A lab, not a lecture hall

Compilea spark of curiosityinto shipped systems

A new way to learn to code, where you set the depth on every step. Move fast with an AI handling execution while you learn the ideas, or go deep and build it yourself with just enough guidance. From your first line to multi-agent AI systems, taught through interactive visuals, not walls of text.

Enter the lab Preview a build

Workshops

Guided paths

100%

Runnable code

Free starters

agent.pyrunning

from langgraph.graph import StateGraph

g = StateGraph(AgentState)

g.add_node("llm", call_model)

g.add_node("tools", tool_node)

g.add_conditional_edges("llm", route)

agent = g.compile(checkpointer=mem)

# → agent ready. streaming…

// agent execution graph

LLM

TOOL

MEM

OUT

// One step, three ways

Struggle, follow, or direct an AI.
Every path lands on the same code.

Each code scene authors three parallel routes to the same result - so you set the depth. Grind it out for recall, type along for a demo, or write a prompt and audit the AI's output. One verify command checks all three.

All paths converge → one verify

$ node -e "require('./auth/passwords')" && echo OK

A wrong answer on any path fails the same import - fail-loud, no branching logic.

auth/passwords.jsHANDS-ON

Your task - build it from the spec

Create auth/passwords.js: • Import bcrypt from bcryptjs • Export hashPassword(password) - hash with 12 salt rounds, return the hash • Export verifyPassword(password, hash) - compare, return a boolean • Both functions must be async

Hint 1Conceptual nudge

bcryptjs gives you bcrypt.hash(password, saltRounds) and bcrypt.compare(password, hash) - both return Promises.

Hint 2Partial snippet

const bcrypt = require('bcryptjs'); async function hashPassword(password) { return bcrypt.hash(password, 12); }

Hint 3Near-complete logic

bcrypt.compare returns a boolean Promise - true if the password matches the hash. Return its result directly from verifyPassword and remember module.exports.

// The learning graph