SmokeyDope

I don’t care what PR rep says, its cost cutting one way or another. I hypothesize that gorilla glass had a 10 year contract with the phone manufactures and now that the contract is up some are choosing to cut out the moddle man in hopes of cheaper in-house solutions.

Would something like this interest you? Gemtext formatted to html is about as light weight as it gets. lots of automatic gemtext blog software on github that also formats and mirrors an html copy. Whenever a news page article gets rendered to gemtext through newswaffle it shrinks about 95-99% of the page size while keeping text intact. Let me know if you want some more information on gemini stuff.

I just spent a good few hours optimizing my LLM rig. Disabling the graphical interface to squeeze 150mb of vram from xorg, setting programs cpu niceness to highest priority, tweaking settings to find memory limits.

I was able to increase the token speed by half a second while doubling context size. I don’t have the budget for any big vram upgrade so I’m trying to make the most of what ive got.

I have two desktop computers. One has better ram+CPU+overclocking but worse GPU. The other has better GPU but worse ram, CPU, no overclocking. I’m contemplating whether its worth swapping GPUs to really make the most of available hardware. Its bee years since I took apart a PC and I’m scared of doing somthing wrong and damaging everything. I dunno if its worth the time, effort, and risk for the squeeze.

Otherwise I’m loving my self hosting llm hobby. Ive been very into l learning computers and ML for the past year. Crazy advancements, exciting stuff.

Sounds like ollama was loaded up with an either overly censored or plain brain dead language model. Do you know which model it was? Maybe try mistral if it fits in your computer.

I run kobold.cpp which is a cutting edge local model engine, on my local gaming rig turned server. I like to play around with the latest models to see how they improve/change over time. The current chain of thought thinking models like deepseek r1 distills and qwen qwq are fun to poke at with advanced open ended STEM questions.

STEM questions like “What does Gödel’s incompleteness theorem imply about scientific theories of everything?” Or “Could the speed of light be more accurately refered to as ‘the speed of causality’?”

As for actual daily use, I prefer using mistral small 24b and treating it like a local search engine with the legitimacy of wikipedia. Its a starting point to ask questions about general things I don’t know about or want advice on, then do further research through more legitimate sources.

Its important to not take the LLM too seriously as theres always a small statistical chance it hallucinates some bullshit but most of the time its fairly accurate and is a pretty good jumping off point for further research.

Lets say I want an overview of how can I repair small holes forming in concrete, or general ideas on how to invest financially, how to change fluids in a car, how much fat and protein is in an egg, ect.

If the LLM says a word or related concept I don’t recognize I grill it for clarifying info and follow it through the infinite branching garden of related information.

I’ve used an LLM to help me go through old declassified documents and speculate on internal gov terminalogy I was unfamiliar with.

I’ve used a speech to text model and get it to speek just for fun. Ive used multimodal model and get it to see/scan documents for info.

Ive used websearch to get the model to retrieve information it didn’t know off a ddg search, again mostly for fun.

Feel free to ask me anything, I’m glad to help get newbies started.

1. Download server files and run them. Modern Minecraft requires java runtime enviroment 17, 1.16 and belore jre8. If you have synaptic package manager search for jre there.

2. I usually make a batch file to execute server. You can run jre as is by executing as an application but a batch file helps set memory size usage, nogui and stuff.

3. It will stop on first server startup to produce a EULA text file. Open it up, set to true, save, restart server. See if you can connect through LAN by using another computer and entering your servers local IP address probably 10.0.0.xx:25565

4. Open up the port for default mc server in your router so others can connect on public network. 25565, TCP+UDP, in out both ways. Make exceptions in your firewall too.

I wrote a guide on here about the differences between alternative search engines. I recommend for you either YaCy or marginalia.nu. searxng supports calling YaCy (I actually contributed to that feature on the github).

The problem with decentralized engines like marginalia and YaCy is that they aren’t good at the things a average user wants from a typical search engine. Ideally a search engine is meant to quickly provide you links to webpages which are strongly related in content to you are looking for. Shopping, weather, map directions, local business hours. On some level you need to prioritize showing the user what they want ideally within the first few results.

Decentralized engines by their nature don’t do this easily. Instead using YaCy or marginalia feels like a scavenger hunt where you get handed a page of random websites loosely connected by your keyword search term and are told to start looking. This is good when your in the mood for blogspam and personal webpage blog dicovery, but not great for finding local buisniess info quickly. YaCy has a user curated priority system but not enough user mass adoption to be worth a damn in practice.

So sadly if you want anything resembling google or bing results for your practical convinence driven daily internet searching needs, you need to scrape them with searxng or use one of their few real search engine company competitors funding their own indexers and web crawlers. So really your options are scraping google, bing, mojeek, qwant, kagi and DuckDuckGo(ish they still use bing for indexing a lot).

Out of those Ive actually warmed up to Kagi over the year. I was put off at the idea of subscription based internet search but its a really good service they provide and they line out their reasoning for pricing well. They seem to be using that monthly sub money to actually improve the service and user experiences while remaining transparent with constant changelogs and blog updates. Kagis recent implimentation of privacy pass protocol, available TOR access, anonymous payment options, and taking fediverse + small net indexing seriously are all green flags to me. Never thought I would pay for a search engine but the way the world is going I’d rather eat the equivalent of a 5-10$ patreon sub to grow a service I believe is respecting me as a customer over fucking FAANG treating me like cattle and absolutely violating user experience so hard just for another nickles worth value in data scraping.

I run a local LLM on my gaming computer thats like a decade old now with an old 1070ti 8GB VRAM card. It does a good job running mistral small 22B at 3t/s which I think is pretty good. But any tech enthusiast into LLMs look at those numbers and probably wonder how I can stand such a slow token speed. I look at their multi card data center racks with 5x 4090s and wonder how the hell they can afford it.

If all computers are on local network you can use warpinator.

Koreader has a plugin to sync with calibre local server and its a REALLY good ereader software

You can self host a local chatgpt like ai known as a local large language model. Searx and Searxbg are great customizable meta search engines that you can customize to scrape whatever you want