Manus: Building the hands for AI

Attendees: Dustin Jeremy Ross Date: October 21, 2025 Type: Research

Summary

Manus AI Company Overview

• World’s first AI agent combining thought and action
• Co-founder: Tao (Kyle John), CPO based in Singapore
• Previous experience: Lightyear, ByteDance (TikTok parent), Census Data
• Raised ~$85M total ($10M + $75M benchmark round)
• Close to $100M ARR after 6 months of charging (started April)
• Moved headquarters to Singapore for cross-border privacy compliance

True AI Agents vs Chatbots

• Real agents must have agency: ability to think, plan, act autonomously
• Not just chatbots with system prompts pretending to be specialists
• General agent vs specific task agents (marketing, sales, video agents)
• Action engine concept: “you come with task, leave with deliverables”
• Can handle open-ended problems requiring creative solutions
  • Example: Research which top 500 company CEOs have partners who went to top universities

Three Frontiers of Agentic AI

• Atomic capabilities: Expanding what agents can do
  • Content creation (image, video, text generation)
  • Browser automation and web operations
  • Website building with natural language (v1.5 release)
  • Desktop software integration (Microsoft collaboration)
• Architecture: Long-running tasks and parallel processing
  • Current longest task: 1.5 hours
  • Goal: 24/7 agents running for individual users
  • Parallel subtask decomposition for complex research
• Proactive intelligence: Acting before being asked
  • Manus Connectors: Calendar, email, Google Drive, Linear, Asana
  • Internal prototype: Reviews 24-hour email, identifies undone tasks
  • Solves tasks proactively, asks for authorization to proceed

Geographic Distribution & Growth

• Largest markets: US, Brazil, UK, Japan, Germany, Saudi Arabia, France
• Viral growth through word-of-mouth and influencers
  • Egypt: Facebook influencer with 20K followers
  • Brazil: Two YouTubers created viral videos
• No formal marketing team until recently
• Key insight: Deliver great product that people want to talk about

Business Model & Cost Optimization

• Initially 100x more expensive than chatbots (token consumption)
• Reduced costs 8x since March through:
  • Direct partnerships with model providers (Anthropic, OpenAI) for discounts
  • System re-architecture to reduce token usage
• Waitlist system initially to control burn rate
• Started charging in April, approaching $100M ARR

Key Insights & Takeaways

• AI adoption mindset: Enhance humans, don’t replace them
• Focus on benefits and opportunities, not risks
• CEO-level commitment required for enterprise AI transformation
• Transparency as feature: Show users exactly what agent is doing
• Working backwards product development: Build working demo first, then scale
• Competition advantage: Speed and user focus over pure technology

Personal Notes from Dustin

• Asked question about geographic makeup of customer base
• Download the Manus app and start using
• Reach out to Tao - described as “freaking awesome”
• Tao’s learning approach: Jump in, read papers for 2-3 months despite non-CS background
• Want to “jump in and learn learn learn learn learn” following Tao’s example