Getting Started with Manus 2.0

Manus shipped 2.0. The useful framing is not "more AI." It is what happens after the first result.

First drafts are cheap. Final cuts, live prototypes, overnight workflows, and agents that keep moving while you are away — that is where most teams stall. Manus 2.0 is aimed at that second half.

What you actually get

Piece	Use it when…
Cascade	You want the new agent engine under the hood (faster / leaner runs on their published test config — treat as directional, not a guarantee)
Manus Studio	You have real files on disk and need a site, short video, or game you can finish by hand
Video Editor	The AI cut is close, but you must swap audio, trim, or drop in real footage
Game Dev	You want a playable prototype this session; multiplayer only after solo works
Cloud Computer	Something must stay up — game servers, long-running automations
Automations	Work should start on a signal (email, Slack, calendar, ads, Notion)
Remote Control	You need the desktop from your phone without opening a laptop
Cue	You want a separate personal-agent app (invite / early access), not Studio creation tools

Three rules matter more than the feature list:

1. Keep one connected project — brief, page, video, and automation stay together.
2. Let AI do the first pass; you own the final cut.
3. Only automate or remote-control what you would trust a junior to do overnight.

Start in Studio, not in a blank prompt

Most real work starts in a folder: phone footage, product shots, a half-finished site, notes you already have.

Open Manus Studio on desktop, point it at that folder, and pick one narrow job:

• a landing page,
• a 30–60s product clip, or
• a tiny single-player game.

Manus produces the first version. You shape it with direct controls. Hand it back for another AI pass only when talking is faster than clicking.

For video, that means the timeline first — not regenerate-everything. Warmer song, your own product shot, shorter intro: edit, then ask for another pass if needed. Alchemy is for when the cut is already close and quality matters.

For games: start from a playable template, change one thing, play again. Buy a Cloud Computer only after solo play feels right and you actually need multiplayer.

First week: six steps

1. Connect Studio to a real project folder.
2. Ship one narrow artifact (page, short video, or tiny game).
3. Make at least one hand edit on the timeline or Edit panel.
4. Ask for a second AI pass on the same project.
5. Add one Automation that drafts only — human before publish.
6. Optional: try Cue with one small agent team if you have invite access.

Skip Cloud Computer, wide Remote Control permissions, and Cue wallets until that loop is boring.

What to avoid

• Treating Cascade's published efficiency numbers as a platform-wide SLA.
• Regenerating a whole video when a timeline edit would do.
• Buying always-on compute before a solo build works.
• Automating an unverified workflow end-to-end.
• Broad Computer Use / Remote Control on a shared machine.
• Cue agents with soft spend limits.

Bottom line

Use Manus 2.0 for the second half of creation: after the demo, when you need to swap the song, ship something friends can join, or keep work moving from your phone.

This week: one Studio project, one hand edit, one second AI pass. Automations, Cloud Computer, and Cue wait until that is routine.

References

• Introducing Manus 2.0
• What is new in Manus 2.0?
• Red Xiao on LinkedIn
• THE DECODER coverage
• Cue