In the Classroom¶

MEADOWS lets students write software that participates — not software that sits in a folder waiting to be graded. A student's bot talks to other students' bots, reacts to messages it never anticipated, and becomes part of a shared conversation. The reward loop is immediate and social.

First contact with real distributed software¶

Students write a bot that answers ping with pong, then grow it. Because the server handles the plumbing — auth, routing, persistence — they spend their attention on logic and meaning, not YAML and TLS. First working bot in minutes, not weeks.

# The smallest MEADOWS bot — copy, rename, run
from meadows.bot import BaseBot

class PingBot(BaseBot):
    BOT_NAME = "ping"
    BOT_DESCRIPTION = "Answers ping with pong"

    def should_handle(self, command, args):
        return command == "ping"

    def handle(self, command, args, raw_args, message, thread_context):
        if command == "ping":
            return "Pong!"
        return None

if __name__ == "__main__":
    PingBot().connect()

See the first bot tutorial for the full walkthrough.

The living systems-thinking lab¶

A class populates a room with dual-context bots — each a "person" with beliefs, reacting to a shared event stream. Students watch consensus, polarisation, and adaptation emerge, then intervene at different leverage points and see what actually shifts. Donella Meadows, but you can poke it.

Classification and ML you can argue with¶

Students write label subscriptions with JSON Logic predicates, disagree about what a message "is," and discover that categories are choices with consequences. A perfect on-ramp to machine learning that starts from judgement, not maths.

# A label subscription that matches sentiment labels from bot-sentiment
bot.register_label_subscription(
    "sentiment_alerts",
    {
        "and": [
            {"regex_match": [{"var": "origin"}, "^bot-sentiment$"]},
            {"regex_match": [{"var": "label"}, "^sentiment$"]},
            {"semver_match": ["^1.0.0", {"var": "semver"}]},
        ]
    },
    scope="global",
    deliver="label_only",
)

Media and information literacy¶

Bots that summarise search results — and sometimes get it confidently wrong. Students learn to check sources, spot fabrication, and evaluate what an AI hands them, by building the thing that hands it to them.

Model a social system — a market, a rumour spreading, a negotiation — as interacting bots. The dynamics become tangible instead of described.

What students actually learn¶

Concept	Where it shows up in MEADOWS
Message-passing	Bot receives messages, emits responses
Remote procedure invocation	`call_rpc()` — call another bot's service
Distributed systems	Dumb coordinator, no direct bot-to-bot communication
Classification	Labeling with JSON Logic predicates
Systems thinking	Dual-context bot, feedback loops, temporal dynamics
Critical thinking about categories	Who decides what a label means? What happens when they disagree?
Source evaluation	RAG bots that sometimes fabricate; students catch them

Check your understanding¶

Why does a MEADOWS bot not need its own server or database?
What is the difference between a bot's should_handle() and a label subscription?
How does the append-only property change what students can do with conversation history?

See the Concepts page for the building blocks, or jump to the first bot tutorial.