Praying Manus

Praying Manus

Glimpse of AGI | X Thread 

TL;DR: Manus, a new general AI agent from China’s Butterfly Effect, blends multiple AI models to tackle diverse tasks autonomously, showing impressive adaptability despite occasional crashes and server hiccups.

Key Points:

• Launched by Wuhan-based startup Butterfly Effect, Manus combines models like Claude 3.5 Sonnet and Alibaba’s Qwen to act independently on tasks, distinguishing it from single-model chatbots like DeepSeek.

• It’s gained buzz globally, with tech influencers like Jack Dorsey praising its potential, though access remains limited—under 1% of waitlist users have invite codes.

• Testing revealed Manus excels at research-based tasks (e.g., compiling journalist lists or property searches) when given clear guidance, but struggles with paywalls, large-scale projects, and system stability.

• Unique “Manus’s Computer” feature lets users watch and tweak its process in real-time, offering transparency and collaboration not found in tools like ChatGPT DeepResearch.

• Priced at $2 per task it’s poised to appeal to professionals and small teams if server issues are resolved.

Why It Matters: Manus’s marks one huge step towards AGI. With a team of agents, “swarm” working together to complete a task. Unlike flashy, overhyped tools, Manus’s transparent process—showing its steps and even admitting laziness—builds trust, which could nudge businesses toward integrating AI agents into daily tasks.

That said, its crashes and limited scope (e.g., stalling on a 50-person Innovators Under 35 list) suggest it’s not yet a full replacement for human effort, especially on sprawling projects like award nominations or deep investigative work. At a tenth of DeepResearch’s cost, it could disrupt pricing norms, pushing competitors to lower barriers and forcing media outlets to rethink budgets for tech adoption.

If Butterfly Effect bolsters its infrastructure, Manus might not just follow Western AI trends but carve a niche where Chinese innovation dominates.