From using AI to building it.
Three levels, one path. Start by using AI to get more done. Move up to building agents that automate real work. Take it as far as building the models themselves. Find the level you're at and start there.
Find your level.
Read the three levels below and identify where you are — that is your starting point. Each level is a complete outcome; go as far as you want.
Use
You want to get more done with AI today.
Learn to use AI well — both the AI you rent (ChatGPT, Claude) and the AI you run on your own machine.
No coding required.
architecture ▸ foundation · runtime + hardware + node
Automate (primary trunk)
You want to build agents that automate real work.
Build a working agent around a local model — the runtime, the tools, the loop, end to end.
For any software engineer. No deep ML math.
architecture ▸ the value layer · agent harness
› where most people are headed
see the agent trackBuild
You want to build the models themselves.
From inference through fine-tuning and training from scratch — and into running that stack yourself, in production.
The deepest path. Optional, and not for everyone.
architecture ▸ foundation · runtime (quantize/serve) + hardware
› the production track — optional, not required
Levels 2 and 3 are hands-on: every lab runs on the mpl toolchain — the local bench that turns your own hardware into the classroom.
mpl agent runmpl lab D1…D8mpl chatProvisions the harness lab, runs your agent loop locally, and traces every episode — the coach helping you is itself a local agent.
mpl benchmpl lab A1… / B1…mpl doctormpl burstInstalls and pins the local stack, benchmarks your rig, and scaffolds the inference / from-scratch labs — your hardware becomes the lab bench — and mpl burst sends the rare job too big for your rig to managed cloud.
Find your level and start
Use AI, build agents, or take it to the model layer. Connect your wallet, then enroll through your mentor to begin.