ZazenCodes Course
Coming soon
Jev: Introduction to System 1 AI
LLMs are too slow and expensive for reflexes. Learn to hand fast, typed decisions to Jev, a decision model, and save the LLM for real reasoning. Seven lessons, about an hour.
The lessons
Slides, runnable notebooks, and one live build.
System 1 vs System 2 AI
Why agent loops are slow, and Kahneman's two modes of thinking applied to AI.
What is Jev? TypeSafe & RLCD
Who built Jev, and how it is trained for calibrated decisions instead of text.
The three typed primitives
Noul, Choice, and Score in the Python SDK, plus a benchmark against frontier LLMs.
Four categories of use cases
AI-powered workflows, map-reducing big data, real-time applications, and verify everything.
Jev use cases in code
One worked example per category, on real data.
Custom Pi model router
A live build: a /jev command for the Pi coding agent that has Jev pick the model for each prompt.
Jev in the wild
Real projects built with Jev, and the course conclusion.
Before you ask
Where can I get it?
Here, for a one-time fee. It is also included in the Premium tier of the ZazenCodes Agentic Coding Club on Skool.
What do I need?
Python with uv and a TypeSafe API key. The comparisons use OpenAI and Anthropic keys, and Lesson 06 uses the Pi coding agent.