Drill · lap 2 · 6/10 · MAS-102 · A swarm that does not know how to stop

[akash-coded]

6 of 10 · MAS-102 · A swarm that does not know how to stop

multi-agent · medium · fix · ~12 min · no AWS account

This swarm works beautifully when the agents converge. When they do not, it is a spending rate with a diagram.

def run_swarm(step, task):
    """step(task, state) -> (state, done, tokens_used). Loops until done."""
    state, spent = {}, 0
    while True:
        state, done, tokens = step(task, state)
        spent += tokens
        if done:
            return {"outcome": "completed", "state": state, "tokens": spent}

Fix it

Add two bounds, both in code — a cap in a prompt is a suggestion:

• max_rounds (default 8) — stop after this many calls to step
• max_tokens (default 50_000) — stop once spent exceeds it

When a bound binds, return {"outcome": "stopped", "reason": "<which bound, and the numbers>", "state": state, "tokens": spent, "rounds": n}. A converged run returns "completed" exactly as before, plus "rounds".

/drill MAS-102

```python
def run_swarm(step, task, max_rounds=8, max_tokens=50_000):
    ...
```

What this proves

That you never ship a loop whose termination depends on the model's cooperation, and that a stop is reported as an outcome — not a crash, not a fake completion.

---

Submit in this thread

Copy the block from the brief, edit it, and post it as a comment here. The bot replies within a couple of minutes with every check, the skill you demonstrated, and what to do next.

_{Brief in the repo: labs/drills/multi-agent/MAS-102 · next in the sequence: EVAL-102 · how grading works · scoreboard}