Manus AI Founder, Tao Zhang: Turning Intent into Intelligent Action

Tao, co-founder and Chief Product Officer of Butterfly Effect (Manus AI), brings 15+ years of startup experience to the AI sector after joining the trend two years ago.

• His founding vision was to create a product capable of utilizing 24 hours of GPU time daily per user, aiming to move beyond the current average of 30 minutes of consumption.
• Manus AI launched on March 5, 2025, following a pivot from a previous product, "Monica."

The company evolved through three distinct product iterations based on user friction points:

• Monica (Extension): Launched a Chrome extension to bridge ChatGPT and existing workflows; grew to 20 million users generating ~50 million in AR, requiring minimal maintenance (~1 hour/week).
• AI Browser: A seven-month project (March–October 2023) abandoned because the "hands-off" experience felt unnatural to users who had to stare at screens waiting for AI completion.
• Manus (Autonomous Agent): Inspired by Cursor's adoption by non-coders, the team built a "right panel of Cursor" for average users, launched in November 2024 and released publicly in May 2025.

Manus AI achieved viral status immediately post-launch:

• Accumulated 3.5 million waitlist signups within the first two weeks of the March 2025 video launch.
• User base split evenly between AI pioneers testing new tools and non-technical new users.
• Transitioned from waitlist to public release in May 2025 by resolving significant infrastructure constraints.

Operational and Technical Strategy:

• Capacity Constraints: The initial waitlist was necessitated by an industry-wide shortage of LLM capacity; autonomous agents consume approximately 1,000x more tokens than standard chatbots.
• Infrastructure Solutions: Capacity was secured through deep partnerships with Anthropic, AWS, and Google Cloud, securing private resources and utilizing prompt caching optimization.
• Model Selection: Currently utilizes Anthropic Sonnet and Google Gemini 2.5 Pro, benchmarked solely on performance rather than ranking positions.
• Benchmark Philosophy: The team runs GAIA benchmarks occasionally to validate agent frameworks but does not actively chase rankings; they accept being consistently second-best in public leaderboards as a signal of market standards.

Product Evolution and Feature Development:

• Slides Generation: Initially generated via Python libraries, the output quality was poor; the team pivoted in late May to use web technologies (HTML, JS, CSS) generated by frontier models, converting them to PowerPoint for superior layout and styling.
• Simplicity over Flexibility: The product intentionally hides model selection options from users to avoid complexity, prioritizing a consumer-grade experience for non-technical users rather than enabling expert configuration.
• Non-Developer Focus: The core philosophy targets "average users" rather than coders, enabling tasks like data visualization and file conversion without requiring code writing or quality evaluation.

AGI Vision and Strategic Outlook:

• Pragmatic Approach: The team views the "AGI" label as a shifting standard rather than a reachable destination, focusing instead on immediate revenue and user experience improvements.
• Proof of Autonomy: Demonstrated advanced agent reasoning in January 2025 when the system autonomously solved a GAIA task requiring video playback and frame analysis using obscure YouTube keyboard shortcuts (e.g., 'K' for play/pause, '3' for progress jumping) without human intervention.
• Call to Action: Tao encourages users to transition from iterative prompting (ChatGPT style) to goal-oriented automation, framing the shift as a new era of AI tool design.

Leadership and Company Culture:

• Founding Team: Co-founders Tao (CPO), Chao Hong (CEO), and Peak Ji (Chief Scientist/Former CEO) operate with blurred role definitions.
• Culture: The team maintains a "paranoid" culture regarding innovation, capable of "zooming in" on technical details or "zooming out" for high-level strategy due to their diverse backgrounds in product, engineering, and business.