jeena
jeena on Hacker News
As long the harness is not open source a proxy won't help, it's easy to phone home on a different subdomain also.
Did you look into self-hosted open source decision models because they are quite capable also much faster and much easier to integrate and you don't need to pay anything.
Posted by
jeena on Tuesday October 06, 2026 @08:32PM
from the jeena.net dept.
Posted by
jeena on Sunday October 04, 2026 @01:36AM
from the self-post dept.
About 13 years ago I made https://jeena.net/catdog using machine learning algorithms to classify pictures of faces of dogs and cats and decide if we’re looking at a cat or a dog. I was in university and this was what I came up with doing for one of the assignments in the ML course.Another guy came up with a evolutionary way to find math formulas which would over time be able to get closer and...
I'll try it. It's cool that it's small and optimized to be a assistant, not sure it will work well with pi as a harness or if it really is only optimized with his harness, I'll find out, I only have a RTX3060 with 12 GB VRAM and would love to use it for assistant stuff.
One is pi.dev on the console, but because the KB is a git repository I have it on each computer. On top of that I also run Hermes Agent and have set up there one agent to be responsible for the KB stuff and that one I connected to a Matrix channel and can use it from the phone and any device where I have a matrix client like Element set up.
With pi.dev I use it in a container https://git.jeena.net/jeena/agent-container and selectively give it access to Firefox via https://github.com/ThinkerYzu/foxmcp when I want it to help me with something in the browser (doing taxes, applying for things online in a foreign language, etc.)
I'm doing something similar but not for the whole family, just for myself, both for home stuff and for work stuff (seperated), but I just give pi access to the tools and accounts I want it to use. I don't use google for anything other than watching YouTube so it would not work for me anyway. I have a knowledge base git repo which the agents use together to do things. The PDF filling thing is especially cool with a KB in the background because the Agent can fill in 99% of the things by itself. It also organizes my todo lists and my calendar.
The owners usually, which means the capitalists, not the working class.
Ah haha that is kind of funny it finds exactly what I was looking for, just much slower than 15 years ago.
I was reading and reading and thinking that's the ages old advise POSSE, and then they even mention the Indieweb, nice.
I tried to do that for very long translations, I had a sliding window, I had a memory for the important things to keep it consistent, a loop for repairs etc.
https://jeena.net/loop-engineeringBut for some reason the local models I used back then that was almost 2 years ago) weren't good enough so none of my optimizations did anything good for the translation quality.
I always liked their voice dictation because it's better than Samsungs or Googles while it's on device. It's super fast and puts in punctuation in, etc.
The keyboard on the other hand I never really liked for some reason which I can't even explain clearly. So for the keyboard I've been using https://github.com/HeliBorg/HeliBoard for ever, but it does not have sviping.
I might try FUTO Swipe just to test it.
Posted by
jeena on Thursday June 18, 2026 @11:00PM
from the jeena.net dept.
I like that a project only qualifies if it's hosted on GitHub.
I made myself a very simple one from the start when I realized it can access everything on my computer
https://git.jeena.net/jeena/agent-container my goal was that it would work transparently and the paths and user, etc. would be just the same as on the host but inside of a docker container.
Off topic, but when I opened the website I was instantly teleported back to a better website design time. This site has character, I will recognize it, it's not like what I was complaining about in
https://jeena.net/content-is-king
We switched from Bit bucket to Gerrit internally and it was a steep learning curve for the des but it's fine.
At a customer we're implementing GitHub Actions and even on our Dev environment there are so many hickups with GitHub.
I've been using OpenCode until yesterday (with some plugin to let me use their model until they implemented what it seems very sophisticated detection to reject you).
It just has a sane workflow it's easy to use, doesn't bother you with 1000 questions if you allow this or that to run and generally it feels like the model is dumber and makes more mistakes since yesterday since I have to use claude code.
About 10 years ago i had HomeAssistant running and thacking my bluetooth devices. It does so per default by jus memorizing a mac adress an recording when it's visible and when not. No need for pairing or anythung. It also stores the custom name if available.
Anyway, the default dashboard also automatically generated a view when my neighbours "Katie's iPhone' was at home and when not, until I actively deleted it and the data it stored.
I use Tubular because of sponsorblock.
My Main Problem is To keep the bridges up to date. I just switched my phone number in WhatsApp and Signal and that lead to a huge ton of trouble for my bridges. After a month of fiddling with it, deleting things, updating, logging in and out of accounts and puppets, I still don't get any messages from signal into element. While it was working for years just fine it gives me the most trouble now that my dad is in the hospital in another country and I have to coordinate with my siblings a lot.
I never thought that the writing HTML is the hard part. Instead where my Sister struggled was to get a domain, get a server, deal with DNS pointing to that server. That shit is also omitted on posts like that one because it's done differently for each domain registrar and you need quite some knowledge in network stuff to be able to understand what you need to do.
Posted by
jeena on Tuesday January 13, 2026 @06:35PM
from the jeena.net dept.
OP here
I was afraid of trying any online models with opencode because per default it has full access to your computer. I also never want the AI to push code anywhere, I want to do it myself to keep this control, afterall it's easy just to nuke the local repo.
This is where I came up with this docker container for it and a alias which opens opencode in the docker container and mounts things, etc. I'm just afraid I missed some safety things, because I'm not good at it.
Posted by
jeena on Tuesday January 13, 2026 @12:51AM
from the git.jeena.net dept.
I self host a lot of stuff myself:
https://uptime.jeena.net/status/everythingAnd until now without AI, but I'm kind of curious but afraid that it will bring my servers down and then I can't roll back :D But perhaps if I would move over to NixOS, then it would be easy to roll back.
I'm not sure if it's abandon ware but I found a gist which attempted to get pictures from this Korean Kindergarten app for backup which we also use.
I tried it and it only half worked and you had to copy a cookie out of the browser which would expire after a couple of days.
I took it and extended it added a login with selenium or what it is called and now I can run it daily and get those pictures out and into my immich. Instance.
The script was already a couple of years old and outside of Korea nobody uses KidsNote, I made it used ware again: https://git.jeena.net/jeena/kidsnote-backup