Bloome

Verdict: the current evidence supports viewing Bloome as a chat-native multi-agent collaboration workspace worth testing, but still as an early-stage adoption candidate rather than a proven mature platform. It is not just an AI chat wrapper, not a standard enterprise messenger, and not a dependable RPA substitute. Proof of attention mainly comes from three X posts with strong sharing; proof of usefulness comes more from one longer Zhihu hands-on write-up and a small number of user demos. That is enough to show plausible workflows, but not enough to confirm long-term reliability, governance, or enterprise readiness.

In practice, the evidence suggests it works best when one task is split across several roles in sequence: one agent drafts a competitive analysis, another checks for gaps, and a third enriches it with new information. The Zhihu test also describes chaining roles like curriculum research, product, design, and engineering inside one group chat. A more accurate analogy is not “a tool for sharing Claude” and not “Slack with AI,” but an experimental IM-style product for running multi-role AI workflows.