CO/AI Subscribe
Monday · August 10, 2026 · Issue No. 953
Video

DeepAgent: NEW Super AI DESTROYS Manus & Genspark? 🤯

Watch on YouTube

DeepAgent: Is this new AI really better than Manis and Genspark?

In a recent review, tech blogger Julian Goldie tested DeepAgent from Abacus AI, a new contender in the autonomous AI agent market that's been marketed as a "god tier" alternative to popular options like Genspark and Manis – but at a more affordable price point of $10/month compared to Manis's $49 entry tier.

What is DeepAgent?

DeepAgent claims to automate a variety of tasks including:

  • Building websites and applications
  • Creating SaaS tools
  • Generating presentations and reports
  • Connecting with Gmail to manage emails

The test results: Not quite "god tier" yet

When put through real-world testing with identical prompts used across platforms, DeepAgent showed some concerning limitations:

The good

  • Successfully created a functioning Space Invaders game on first attempt
  • Generates dedicated subdomain deployments
  • Provides detailed thinking process and research capabilities
  • Pricing is competitive at $10/month

The significant limitations

  • Limited to only 2 tasks per day
  • Cannot run multiple tasks simultaneously (unlike Genspark)
  • Tasks have compute point limits that can expire before completion
  • Some advertised features (like Gmail integration) were difficult to locate

Head-to-head comparison

When tasked with creating a website about Julian's SEO agency and a Space Invaders game, Genspark consistently outperformed DeepAgent:

  1. Website creation test:

    • Genspark: Created a complete, modern website with actual details from Julian's business
    • DeepAgent: Created an extensive plan but ran out of compute points before completion
    • Manis: Created a functional but generic website with stock images
  2. Game creation test:

    • Genspark: Created a more sophisticated, almost "PlayStation level" game
    • DeepAgent: Created a basic but functioning Space Invaders game
    • Winner: Genspark by a significant margin

Current rankings according to the test

  1. Genspark: Best overall performance, unlimited simultaneous tasks
  2. Manis: Less restricted than DeepAgent but not
Share: X LinkedIn Email
Video Feed

More videos

All videos →
Poolside’s Eiso Kant on Why Small Models Can Do Big Things (And Why MCP Is “Stupid”)
Video

Poolside’s Eiso Kant on Why Small Models Can Do Big Things (And Why MCP Is “Stupid”)

Poolside released Laguna S, a 118B parameter model (8B active) trained in 8 weeks that outperforms models 2-3x...

Garry Tan Says He’s 400x More Productive Now. Here’s the ‘Brain’ Behind It.
Video

Garry Tan Says He’s 400x More Productive Now. Here’s the ‘Brain’ Behind It.

Garry Tan’s real argument is about rebuilding how a company works, from the org chart down. Here’s his...

Andrej Karpathy – Bigger Models Aren’t the Answer
Video

Andrej Karpathy – Bigger Models Aren’t the Answer

Andrej Karpathy makes the case that small models, the right tools, and tight feedback loops matter more than...

CONSULTING

Outsider
Labs.

A management consulting team focused on AI transformations for executives and business owners.

Work with us →