Lucas Academy
Bringer of Light

Coding · Snake-Lab AI Camp

Meet the cleverest creature. Then learn to rule it.

The snake is just the beginning. The cleverest things on earth today are machines, games, and AI — and they get cleverer every year. Snake-Lab is where kids learn to understand them, command them, and out-think them.

“Now the snake was the most clever of all the wild animals the Lord God had made.”

Genesis 3:1a
Play now Contact us for dates & enrollment See camp recap

Think through coding.

A Snake-Lab camper at the bot-anatomy lecture, with real nextMove(state) code on the screen
Snake-Lab AI Camp in session — an instructor walking students through bot code

Real moments from Snake-Lab camp in Orange County, CA.

Coding is the tool, not the goal

The real subject here is thinking: how to break a problem down, weigh trade-offs, and decide. Code is just how kids make their thinking run.

Real slides from the camp deck — bright lectures and the dark “learns by itself” lesson. Swipe, tap the arrows, or use your ← → keys.

So we start by looking the clever creature in the eye — and seeing what it really is.

L1 · Part 2 · Lecture

What IS the snake?

⏱ 15 min
Snake.js
constructor(startX, startY, length) {
  this.body = [];
  for (let i = 0; i < length; i++) {
    this.body.push({ x: startX - i, y: startY });
  }
  this.direction = 'RIGHT';
}

In plain English

  • body is a list.
  • Each item is a cell — an {x, y}.
  • body[0] is the HEAD.
  • The rest of the list is the body.
💡A “snake” is just a list of {x, y} cells. No creature — only data.
🐍 Snake Lab · Code CampL1

L2 · Part 2 · Lecture

The phone call

⏱ 15 min

Client → Server

📱iPad
📱iPad
📱iPad
💻 THE SERVERthe truth lives here

Your iPad is a CLIENT. One laptop is the SERVER. Everyone connects to the same server — and the server is the truth.

Watch it live — the server log

terminal
[room cosmic-viper]
  +p1 Curly   #4ade80  (1/8)
  +p2 Wiggles #fbbf24  (2/8)
👀Each kid who joins shows up here, live.
🐍 Snake Lab · Code CampL2

L3 · Part 5 · Hands-on

Your first AI conversation

⏱ 15 min
🤖 Your turn
Open Claude / ChatGPT. Ask it to write a bot that does something SILLY.

Try one of these

  • “A bot that tries to spell my name with its body”
  • “A bot that runs AWAY from food”
  • “A bot that only ever turns left”
💡The AI is the partner — but YOU are the boss. You decide what it does; you debug what it actually does.
🐍 Snake Lab · Code CampL3

L4 · Part 1 · Play

The Tunable bot — 4 knobs

⏱ 10 min
tunable.js
const W_FOOD     = 1.0;
const W_SAFETY   = 5.0;
const W_BLOCKING = 0.0;
const W_OPEN     = 0.5;

Don’t change anything yet

Click the Tunable sample, create a room, and just play. The bot is competent — it survives and eats. The four weights at the top are its whole personality.

💡Four numbers decide everything this bot does. Next: how.
🐍 Snake Lab · Code CampL4

L4 · Part 3 · Hands-on

Tune your bot

⏱ 25 min

Open “Edit bot” and try these

  • Reckless eaterW_FOOD=5, W_SAFETY=1
  • CowardW_FOOD=1, W_SAFETY=10
  • BullyW_BLOCKING=5
  • ExplorerW_OPEN=5, W_FOOD=0.1

After each change, WATCH:

  • How long did it survive?
  • How many foods?
  • Longest length?
💡Change → run → observe → change again. Force 30 seconds of watching between tweaks!
🐍 Snake Lab · Code CampL4

L4 · Part 5 · The reveal

You are the gradient

⏱ 10 min
“What you just did — try numbers, watch what happens, pick better numbers, run again — has a name. It’s how MACHINE LEARNING works. The math version adjusts millions of numbers, but the loop is exactly the same.”
You are the gradient.

After lunch: a bot that does this loop by itself.

🐍 Snake Lab · Code CampL4
L5

Lesson 5 · Day 3 AM · v3 Coder

The Bot Learns by Itself

BIG IDEA  Yesterday you tuned weights by hand. Today we hand the tuning to the bot itself. It plays a game, dies, plays again — and gets a little better every round. That’s reinforcement learning.

What you’ll learn

  • The Learning bot keeps a Q-table: for each situation, how good is each move?
  • After every move it updates the table from a reward
  • It explores first (random), then exploits what it learned

The parts

  • 1Watch it be BAD10m
  • 2The Q-table20m
  • 3Explore vs Exploit10m
🐍 Snake Lab · Code CampL5

L5 · Part 2 · Lecture

The Q-table — the bot’s notebook

⏱ 20 min

Per SITUATION it has seen: 4 numbers — one for each direction.

qTable
situation "1,0|2"   →
   UP:   2.1   DOWN: -4.0
   LEFT: 0.3   RIGHT: 8.7  ← best

// "1,0" food right · "|2" danger left

After every move

+10
ate food
−50
died

Mark the move that led here as good (or bad). Next time in a similar spot, it remembers.

💡See → pick → move → see what happened → update the notebook → repeat. Yesterday YOU were this loop.
🐍 Snake Lab · Code CampL5

Play it now

The games behind the camp — play right in your browser, then work your way up to writing bots that play for you.

🐍 Just play

Classic Snake

The original — eat, grow, don’t crash. See how the game really works.

 🤖 Code a bot

The Challenge

Climb a ladder of rival bots: play by hand, then write a bot that plays for you.

 ⚔️ Compete

The Arena

Live multiplayer — go head-to-head with friends in a shared snake battle.

 🧠 Program bots

The Coder

Write a bot’s brain in code and watch it battle other players’ bots, live.

 ⚫ 🚧 Work in progress

Gomoku Challenge

Five-in-a-row: play by hand, then code a bot to out-think a ladder of rival bots. Early preview — not finalized yet.

Classic Snake and the Challenge run instantly. The live Arena and Coder are multiplayer servers — they can take ~30 seconds to wake up on the first visit.

Three honest AI threads

Three sentences every kid can finish by Friday.

Build with AI

Use Claude or ChatGPT as a coding partner the way engineers do: show it everything, then ask the right question.

“I asked an AI for help. It worked.”

Think like AI

Tune a bot’s weights by hand and feel what training is. You are the gradient.

“A strategy is just numbers I can tune.”

Watch AI learn

Watch a learning bot go from crashing into walls to eating on purpose — live.

“A bot can do my tuning by itself.”

The camp, day by day

Two lessons a day, no filler — three days that take kids from “the snake is just data” to “I trained a bot to play for me.”

Day 1 · Foundations

  • The snake is data, code is a map
  • Two computers talk — and the bot’s brain

Day 2 · Bots + AI

  • Hello, AI — meet your coding partner
  • Strategy — tune the bot by hand

Day 3 · Learn + Compete

  • The bot tunes itself (machine learning)
  • Tournament + parents demo

Bonus · Family Saturday

Parents welcome — King-snake and Fog-of-war play, and kids demo their bots. (~3 hours)

Take-home

A self-paced 10-level bot ladder — Random Randy to Grandmaster — plus an earned Secret Lesson on out-thinking the champion.

“Rule over the fish in the sea and over the birds in the sky. Rule over every living thing that moves on the earth.”

Genesis 1:28b

We don’t fear the clever creature. We learn to rule it.

The snake was the cleverest creature in the garden. Today the cleverest things are machines, games, and AI — and they keep getting cleverer. The camp’s whole arc is one calling, made real: YOU are the boss. You are the gradient. The rule-writer chooses which strategy wins.

The part most camps can’t teach

Real game theory: three strategies that beat each other in a rock-paper-scissors loop, so there’s no single best bot. Kids meet reward hacking and AI safety the way researchers do — and learn that when no single strategy always wins, you stop following one recipe and start reasoning about the matchup.

When there’s no perfect move, kids learn to think — not just follow rules.
Read the ideas behind Snake-Lab →

Clear all ten home levels and an advanced session unlocks — four ideas that turn “I can’t out-survive this bot” into “I’ll out-think it.”

🔒

THE SECRET LESSON

How to beat a bot you can’t out-survive

Earned by clearing Level 10 · an advanced session beyond the camp

1Reward hacking 2Honest play 3Rock-paper-scissors 4Rules make the game
  • Reward hacking — win by exploiting the rule you were given.
  • Honest competence — out-play, or counter-pick when you can’t.
  • Rock-paper-scissors — no single best bot, only a cycle.
  • Rules make the game — the rule-writer chooses which strategy wins.

Keep climbing at home

A self-paced bot ladder of ten named opponents, from Random Randy to the Grandmaster — plus two expert bots, Achilles and Apex, waiting for kids ready to write a smarter bot.

🏆 Keep going after camp

🏆 Homework Challenge

A 10-bot ladder (Random Randy → Grandmaster), self-paced on your own device. Play by hand early; code your way up. Level 8 is the wall where reflexes stop and you must code.

🎓 Secret Lesson

Earned by clearing Level 10. How to beat a bot you can’t out-survive: find the one mistake it always makes — then let a computer search for the win. You’ll have trained an AI to beat another AI.

🎮Play at home — Challenge ladder runs in any browser · the Arena & Coder live online
  • For ages 10+.
  • In-person and remote; schedule varies.
  • We're a non-profit community — tuition covers cost only.
  • iPad-friendly; a play + lecture + guided discussion format.
  • Scripture quietly unlocks as kids climb the ladder.
  • Based in Orange County, CA.

Get camp dates & enrollment info

Parents and adults can leave an email, and we'll reach out when the next Snake-Lab dates are ready.

Prefer email? hello@lucasacademy.org