
echo "Adding task: $line"
if [ -n "$proj" ]; then
task add "$line" "$proj"
else
task add "$line"
fi
done < "$input_file"
which lets me have preset task lists for various things, like:
taskfile.sh ~/tasks/packing-list.txt proj:work-trip
taskfile.sh ~/tasks/xmas-list.txt proj:xmas-presents
Thats an administration cli, not a daily use cli.
And, I mean things like being able to take tasks, pipe them through whatever tooling you like, pipe them back in, sync them from any other task system, etc. TaskWarrior has a fully customizable priority and urgency system, tagging, dependency management, and status system. I also like that TaskWarrior is usable completely offline and then can be resynced whenever you feel like it.
Vikunja is great and probably the right choice for most people. But if you do significant project planning for yourself AND want to be able to deal with you todo list programmatically, the only other thing close is orgmode.
I’ve taken a brief look at it. The ui is fairly slick, but you have to use it their way and it doesn’t have a cli. How much that matters is up to you.
Instead of new and exciting, how about old and boring? https://taskwarrior.org/
Sync server is self hosted, it’s more configurable and programmable than anything else and has a variety of front ends on different platforms.
And give up the best weather and access to 5 star restaurants in the world? Surely you jest!
Oh it’s gotten so weird and depressing for me too. Like a lot of people I played with the free demo of GH Copilot when it came out in 2021. I was… not impressed for the most part. Its python suggestions were usually not any better than the standard Python analyzers, though every once in a while it would pop up a surprisingly decent function. Its C and C++ were either hilariously terrible or more likely just non-existent unless I was writing a textbook algorithm for some reason. Then it might be okish, maybe.
That fall, the company I was working for laid off my entire department, so I spent a few months looking for a job and ended up in management for a few years. The only code I wrote was a few hundred lines of Python to help me manage Jira tickets and I didn’t bother setting up Copilot. I’m not sure if the company was paying for it. A couple engineers mentioned that they used it to do a couple of refactors, but nothing much seemed to happen with it.
Then, right about a year ago I changed jobs to go back to IC engineering at a company that had a strict no-AI code policy. In February I had a crazy work trip where I had three days to bring up some hardware that I had never touched before and script a demo for it. I pulled it off, after a couple of 3AM coding sessions in my hotel room. The C parts were straightforward, but the demo script was a damn mess of some of the worst Python I had written in a long time, due to the time pressure and the fact that I was writing it at 3AM.
Oh, but also in February the company softened its stance and was allowing AI coding assistance in some very limited ways as long as it fit into whatever our GH budget was. I knew that I would probably never touch the Python again as the real version was in process by someone else in Rust, but just in case I wanted it cleaned up. I figured it would take me a day or so after a good nights sleep to do it the right way. In the airport I asked copilot CLI to refactor my script, as a test, not expecting much.
Five minutes later it had refactored it into something much cleaner, properly using properties, and other such niceties. It wasn’t perfect, there was a sign error in one equation, and there was still some minor cleanup to do but damn. It was good enough that if I was the reviewer I would have thought it was a good first pass and only had a few nits.
Since then we’ve all been given a decent LLM budget and told to use it as much as we want. I’m not being told I have to use it, but almost everyone is except for a few of us grey beards. And reviewing the code is terrible, because there is too much of it and it doesn’t matter, I’m not actually teaching or working with anyone, just talking to someone who is talking to an LLM.
Due to some health issues that intermittently leaves me with too much brain-fog to really write code but clear enough to talk about my general knowledge and explain what I want to do, I’m tempted to use them to keep up. And when I do it’s so disheartening. They have the judgement of a very junior engineer, but with perfect syntax in every language and documentation level knowledge of almost every subject. But they’re just so damn fast at it. Previously, if I gave the kinds of instructions to a junior engineer, I’d expect to hear back from them in a day or three, with a passable first attempt and some interesting ideas to talk about. Now I hear back in five minutes with too much code and no reason to talk with them other than to give a few instructions to clean it up.
Eh, there are distros where you will never need to touch a command line if you don’t want to. In fact, most modern mainstream ones are that way: Mint, Fedora, Ubuntu, etc.
Though I don’t understand the fear of the command line. My grandmother owned two computers in her life and both of them were command line only. She was not a researcher at a university, a professional programmer or anything like that, just a normal person with a high school education.
She just wasn’t afraid to read the manual and try things. I still remember the spiral-bound manual with post-its covered in her nurses shorthand.
She used them to run the books for the family business, first an Apple ][, then a PC running DOS. She also used them to write letters and the church newsletter, practice Bridge, organize her family genealogy work, and things like that.
Welcome, brother. But remember that they must walk the path themselves or they will never reach true understanding.
Either a NAS, or a direct storage box plugged into your server. QNAP and Terramaster make reputable ones. Then you can setup your own storage pool across whatever drives you have.