Write and debug code across eight test tasks
Manus has changed hands twice since we tested it. Task: The reviewer executed eight specific coding tasks using Manus, covering Python scripts, debugging, algorithm explanations, HTML/CSS pages, web apps, code translation, React scaffolds, and a browser game. Manus generated code and explanations autonomously, while the reviewer manually tested outputs for crashes, edge cases, and accessibility. The workflow relied on Manus's autonomous agent capabilities to produce first drafts that were then verified by a human.; tools: Manus. Human role: The original post does not say how a person stayed involved.; limitations: Manus's underlying AI models have not been disclosed since March 2025, creating uncertainty about its technical foundation. The company underwent significant ownership changes, including a blocked acquisition by Meta and a return to independent operations in September 2026. Pricing starts at $20/month for 4,000 credits, but hosting and specific model access may incur additional costs..
Tools
Published on Agent Work Atlas:
Overview
Manus has changed hands twice since we tested it.
Workflow
The reviewer executed eight specific coding tasks using Manus, covering Python scripts, debugging, algorithm explanations, HTML/CSS pages, web apps, code translation, React scaffolds, and a browser game. Manus generated code and explanations autonomously, while the reviewer manually tested outputs for crashes, edge cases, and accessibility. The workflow relied on Manus's autonomous agent capabilities to produce first drafts that were then verified by a human.
Reported result
Manus achieved an average score of 4.69 out of 5, with perfect scores on four tasks including a currency converter and a real estate landing page. It failed to execute code in one translation task and lacked keyboard support in a game task. The source notes that Manus reported $100 million in annual recurring revenue in December 2025, though this is a self-reported figure.
Human role
The original post does not say how a person stayed involved.
Limitations
Manus's underlying AI models have not been disclosed since March 2025, creating uncertainty about its technical foundation. The company underwent significant ownership changes, including a blocked acquisition by Meta and a return to independent operations in September 2026. Pricing starts at $20/month for 4,000 credits, but hosting and specific model access may incur additional costs.
Try it
Act as a senior developer evaluating Manus AI for coding tasks. I will provide a specific coding prompt: [replaceable input: e.g., 'Create a Python script to validate email formats']. Use Manus to generate the code and explanation. Do not deploy or publish the output. Instead, perform a dry run by executing the code in a local sandbox or reviewing it line-by-line for edge cases, input validation, and accessibility. Report the results, including any errors or missing features. If the code fails or lacks validation, note these as critical gaps. Only proceed to integration if the human review confirms the code is robust and meets all requirements. Ask for clarification if the prompt is ambiguous.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.