Skip to main content
Canvas replaces the Telegram bot-check with a short sponsored conversation. A human trying to join a registered group exchanges a few messages with the Canvas AI agent, completing a task set by an advertiser or AI lab. The group owner earns USDC per verified join. The advertiser receives the conversation transcript and a Kimi quality score. The user gets access to the group.

​
The two sides

Advertisers

Buy verified human completions in bulk. Pay per passing verification. Receive transcripts, preference signals, or labeled training data depending on task type.

Group owners

Earn USDC on every verified join. Canvas replaces the bot-check — nothing else about group management changes.

​
Why Telegram groups

Telegram groups are among the densest concentrations of crypto-native users on the internet. A DeFi group with 5,000 members is a qualified, self-selected audience. The join moment is when that audience is most identifiable — they’re actively seeking access to a specific community, which means the group topic is a reliable proxy for their interests. Canvas captures that moment. The verification conversation happens anyway; Canvas makes it generate value instead of discarding it.

​
What makes Canvas completions different from standard annotation pipelines

Scale AI and Mechanical Turk produce responses from workers with no verifiable context. A Canvas completion comes with three things those pipelines cannot provide:
  • The group the user was joining at time of completion (audience context)
  • A Kimi quality score on the conversation transcript (engagement signal)
  • An onchain record of the completion on Base (cryptographic provenance)
A DeFi trader ranking protocol responses produces a different labeled example than a random worker doing the same task. The group context is the provenance — and it’s logged before the scoring happens.