Most of us use AI through products like ChatGPT, Claude, or Gemini. That means someone else is making a lot of decisions for us. They choose the model, decide how it behaves, what it remembers, what it can access, and when it changes.
This week, we’re going to get a little closer to the actual model to get a better understanding of what can be customized and give us more control.
Before we start: if you are getting this email forwarded to you and are not yet part of the lab - you can join the tiny AI lab here.
The Challenge:
Run an open model locally and build an AI model locally on your computer. You are not training a model from scratch.
Instead, you can use an existing open model, run it directly on your computer, and customize it for one specific job.
The easiest way to do this is with a tool like Ollama or LM Studio. You can use your usual LLM to help you through the setup, tell you which model makes sense for your computer, and give you step-by-step instructions as you go.
The prompt to get you started
You can still use one of the frontier models to guide you through the set up and instructions.
Drop this into your AI of choice to get started:
I want to run an open model locally on my computer and use it to build a simple AI assistant.I am not very technical, so walk me through this step by step and do not assume I already have developer tools installed.
First, help me choose the best open source tool based on my computer and comfort level.
Then:
Help me check my computer specs and recommend a small open model that should run well.
Give me exact installation instructions.
Help me download and run the model.
Help me create a simple assistant for this job: [DESCRIBE THE JOB].
Show me how to customize its instructions and, if possible, what information it can access.
Give me five test prompts I can use to evaluate whether it is actually good at the job.
Only give me one step at a time. Wait for me to tell you whether each step worked before moving on. If I get an error, help me troubleshoot it before continuing.
Too easy?
Swap the model.
Keep the job, instructions, and test prompts exactly the same, but run them against two or three different local models.
Why this matters?
The way most people still interface with AI models is through products like ChatGPT, Claude, CoPilot etc. But the instructions, memory etc can all be changed independently and local models are a great way for you to have more customized or more private use cases for AI.
Running an open model locally makes those pieces much easier to see so you can decide what works best for the work you are doing.
That’s it for this week.
Hit reply and tell me what you thought of the challenge this week.
I read every response, and your experience helps me shape these challenges. If you get stuck or have questions, reach me at [email protected]