Tools

Tools

A list of tools I use.

Fun/media

I am aggressively de-googling my life in 2025/2026.

  • protonmail - encrypted email, Swiss privacy laws
  • duckduckgo - privacy-respecting search engine
  • immich: google photos clone, self-hosted
  • navidrome: youtube music clone, self-hosted
  • more to come? :)
  • vesktop - lightweight Discord client fork
  • switch.hacks.guide - comprehensive Nintendo Switch homebrew/CFW guide
    • nh-server/switch-guide - GitHub source for the above guide
    • mGBA - GBA emulator for Switch homebrew
    • pemu - multi-system emulator frontend (needs a custom XML file, annoying)
    • retroarch - multi-system emulator frontend (needs older firmware, warning)
  • imgburn, cd burning
  • canirun.ai - can I run this game checker
  • Xbox Emulator Files - Xbox emulator files collection

Programming

  • python: general purpose language
    • conda/poetry/uv/pipenv/pip - package managers. prefer uv for speed/simplicity.
    • pil/pillow: image processing library
    • numpy: numerical computing library
    • pandas: xlsx/csv manipulation library
    • matplotlib: data visualization library
  • rust: faster than python.
  • bash: scripting language
  • just: very good build tool
  • postman: http requests
    • bruno is probably better, TODO use it
    • in general, ditch postman. they're starting to put their client behind a paywall/login wall. move to an OSS tool that's free. it's not rocket science to make something that ingests OpenAPI Spec and just runs curl...so why force corporations or individuals to pay for the simple feature of saving a Postman collection?

Text Editing

  • windsurf: bloated, use zed

  • cursor: bloated, use zed

  • vscode - bloated, only use if you absolutely need to. It's slow and used to be my favorite, but...

  • zed - Fast and simple text editor. Rust backend.

  • nano - command line editor

CICD

  • jenkins
    • AVOID. legacy and lots of design bugs
    • pipeline is useful but...its 2025. try to use github cicd or similar. AVOID jenkins.
  • get good at bash
  • understand yaml
  • github runners
    • (literally just docker containers w/ some yaml command runner. its all the same pattern.)
    • can also do self hosted
  • openstatus - uptime/health monitoring, can self host

OS

i highly recommend never using windows, for many, many reasons

Linux distros

  • ubuntu - server OS
  • nixos - server/desktop OS
  • cachyOS - easy arch based OS, fast. bleeding edge so stuff breaks. avoid if u value ur time or are new to linux.
  • TrueNAS CORE - freebsd based NAS OS that uses ZFS, ZFS is superior for use as a NAS for many reasons
    • AVOID hardware based RAID. if ur card dies and u cant replace it, kiss your data goodbye. software based raid does not suffer from this issue.

Dotfiles

  • chezmoi - dotfile manager, handles templating/secrets/multi-machine setups

Windows utilities

Hardware

I run various cheap laptops and a few HEPC (High-End Personal Computer) setups. Only Linux, except for a Windows computer I use for piano playing/DAW/Frooty Loops.

AI

My current AI stack is here. I used to use ollama but it has issues with ROCm drivers on AMD GPUs, so I switched to llama.cpp!

cachyos-whitedragon-ai-lab

Current working stack:

Corpo AI for coding

  • claude: You can use this, but you can also self-host local AI and use CLI tools like:
    • aider: OSS claude cli clone
    • etc...TODO: Add more coding agents here from testing.

The basic pattern to locally host is just to run llama.cpp (or ollama if you have NVIDIA GPUs) on a powerful PC, set up a VPN with tailscale, and then connect to the LLM endpoint with your less powerful computer via a coding agent like aider or a frontend like openwebui.

Local AI models

These models were intended for ollama but can be converted for llama.cpp. I plan on running my own benchmarks on each.

  • llama3.2:3b
    • TODO: Test.
  • yuiseki/devstral-small-2507:24b
    • TODO: Test.
  • hf.co/bartowski/Qwen2.5-Coder-14B-Instruct-abliterated-GGUF:Q4_K_S
    • TODO: Test.
  • hf.co/mlabonne/gemma-3-27b-it-abliterated-GGUF:Q4_K_M
    • Human remarks: It is indeed abliterated. It will happily generate usable Python malware.
    • TODO: Test.

Anti AI/Anti Slop

  • nepenthes - Markov chain babble generator, wastes the time/CPU of scrapers

Infra

  • docker: containerization platform
  • kubernetes: container orchestration
  • grafana: monitoring visualization
  • prometheus: monitoring metrics/backend

Backup

  • dd - data duplicator/data destroyer

Media

Streaming

Other

  • handy.computer - offline text transcription
  • sshx.io - share terminal (dangerous)
  • https://pomofocus.io/ - pomodoro timer
  • obsidian - markdown note-taking / personal knowledge base

Hacking links

Dark web / anonymity

Hacking tools

Used

  • kali linux - penetration testing distro by Offensive Security

  • parrotos - security/privacy-focused Linux distro

  • dirb - web content scanner / directory brute-forcer

  • nmap - network mapper and port scanner

  • rustscan - fast Rust-based port scanner, feeds into nmap

  • ftp - file transfer protocol CLI client

  • some python idk lol

  • searchsploit - offline CLI search for exploit-db

  • msfconsole - Metasploit Framework interactive console

  • msfvenom - Metasploit payload and shellcode generator

  • evilginx2 - adversary-in-the-middle phishing framework, bypasses 2FA by proxying real login pages and capturing session cookies

To investigate