
What happened
Manus announced Manus 2.0, built on its in-house agent harness Cascade; in a test configuration it cut tokens by 23.2%, execution costs by 32%, and task completion time by 28.2% versus its previous system. It's available on Web, desktop, and mobile.
Why it matters
Lower token use and execution cost per task suggest AI agent work may get cheaper to run for the same jobs, while the 28.2% faster completion points to quicker turnaround on projects.
What to watch
The figures come from a test configuration, so real-world results may differ. Also watch Cue's iOS version, which is slated for release after App Store review.
WHO IT HITSTeams running AI agents for automation and content work — automating email, calendar, Slack, and Notion triggers — stand to see lower execution costs and faster task completion, though the published savings come from a test setup.
Summaries like this, every morning.
Manus is positioning its new release less as a single upgrade than as a rebuild of the layer that drives its agents. The in-house Cascade harness replaces the prior system and is credited with the token, cost, and speed improvements measured in a test configuration, which is the kind of efficiency claim that matters most to anyone paying for agent runs at volume.
Around that core, Manus has added pieces aimed at longer and more independent work: a Cloud Computer that keeps projects running even when the PC is closed, and event-driven automation that starts tasks from updates in email, calendar, Slack, or Notion rather than a fixed time. The desktop app has been rebranded Manus Studio, with dedicated environments for video and game development, and Remote Control and Computer Use let a phone direct work happening on a home PC.
The new Cue app extends the same infrastructure to personal agents that come with their own email address, phone number, wallet, and computer — enough to send messages, pay within a budget, and take on tasks, with several agents sharing a group chat. Whether these features pay off for Manus may hinge on how the test-configuration savings hold up in everyday use and on Cue clearing App Store review, but for teams already delegating work to agents, the pitch is cheaper, faster runs that keep going without someone at the keyboard.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
The author fed the 'desired candidate profile' text from 145 job postings to Claude, turned the extracted labe…

Researchers de-duplicating leak-site data say alleged ransomware attacks reached nearly 3,000 over the summer…

Jamie Dimon said JPMorgan Chase is "not gonna allow just any" outside AI agent inside the company, warning "Yo…

OpenAI is dissolving its standalone robotics team and folding researchers into the main research team under Ch…

Jacob Coxon, a former OpenAI and Anthropic researcher, cited an Anthropic IPO roadshow slide showing pre-train…

Internal data first reported by The New York Times show only 42% of Alpha School students met its goal of lear…
