Manus AI Agent: Can It Solve My 3 Challenges?
Based on All About AI's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.
Manus generated and executed a Python script that parsed JSON sports statistics and exported an Excel file with aggregated player stat counts.
Briefing
Manus AI agent delivered real, task-completing results across three tests—automating a paid coding job, producing a data-driven career report, and sending an email newsletter through Proton Mail—while also exposing the rough edges that still show up in formatting and attachment handling.
The first challenge was a coding job posted on Upwork with a stated $10 reward. After Manus was given a raw JSON example and the job URL, it navigated to the task details, inferred the requirements, and generated a Python script to parse football player statistics from the JSON. The agent iterated through multiple command-line attempts, then produced an Excel output summarizing player counts by stat categories (singles, doubles, and related metrics). The transcript notes the script ran successfully, the Excel file was generated, and a local preview in Google Sheets showed the expected columns and aggregated statistics. The creator then concludes the job was effectively completed and claims the agent “made us $10.”
The second test shifted from coding to analysis and writing. Manus was fed a CSV dataset of post-college salaries (700 occupations) and instructed to recommend the top three careers using arguments grounded in technology progress, geopolitics, and macroeconomics. It also had to produce results in an MDX document with graphs and visualizations. Manus pulled in external trend research—citing items like Gartner’s strategic technology trends for 2025 and Accenture-style reporting—then combined those signals with the salary dataset and broader economic expectations (including inflation easing and rising unemployment projections). The resulting report emphasized criteria such as financial potential, meaningfulness/fulfillment, and resilience to future change, including technological resilience, geopolitical resilience, and economic resilience.
The top three careers it recommended were petroleum engineering, information and computer science, and building science. Petroleum engineering drew skepticism from the tester, but the report’s rationale leaned on financial rewards and “future change” resilience, arguing that energy demand remains relatively inelastic and that geopolitical competition for resources keeps the field valuable even amid energy transition. Information and computer science was framed around AI and automation and the growing geopolitical importance of digital systems. Building science was positioned as a response to demographic shifts, economic cycles, and long-run infrastructure needs.
The third challenge tested email automation. Manus logged into Proton Mail inside a containerized browser session, opened the latest email, drafted a polite reply, and sent it. It then gathered geopolitics research (including topics tied to Donald Trump, Russia/Ukraine, NATO leadership, the Middle East, and energy-related issues) and attempted to send a newsletter-style email back to the tester. The process worked end-to-end—an email arrived—but formatting was imperfect and attachment behavior was inconsistent: the agent initially seemed to think it was done without correctly attaching or rendering the newsletter, requiring a retry. Overall, Manus wasn’t portrayed as “AGI,” but it was shown as capable of producing useful outputs with meaningful autonomy, plus clear room for improvement in reliability and presentation.
Cornell Notes
Manus AI agent completed three practical tasks with minimal human intervention: generating a Python script to satisfy an Upwork coding job, producing an MDX career report from a 700-row salary dataset plus external trend research, and automating Proton Mail to reply and send a geopolitics newsletter. In the coding test, it parsed JSON football stats and exported an Excel file with aggregated player statistics. In the career test, it combined salary signals with technology, geopolitics, and macroeconomic resilience criteria, recommending petroleum engineering, information and computer science, and building science. The email test largely worked, but newsletter formatting/attachment handling was inconsistent, highlighting where automation still needs polish. The results matter because they show what “agentic” workflows can already do—and where they still fail under real-world constraints.
How did Manus handle the Upwork coding challenge, and what was the measurable output?
What inputs and constraints shaped the career recommendation report?
What criteria did the MDX report use to rank careers?
Why did petroleum engineering appear in the top three despite an energy-transition narrative?
What succeeded—and what broke—during the Proton Mail newsletter automation?
Review Questions
- In the Upwork coding test, what chain of steps led from JSON input to an Excel deliverable, and what evidence suggests the script met the job requirements?
- Which three careers were recommended in the MDX report, and how did the report connect each one to technology, geopolitics, and macroeconomic resilience?
- During the email automation test, what specific failure mode occurred with the newsletter (formatting vs. attachment), and how was it resolved?
Key Points
- 1
Manus generated and executed a Python script that parsed JSON sports statistics and exported an Excel file with aggregated player stat counts.
- 2
The Upwork-style workflow relied on Manus opening the job URL, inferring requirements, and iterating through command-line attempts until the deliverable was produced.
- 3
The career report combined a 700-occupation salary CSV with external technology and geopolitical trend research, then wrote results in MDX with visualizations.
- 4
The MDX report’s ranking framework emphasized financial potential, meaningfulness/fulfillment, and resilience across technological, geopolitical, and economic dimensions.
- 5
Manus recommended petroleum engineering, information and computer science, and building science, with petroleum engineering justified by energy-demand inelasticity and resource geopolitics.
- 6
Proton Mail automation worked for login, replying, and sending, but newsletter delivery showed reliability issues around formatting and attachment behavior.
- 7
The transcript frames Manus as capable of useful agentic automation today, while still falling short on polish and robustness in real-world communication tasks.