inside the rabbit hole
all courses
Every open course in The Rabbit Hole classroom. They are free to every member, and most are a handful of short text lessons.
Start here, no developer needed
Never opened a terminal? One boring task off your plate, on the phone you already have.
open on skool →Start Here: Welcome to The Rabbit Hole
How the community works: check wild claims, find the catch, match projects to your hardware, and run one small experiment tonight.
open on skool →AI Engineering: The Full Course
Every module of the AI engineering course in one place, from LLM foundations to shipping secure, observable systems.
open on skool →AI Foundations
What an AI model actually is, what it costs you, and how to pick one, ending with your own setup working.
open on skool →The Coding On-Ramp
The terminal, git, a real editor, enough Python and TypeScript, and your first thing live on the internet.
open on skool →Run 'Impossible' AI on Ordinary Hardware
How RAM, GPU memory, storage and architecture decide what runs, worked through three real cases from the community.
open on skool →Local AI Without a GPU
What CPU inference can do, why GGUF and quantization matter, and how to choose a first model that won't turn your laptop into a space heater.
open on skool →Fine-Tuning on Low-VRAM Hardware
Why fine-tuning and inference are different memory problems, and how adapters and honest benchmarks let small GPUs do real training work.
open on skool →How to Verify Viral AI Claims
Find the primary source, separate the memory numbers, reproduce the smallest test, and report the catch.
open on skool →Build an AI Research System
Turn weekly link hunting into a bounded job: define the question, collect, filter, keep the citations, and deliver a brief.
open on skool →Useful AI Agents for Boring Work
Automate one chore you already repeat, set its permissions, handle failure, and decide after seven days whether it earned its keep.
open on skool →Build Your First Local AI App
A blank folder to a small local app you can run, inspect and improve, calling a model through Ollama's local API.
open on skool →Chat With Your Files Using RAG
A private question answering system for your own documents, with citations, and tests for where RAG confidently gets it wrong.
open on skool →AI Coding Agents for Real Projects
Use coding agents without handing them your repo and hoping: a tight brief, limited context, a testable loop and a diff you review.
open on skool →Use MiMo-V2.6-Pro without a GPU
A 1.02 trillion parameter open model, used without owning a GPU. Six lessons, from your first call and the price list to Command Code, your own key and a five task check on your own work.
open on skool →Before you rent a GPU for MiMo-V2.6-Pro
What it takes to serve the 1.02 trillion parameter model yourself: why the weights need 566 GB of GPU memory, and what an 8 GPU box costs per hour on Modal, RunPod and AWS.
open on skool →