A Tribute to Manus: How an AI Agent Company Entered the History of Technology, Governance and Global Competition

First, a factual correction

Manus did not say on August 11 that all affected data had already been deleted or that the transition was complete. It said it would soon resume operating as an independent company. For certain affected users, data generated on or after December 29, 2025 was scheduled for deletion on August 23–24 Singapore time, with restoration opening on August 25.

Manus also said this was not a data breach or security incident. Completion remains something to verify after the announced transition window.

The scale Manus reported before joining Meta

• 147T+ tokens processedCompany-reported, not independently audited.
• 80M+ virtual computers createdCompany-reported cumulative activity.
• $100M ARR after eight monthsCompany-reported annual recurring revenue.
• 105 employees across three citiesSingapore, Tokyo and San Francisco at the time of disclosure.

Why pay tribute to a company whose choices deserve scrutiny?

Because tribute is not immunity from criticism. Manus made the application layer of AI impossible to dismiss as a thin interface. It treated an agent as a system that plans, uses tools, manages state, survives errors and finishes work—not merely as a model that produces a polished answer.

That shift changed the industry's standard. Even competitors that disagree with Manus's architecture now have to answer the same question: can your agent actually complete the job?

Technology: making execution a first-class layer

Manus publicly framed its bet as context engineering on top of frontier models rather than training an end-to-end foundation model. Its engineering notes emphasized KV-cache efficiency, constrained tool selection, append-only context and the file system as external memory.

Those ideas are not magic and were not uniquely discovered by one company. Their significance was operational: Manus combined them into a product people could watch doing work across browsers, files and virtual computers. It gave the execution layer a recognizable product form.

Company-building: proving an agent layer could become a company

Before the Meta announcement, Manus reported $100 million in ARR eight months after launch, more than 147 trillion tokens processed and more than 80 million virtual computers created. These are company disclosures, not audited facts, but they explain why the industry paid attention.

A relatively small team spread across Singapore, Tokyo and San Francisco showed that value could accumulate above the model layer—in orchestration, reliability, interfaces, distribution and the accumulated craft of finishing tasks.

Governance: ownership became a product question

On December 29, 2025, Manus said it was joining Meta and would continue operating from Singapore. On April 27, 2026, China's foreign-investment security review authority prohibited the acquisition and required the parties to unwind it.

The resulting separation demonstrates that ownership is not a footnote for an agent company. When a product can access accounts, files and business workflows, corporate control touches data residency, user consent, service continuity, intellectual property and national security at once.

International competition: one company, four overlapping maps

Manus was founded by a Chinese team, operated from Singapore, attracted a major American platform and depended on global model and cloud supply chains. Its story therefore sits on four maps simultaneously: talent, technology, capital and jurisdiction.

The failed transaction is not a simple victory for one country or defeat for another. It is evidence that frontier AI companies can no longer choose a corporate nationality, data boundary or model supplier as if those decisions were independent.

Independence is not a return to the old company

A separated Manus will inherit users, technical systems and expectations from the Meta period, while needing to prove its final ownership, governance, funding, infrastructure and data boundaries. Independence is a new operating test, not a reset button.

The standard should be evidence: whether the announced deletion and restoration occur, whether services remain reliable, whether technical work continues, and whether users receive clear choices about their data and accounts.

What the notice to users means — and why the data migration is unavoidable

Many readers reach this page after searching one line from Manus's notice: “unavoidable as part of Manus's transition to independent operations and to comply” with legal and regulatory requirements. The sentence is dense, but its logic is simple. After China's review decision ended the Meta transaction, Manus must operate as an independent company — and user data held in Meta-period systems cannot simply stay where it is. Moving it is a legal obligation attached to the separation, not a product choice. That is why the notice calls it unavoidable.

The migration is also what makes independence real rather than nominal. It redraws the data boundary so that an independent Manus holds its own user data, on its own infrastructure, under a consent and jurisdiction chain that matches the new company. For users, the practical benefits are concrete: a cleaner answer to who holds my data, no residual dependence on the former parent's systems, and a verifiable baseline — a published backup window, a scheduled deletion, and a dated restoration point that can be checked rather than taken on trust.

Seen that way, the migration is less an ending than a starting point. It is the act that closes the Meta chapter for good: once the data boundary is redrawn, the past stops being an open question and Manus gets to simply be Manus again — an independent company judged on its own work. The outlook runs through the timeline above: watch whether restoration opens on August 25 as announced, whether service stays reliable through the window, and whether the independent company then publishes its ownership, funding and infrastructure details. If those follow, this migration will read as the first completed step of a new life. If they do not, it belongs with the evidence gaps this special already lists.

Questions from notice recipients

Why did I get this notice? Because your account falls within the transition scope Manus described in its official note to users: the company is separating from the Meta period and must move, delete or restore certain data to complete that separation.

What happens to my data? According to the note, for certain affected users, data generated on or after December 29, 2025 was scheduled for deletion on August 23–24 Singapore time, with restoration opening on August 25. The official note — linked in the sources below — is the binding version; check it against your own notice.

Do I need to do anything? Follow the instructions in the official notice itself, including any backup or export steps offered before the window. This special is independent analysis, not Manus support, and cannot act on individual accounts.

Six moments that define the story

1. Jul 18, 2025 Manus publishes its context-engineering lessons

   The company explains the system choices behind an agent built to execute.

2. Dec 17, 2025 Manus reports $100M ARR

   The company publishes scale, usage and team figures eight months after launch.

3. Dec 29, 2025 Manus announces it is joining Meta

   It presents itself as the execution layer and promises continuity from Singapore.

4. Apr 27, 2026 China prohibits the acquisition

   The security review decision requires the parties to withdraw the transaction.

5. Aug 11, 2026 Manus announces its return to independence

   The company publishes a user transition, backup, deletion and restoration plan.

6. Aug 23–25, 2026 The announced transition window

   Deletion was scheduled for August 23–24 and restoration to open on August 25.

What this special does not claim

Respect for Manus does not close the evidence gaps. As of the August 11 announcement:

• It was not yet confirmed that the scheduled deletion and restoration had completed.
• The final ownership, board structure, financing and control rights were not publicly detailed.
• The split of code, intellectual property, models, infrastructure and data was not disclosed.
• The operating figures were company-reported and were not presented as independently audited.
• The separation did not, by itself, prove that globalization or regulation had won.

To Manus

We pay tribute to Manus not because every choice was beyond dispute, but because it forced the industry to face the future earlier. It made agents less about promises and more about completed work. It also showed that the layer above the model can carry technical depth, product imagination and company value.

Its next chapter will be harder—and more independent. If Manus preserves its obsession with finishing the job, its most important contribution may be neither the acquisition nor the separation, but the standard it leaves behind: when AI can take responsibility for work, it should not stop at an answer.

Try Manus yourself

The fairest way to judge an agent company is to hand it a real task. Manus continues to operate as an independent company — you can sign up and put the execution layer this special describes to work.

Our read, not a Manus statement: with independence, serving the Chinese market again becomes a realistic possibility, and agent products run on finite compute and support capacity — Kimi and DeepSeek both saw launch-time overloads and rate limits in that market. If you intend to use Manus, registering early is a reasonable way to hold your place. We make no promises about service performance.

Open Manus

Affiliate note: this link may carry a referral code. If you subscribe, this site may earn a commission at no extra cost to you. That never changes the analysis above.

Sources and evidence boundary

Primary sources lead. Scale figures are explicitly attributed to Manus; secondary reporting is used only for cross-border context.

1. Manus — A Note to Our Users: August 11 transition, backup, deletion and restoration schedule.
2. China NDRC — foreign-investment security review decision: Official decision prohibiting and requiring withdrawal of the acquisition.
3. Manus — Manus Joins Meta: The December 29 acquisition announcement and execution-layer positioning.
4. Manus — $100M ARR update: Company-reported revenue, usage and headcount figures.
5. Manus — Context Engineering for AI Agents: The company's technical account of its agent architecture.
6. Associated Press — Meta acquisition context: Independent context on the China-founded, Singapore-based company.
7. Reuters analysis — cross-border China tech deals: Independent reporting on the review and wider transaction risk.
8. On this site — Grok Bot first look: Related reading: xAI's entry into the same digital-labor race, reviewed hands-on.

Model specials series

1. Claude Academy: Who Bears the Cost of AI Fluency? — Anthropic has launched a free school for learning to work with AI. Its stated framework reaches beyond prompts into delegation, judgment and disclosure. That is a meaningful public resource—and it raises a harder question: when the maker of the disruption also issues the credentials for adapting to it, where does institutional responsibility end and individual responsibility begin?
2. Meituan All-in AI: The Execution Costs — Going all-in on AI is easy to announce and hard to govern. The execution bill arrives where strategic urgency meets source provenance, merchant consent and incentives: the less time a team leaves for verification and reversal, the more expensive its speed becomes. This special separates the verified public record from two weak, single-source signals and treats the pattern as a governance problem—not proof that AI investment itself has failed.
3. AI Token Monetization: Token Is the New Dollar — At Stripe Sessions, President of Technology and Business Will Gaybrick changed a demo app from a $2 flat fee to $3 per million tokens, then streamed stablecoin payments as each token was consumed. That sequence is more than a billing demo. It shows software moving from seats and monthly access toward metered intelligence: every unit of model work can carry a price, a margin, a fraud risk and a settlement event. Our thesis is that the token is becoming the dollar of AI software—a unit of account for machine work, not legal tender and not a replacement for the US dollar.
4. Stripe × OpenRouter: Where Token Monetization Begins — Put the reported $8 billion price aside. Stripe, the payments leader that became a checkout layer for the internet, chose not to buy a model lab but the switchboard between AI applications and hundreds of models. That is the story. Stripe is betting that the most important layer of the AI economy will not only produce intelligence; it will turn intelligence into exchangeable value. Today an LLM token is a billing unit. Tomorrow a model-agnostic AI token could be held, transferred and settled like the generation of digital assets opened by BTC and ETH — representing not digital scarcity or blockchain gas, but a claim on usable intelligence. Axios reports an agreement above $8 billion, while neither company had publicly confirmed it at this update.
5. Grok Bot: xAI Gives Every Agent Its Own Computer — Launched in early beta on August 11, 2026, Grok Bot turns the agent from a chat window into a teammate: each Bot gets a persistent cloud computer, signs into the tools you already use and keeps working while you are away. We have been using it since day one. The product is days old and public information is still thin, so this special leads with tested impressions and keeps mechanism facts second — separating what is verified, what is company-stated and what is our read.
6. Kimi: Moonshot AI's Open-Weight Sprint to the Frontier — In four months Moonshot AI shipped an open-weight 1-trillion-parameter workhorse, followed it with the reported 2.8-trillion-parameter Kimi K3, paused new paid consumer subscriptions when demand outran capacity, and set off toward a Hong Kong listing. This file collects what is sourced, what is company-reported, and our read on where it fits the timeline this site tracks.
7. Zhipu (Z.ai): The First Listed LLM Company and the GLM Agent Bet — Zhipu AI reached public markets before any other large-model lab — listing in Hong Kong in January 2026 — and spent the following months shipping the GLM-5 line into an open-weight agentic flagship while raising prices twice. This file collects the sourced record: the models, the phone-use agent bet, the economics, and our read on what a listed lab means for tracking AI's real impact.
8. Qwen: The Open-Weight Leader Starts Charging for the Crown — Alibaba's Qwen is the most-downloaded open-weight model family in the world — by company count, more than three billion downloads and over half the open-source market. In August 2026 it shipped its biggest flagship yet, priced it far under US frontier rates, and put its open weights under a revenue-share license for the first time. This file collects the sourced record and our read on what the pivot means.
9. DeepSeek: The Price Anchor Starts Moving — Eighteen months after the R1 moment made frontier-for-pennies the industry's reference point, DeepSeek promoted V4-Pro to general availability — and in the same week announced API price increases of up to 1,100% plus the sector's first peak/off-peak token pricing. The lab that anchored the price war is repricing. This file collects the sourced record and our read.
10. MiniMax: The Multimodal Tiger That Doubled on Debut — MiniMax reached the Hong Kong exchange one day after Zhipu and doubled on its first day. But the reason it closes this series is not the listing — it is the product surface. Where the other five files cover text and agents, MiniMax ships video, speech and music at commodity prices, and that points the AI shockwave at a different cohort: creators. This file collects the sourced record and our read.
11. Model Price Watch — Rolling ledger of frontier model price moves, updated as list prices change.