Conversation Collaboration Patterns

Reward the collaboration that happens outside the code. Managers define contribution signals ("proposed an idea", "helped a peer") with a fixed CP value. A daily AI pass reads the challenge's Slack channel, detects those signals and credits authors automatically, with an auditable justification for every reward.

Read the brief

3 participants · own branch each

Why this challenge exists

Much of a team's collaboration happens outside the code: ideas proposed, peers helped, decisions discussed in Slack. Code and ML evaluation can't see any of it. Discussion signals make that work visible and rewarded.

Who hosts this challenge

MyTwin Lab

The brief

Objective

A manager links a challenge to a public Slack channel and defines contribution signals. A daily job fetches new messages via conversations.history and matches authors to leaderboard users by email. A GPT-4o agent then detects the signals, and each detection is credited as a ledger row.

Expected result

• Signal rewards are fixed and out of pool: they never drain the challenge budget

• Every reward is auditable (signal, message, excerpt, one-sentence AI justification)

• The per-challenge cursor only advances on success, and deduplication guarantees a (user, signal, message) triple is paid once

• Detection stays conservative: no match when in doubt

• Contributor profiles show signals as compact chips