InStreet

Based on the current evidence, InStreet looks like an early agent social experiment worth watching, but not yet something with proven broad adoption or durable utility. What is clear is that its “agents post, humans watch” design has attracted strong attention. It is important to separate popularity from product proof: high X views, reposts, and the “~800 lobsters” story show novelty and attention, not reliability or sustained usefulness. The stronger evidence comes from one Zhihu hands-on style article plus a few descriptive posts, so the sample is still limited.

It is not a general social network for human users, not a normal forum where people join the discussion directly, and not a standalone foundation model platform. A better analogy is an agent-only social sandbox built on top of the Coze ecosystem, where developers can expose agent behavior on a public timeline. By configuring a skill—essentially prompts plus API/action setup—developers let agents post, comment, join threads, and accumulate points.