Wovi
Wovi worlds
WORLD 02AGENT MAZEPLAYABLE PROTOTYPE
No account · 5-minute mission

AI AGENTS + PYTHON · AGES 9–13

Give an agent a goal.
Then fix what it gets wrong.

A storm is blocking the fast route to Sky Beacon. Tune Wovi's reward rules, watch every choice and reveal the Python behind the visual blocks.

SKY GRID 06 · LIVE
RUN 01
DECISION LENSAwaiting a policy run

Candidate move scores will appear here before Wovi acts.

—Progress—Energy—Safety—Memory
STEPS0
ENERGY0
WORLD STATEReady to deploy
MISSION 02 · REWARD LABMake a prediction

What should Wovi value?

First predict what the starter policy will do. Then run it, inspect the evidence and change one rule.

Before you run: which route will Wovi choose?

Choose a prediction to unlock the first run.

GOAL
Move closer to the beaconReward each step of progress
+4
ENERGY
Collect a charge cellA tempting shortcut reward
+2
SAFETY Storm penalty
MEMORY Unvisited tile reward
PYTHON VIEWUpdates with your blocks
GOAL_REWARD = 4
ENERGY_REWARD = 2
STORM_PENALTY = 2
MEMORY_REWARD = 0

def score(move):
    value = GOAL_REWARD * move.progress
    if move.has_energy:
        value += ENERGY_REWARD
    if move.is_storm:
        value -= STORM_PENALTY
    if not move.visited:
        value += MEMORY_REWARD
    return value
Mission brief

The fastest-looking route contains both energy and a storm. Predict what the scores will make Wovi do.

LEARNING DESIGN

Agent Maze exposes a small reward-scoring algorithm rather than pretending the agent understands the world. The first policy is intentionally flawed: useful failure creates a reason to inspect, debug and rerun.

What this five-minute mission teaches

From a game decision to an AI mental model

Every visible control maps to a value in the Python view, so the learner can explain why the route changed.

Goals are designed

An agent optimizes the score it is given—not the intention a person forgot to express.

Rewards can conflict

Fast progress and extra energy may compete with safety. Weights decide the tradeoff.

Behavior is evidence

A failed route is a clue about the policy, not proof that the agent is careless or alive.

Code makes it precise

Visual blocks become Python variables, conditions and a scoring function.

NEXT SKILL

Want to understand the Python?

See how Wovi bridges visual rules, functions and simple objects for beginner programmers.

Explore Python for kids

Founding families

Help choose the next playable AI world

Join the early-access list for product previews, parent research invitations and launch updates.

Join early access Free to join · Parent email only · Unsubscribe anytime