My AI Stack, Usage and Resources
Last Updated on 5/1/26
The world of AI moves fast. I monitor advancements and test many tools. Here’s which tools I currently use and the various methods I use them for.
My Posts on AI
These are my posts on new tools, guides, and interesting news.
My AI Stack
My AI General Usage
- My quick go to for answering questions and quick research is using Gemini. I have a Pro subscription and use it for quick chats and using the chrome extension directly on pages in the browser. I find the personal intelligence feature by connecting to my Google suite for Gmail, docs and more is useful.
- I will also sometimes use ChatGPT, Claude, Perplexity and Grok to test their answers and compare.
- The biggest challenge is wanting to maintain my chat history as a reference that I can search against which is why I primarily use Gemini. I can be difficult if your discussions are dispersed amongst various systems. I’m sure there are ways around that by aggregating all discussions but I haven’t pursued that yet.
My AI Use Cases
- Research is my most common use case
- Shift from search engine to answer engine using Google. The AI summary gets it right 80% of the time
- Looking for products to serve a specific need and adding a feature comparison matrix
- Travel research
- Excel and Google Sheets
- This has been one of the greatest unlocks. AI has enabled me to leverage all the coding power of spreadsheets I was unable to utilize before by creating formulas, macros, power queries and vba scripts.
- For work I created a reporting tool that pulled content records from a JSON feed and then transformed the data with custom rules into a spreadsheet. Another script generates static reports from the data into new files
- I’ve created a spreadsheet that lists all the Hollywood Bowl concerts and allows multiple friends to show interest in different shows and then add shows they go to with seat locations which helps everyone in our group co-ordinate events.
- Redesigned and optimized my WordPress Sites
- I redesigned and optimized my websites using several tools to reduce plugins and minimize theme dependencies. Full article with details available here.
- Install Immich using a Docker instance
- This isn’t a simple task. It takes quite a bit of technical effort and was even more difficult since I’m using the external library feature. I’ve used both Claude for the initial installation as well as Codex to do a version upgrade which is also challenging and was able to complete these tasks easily with AI.
My AI Tools and Services
- Kling has been my favorite tool for video generation
- Suno is my favorite for music creation
- PhotaLabs is great for creating personal avatar images
- Huggingface (my likes)
- Replicate (my likes)
- Fal
My AI Coding
- I’ve recently started using Codex Desktop with a Pro subscription to compare to Claude Desktop
- I heavily used Claude CLI and Claude Code Desktop with Opus 4.7 for planning and Sonnet 4.6 for execution
- I use the /superpowers skill for planning and executing projects
- I use the /frontend-design skill to help with design
- I’ve also recently been testing out Claude Desktop as well now that they support skills and plugins
- I’ve been hosting projects using Railway and Vercel
- Most of my apps are shared on Github
Agent
I was testing out the OpenClaw agent but was having some issues especially when upgrading it. I then tested out Hermes and had a much smoother experience and have since switched.I found it easier to setup, configure and maintain. I’m still in the early phases of running this but it seems like a very promising tool I will be using daily. Below is my current configuration.
- My installation is on a Windows machine using WSL
- Model: Using a Github Copilot Pro subscription setup with the Claude Sonnet 4.6 model as my default. I do this to keep my Claude subscription dedicated for coding and isolated from agent usage.
- I installed the Web UI to provide a browser interface to track things easier
- I’ve activated the built in llm-wiki skill to populate a dedicated Obsidian vault for the agent. This X post explains the value well.
- I can communicate remotely with the agent through Telegram and conversations are broken up using the Topics feature
- I use LMStudio if I want to switch to using a local model. Some details on why you may want to do this.
Keeping up with AI News and Announcements
The world of AI moves very very fast. Major advancements are announced on a weekly basis. I have several places where I monitor this information to stay on top of it listed here. For my latest findings you can follow me on X and if you want a real-time firehose of thousands of resources check out Robert Scoble’s AI Lists.
- Podcast and Newsletter: AI for Humans
- Youtube: Matt Wolfe
- Reddit: artificial, ChatGPT, ClaudeAI, ClaudeCode, vibecoding, Codex
- Newsletters: Ben’s Bites, The Rundown, The Algorithm
Running AI Tools on My Computer
You can run AI tools for free on your own computer if you have a powerful system with an Nvidia GPU or Apple M series processor. Installation is fairly easy and one reason I do this is when I require privacy by analyzing personal documents. These tools are a great way to learn and experiment.
- I use Pinokio to run many different AI tools
- Here’s an install tutorial video. Support is available on the Discord support channel.
- I use LMStudio to run different models locally which can then be accessed by other tools when necessary. For example I can tell my Hermes agent to use models running through LMStudio. This is primarily done for privacy and cost savings since there is no charge for me to run them.
Services to Host and Manage my Projects
- Vercel – Hosts most of my projects and one is hosted on Railway
- Resend – to manage sending emails
- Umami – for tracking traffic to the sites
- Neon – for hosting databases
- Cloudflare – for DNS and asset storage


