Implication: The strategic bet is that agent usefulness improves when execution, tools, and revision live in one environment, but that also raises expectations for permissions, reliability, and ongoing operational control.
Watch condition: Look for independent benchmarks and user evidence showing whether event-triggered workflows and persistent execution remain reliable under real workloads, and whether users can inspect and constrain agent actions effectively.
New branch reason: No prior New Runtime record is supplied, so this announcement starts a distinct analytical record focused on Manus 2.0’s shift toward persistent, editable agent workspaces.
Thesis
Manus 2.0’s significance is the attempt to make an agent platform a persistent, editable work environment rather than a prompt-to-task service.
Observed facts
Manus says Cascade is the latest iteration of its in-house agent harness and reports that one tested configuration used 23.2% fewer tokens, finished tasks 28.2% faster, and cost 32% less than its previous system.
1The company says Automations can trigger from connected-service events including new email, Slack messages, and Notion updates.
1Manus Studio adds Video Editor and Game Dev environments; Cue is a separate personal-agent app in early access, free with an invite code.
1
Mechanism
Manus is combining agents with durable execution contexts and interfaces for human correction. Persistent Cloud Computers can host projects or automation beyond a user’s local session; event triggers let workflows respond to external state; timeline and in-editor controls let people refine generated media or code instead of restarting from prompts. Cue extends this model to agents with their own communication and compute identities.
Why now
This is a broad product launch presented as available now across web, desktop, and mobile for Manus 2.0, while Cue is in invite-only early access. It is a timely point to assess whether the agent category is moving from isolated task completion toward ongoing, user-supervised work.
Implications
The strategic bet is that agent usefulness improves when execution, tools, and revision live in one environment, but that also raises expectations for permissions, reliability, and ongoing operational control.
Evidence boundary
The feature descriptions, availability statements, and efficiency figures come from Manus’s own announcement. The performance comparison is limited to one tested configuration and is not independently validated here; the announcement does not establish broad comparative quality, reliability, or user adoption.
Watch conditions
Look for independent benchmarks and user evidence showing whether event-triggered workflows and persistent execution remain reliable under real workloads, and whether users can inspect and constrain agent actions effectively.