---
title: 'Ai Things'
url: 'https://grav.munn.me/articles-books/ai-things'
markdown: 'https://grav.munn.me/articles-books/ai-things.md'
date: '2025-11-09'
description: 'Well, everyone seems to get excited about the ai thing these days, so I thought I would create an ai links page. Kind of like my old ones. I will start putting pro and con articles and models here. Personally I have found ai useful for: translation, removing noise from microphones, occasional scrip…'
---

# Ai Things

Well, everyone seems to get excited about the ai thing these days, so I thought I would create an ai links page. Kind of like my old ones. I will start putting pro and con articles and models here. Personally I have found ai useful for: translation, removing noise from microphones, occasional scripting help, poetry, ocr, and thats about it.

| Link | Description | Rating |
|---|---|---|
| [lmstudio](https://lmstudio.ai/) | EASY way to get into ai | 5 |
| [granite model](https://huggingface.co/collections/ibm-granite/granite-code-models-6624c5cec322e4c148c8b330) | IBM provided model which allows for commercial use. | ? |
| [UNSAFE llama3 model](https://huggingface.co/meta-llama/Meta-Llama-3-8B) | Its face book's somewhat unsafe (It told me how to make sarin gas!) | 3 |
| [somewhat safer mistral model](https://huggingface.co/mistralai/Codestral-22B-v0.1) | Seems to not let you make sarin gas. Kind of moralistic. | 4 |
| [web archives with ai article](https://lil.law.harvard.edu/blog/2024/02/12/warc-gpt-an-open-source-tool-for-exploring-web-archives-with-ai/) | VERY interesting article on using web archives and training models on specific datasets. | 5 |
| [Granite Model](https://huggingface.co/collections/ibm-granite/granite-code-models-6624c5cec322e4c148c8b330) | IBM's granite model, suitable for commercial use | ? |
| [red flag](https://github.com/Addepar/RedFlag?utm_source=cloudseclist.com&utm_medium=referral&utm_campaign=CloudSecList-issue-243) | evaluates the safety of software using ai models locally | 4 |
| [fuzzy dedupe paper](https://blog.nelhage.com/post/fuzzy-dedup/) | uses ai to dedupe documents | 4 |
| [ai models collapse](https://www.nature.com/articles/s41586-024-07566-y) | ai models collapse if they use non-human generated base data | 5 |
| [llm2sh](https://github.com/randombk/llm2sh) | converts plain spoken things into shell commands | 4 |
| [microagent](https://github.com/BuilderIO/micro-agent) | automates tests based upon criterion | 4 |
| [change](https://www.theregister.com/2024/08/18/self_hosted_github_copilot/) | lets you autmatically do copilot like chatting locally with ollama and various local models | 5 |
| [run llm locally](https://nullprogram.com/blog/2024/11/10/) | everything that this person has learned about running local llms | 5 |
| [project mantis](https://github.com/pasquini-dario/project_mantis) | project mantis hack back at the ai hacker | 4 |
| [llm from scratch](https://andrewkchan.dev/posts/yalm.html) | helps you train and do an llm from scratch with cuda and c++ | ? |
| [LLM locally](https://abishekmuthian.com/how-i-run-llms-locally/) | A bit expensive, but interesting how to do llm locally | 4 |
| [tuskira](https://www.tuskira.ai/) | an automated threat feed that uses ai | 4 |
| [modernbert](https://huggingface.co/blog/modernbert) | a encoder only model very fast | 5 |
| [UNSAFE llama model](https://huggingface.co/huihui-ai/Llama-3.3-70B-Instruct-abliterated) | Basicically a completely polly-anna removed LLM | 5 |
| [chatgpt Temporal model break](https://www.bleepingcomputer.com/news/security/time-bandit-chatgpt-jailbreak-bypasses-safeguards-on-sensitive-topics/) | Basically a way that they broke chatgpt by using fake time travel | 5 |
| [deepseek original pdf](https://github.com/deepseek-ai/DeepSeek-V2) | Has the original pdf with deepseek's models etc. | 4 |
| [ramalama](https://github.com/containers/ramalama) | models in containers to make them safer | 5 |
| [exo](https://www.cnx-software.com/2025/02/18/exo-software-a-distributed-llm-solution-running-on-a-cluster-of-computers-smartphones-or-sbcs/) | lets you run models on distributed computes with lots of gpu's slow but lets you run bigger models more cheaply | 4 |
| [voice cloner register article](https://www.theregister.com/2025/02/16/ai_voice_clone/) | nice article on how to clone voice, needs nvidia alas with a bit of vram (16gb or so1) | 3 |
| mcp registry | article on ai mcp regestries |  |

---

## Navigation

- Parent: [Articles and Books](https://grav.munn.me/articles-books.md)
- Previous: [Security](https://grav.munn.me/articles-books/security.md)
- Next: [random kernel patch to make linux faster for intel](https://grav.munn.me/articles-books/random-kernel-patch-to-make-linux-faster-for-intel.md)