News for hackers. Stuff that matters.

sebion7125

I've been working on pi-protokoll, an extension for the Pi coding agent. When I ran long sessions on a local model, compaction kept stalling for minutes because the model had to dismiss its cache and read the whole conversation as well as its compacted version. So now it writes a short protocol at the end of each turn, and compaction mostly just assembles those. Started as a quick fix, but by now it's gotten pretty polished: https://github.com/sebion7125/pi-protokoll
Re:Ask HN: How do AI agents help you in your daily life? on Friday November 06, 2026 @10:25AM (#50393478)
I started with Debian in the VM and now Ubuntu is the distribution I boot into.
Does it depent on the country you're located?
Re:Ask HN: Local AI Models for Coding and General Tasks? on Thursday November 05, 2026 @06:42PM (#50386193)
what is the ammount od system RAM you have? If you've got enough Strata with Qwen3.8 Flash Next could be a great option. For me it works great on Pi. I've developed an extension, that made it useable for very long chats and it now helps developing on huge Repos like GrandOrgue. Sometimes it beat Sonnet 5.5 in this configuration, but the most efficient way is to combine it with Opus 5.5, when it gets stuck, or its plans are not optimal. After having the Max Subscription for 2 month I'm now able to keep the Pro subscription and still get good results.
Re:Ask HN: How do AI agents help you in your daily life? on Thursday November 05, 2026 @06:34PM (#50386130)
It's the new better search engine for me. It helpf me avoid click bait videos on YT, as it can summarize and fakt check them with a click. I use it to organize all my paper work as far as legally possible ;) And it was the game changer, that made Linux the perfect OS for me. If there's anything not working or not working as I would like it to be, it helps me fix and or modify anything I want. Of course it made programming with foreign code bases a lot easier, such that I now am an open source developer in my free time, trying to make the world a little better step by step. And it is really a great tool if you want to learn new things. No need to mention you have to stay skeptical with the answers you get...
I'm wondering if that could be part of a prompt injection warn system in the future. Did I get it right, that it does not need an LLM to evaluate sessions?
This is awesome. Thanks for sharing weights on HF :-)