LOCAL MODELS
Start small.
Learn something real.
The Ollama lab makes local AI tangible: get a small model running, give it a useful task, and notice the limits before investing in more hardware.
Explore the Ollama lab05 / Build AI With Me
Ideas are easy.
Make one work.
Build AI With Me is Steven’s public workshop for practical AI: local models, connected tools, and systems worth understanding. The work turns an experiment into something you can follow, question, and build on.
HANDS-ON LABS / BUILD NOTES / PRACTICAL AIThe process is part of the work.
SELECTED WORK
Three builds.
Three useful questions.
Where should the model run? What should it be allowed to do? How do you know the result is useful? These projects give those questions something concrete to work with.
Start small.
CONNECTED TOOLS
A model that can act.
A boundary you can see.
The MCP lab follows a tool connection from its contract to failure cases and a human check. The interesting part is the handoff—and what keeps it under control.
Explore the MCP labBUILD IN PROGRESS
Your infrastructure.
Your decisions.
The private AI server project follows the choices behind running a model, adding an interface, and tracing the data path. The build is still being documented.
Follow the server build
A SAMPLE OF THE APPROACH
The first win
is understanding
what you built.
The introductory Ollama lab starts with a deliberately modest outcome: experience local inference before buying more hardware. Its four build notes make a useful rhythm—set up, try, observe, then decide.
That is the appeal of the project: the choices and limitations are part of the learning, alongside the result.
Read the labINSIDE THE LAB / FOUR BUILD NOTES
Open a step for a closer look.
01Install the runner
+
Begin with the component that makes a local model usable. Keep the first setup small enough to understand.
02Pull a small model
+
Choose a manageable starting point. A larger download is not the same thing as a more useful result.
03Ask a useful question
+
Give it a task you can judge. A fluent answer still needs to be checked against the purpose of the task.
04Observe the tradeoffs
+
Notice response time, answer quality, and hardware limits. Those observations guide the next change.
Step titles from the published Ollama lab. Expanded notes are portfolio commentary on the learning approach.
WHY THE HANDOFF MATTERS
Giving AI a tool
changes the question.
An answer is one thing. A system that can use a tool asks more of its builder: a clear contract, visible inputs and outputs, and a plan for failure.
The Claude MCP lab brings that boundary into view. It is a useful example of Build AI With Me’s focus on understanding a working system, including where human judgment belongs.
Model→Tool→Human check
Explore the tool connection 
THE THREAD THROUGH THE WORK
A result worth showing.
A process worth sharing.
Build AI With Me connects practical experiments with explanation. The portfolio spans local models, tools, and infrastructure; the value is in making the decisions visible enough for someone else to understand and extend the work.
Explore AI Builder FoundationsLET’S MAKE THE NEXT THING WORTH EXPLORING
See a connection?
Let’s talk.
Relevant sponsorships, useful products, thoughtful collaborations, and ideas that fit the work.
Partner around Build AI With Me