LIVE: Vibe Coding with Claude Fable 5!

Introduction to Fable 5 and Hermes Update

Overview of the Session

• The session begins with a brief introduction, indicating that the focus will be on testing Fable 5 and discussing the latest update from Hermes (v 0.18) .

• The host mentions plans to explore how to use Fable 5 for AI SEO and compare it with other agents like Mixer and Fusion .

Initial Observations

• Users have reported issues with debugging in Fable 5, noting that it sometimes redirects to Opus, which can be frustrating for developers .

• Concerns are raised about token limits in Fable 5, suggesting that users may run out of credits quickly depending on their plan .

Agent Operating Systems Discussion

Features of Agent Operating Systems

• The host describes an agent operating system setup where various agents work together seamlessly, including new features like a music agent and video agent being integrated into the system .

• A question arises regarding what people are using Fable 5 for; initial feedback indicates limited usage due to rapid depletion of tokens after just a few prompts .

Performance Testing

• The session includes running benchmark tests (Goldie Bench) for Fable 5 against other models like Fusion and Mixer agents, aiming to see if it can outperform them in practical applications .

• Users report frequent bugs with Claude's service being busy during attempts to access Opus versions, highlighting potential limitations in availability or performance at this time .

Hermes Desktop vs. Agent Operating System

User Preferences

• A user requests a masterclass on Hermes desktop; however, the host prefers using the agent operating system due to its superior customization options compared to Hermes desktop .

• The discussion emphasizes that while both systems have their merits, the agent operating system allows for more tailored setups based on individual needs .

Customization Examples

• An example is provided where a member has created an outreach tool within their customized agent operating system that analyzes job opportunities and forwards them efficiently .

New Features in Hermes v0.18

Mixture of Agents

• One significant update highlighted is the "mixture of agents," allowing users to combine multiple models (e.g., Claude, Opus 4.8, GPT 5.5) for improved output quality by leveraging diverse strengths from each model .

Enhanced Functionality

• New functionalities include goal-oriented task loops where outputs are judged automatically until they meet quality standards without requiring user intervention .

Learning Capabilities

• A feature called forward/learn enables Hermes to learn from external guides or resources quickly by saving learned information as skills for future reference .

This structured approach provides clarity on key discussions surrounding Fable 5's capabilities alongside updates from Hermes while ensuring easy navigation through timestamps linked directly back to specific moments in the transcript.

How Does the Delegate Tasks Feature Work?

Overview of Delegate Tasks

• The delegate tasks feature allows users to run complex tasks, such as SEO funnels, by automatically creating multiple sub-agents that work collaboratively.

• This system enables parallel processing; for instance, one keyword can generate five articles simultaneously without further user input.

• Use cases include SEO, video production, content creation, and website development—essentially any task requiring a team of agents.

Efficiency and Cost-effectiveness

• The new updates have made the system faster and more efficient while also being cost-effective compared to previous versions.

• Users can track their progress through a comprehensive journey breakdown provided by the system.

Updates on Hermes Agent System

Proof of Work Feature

• The latest update (v 0.18) introduces a proof of work feature that records evidence for coding tasks by running project checks instead of merely claiming completion.

• A standing goal loop continuously evaluates work against established evidence rather than stopping prematurely when an agent claims to be done.

Features and Tools Available

• Hermes integrates various tools including voice-powered agents, news updates via Hermes Oracle, automatic content generation for social media and websites.

• An outreach tool is available for lead generation through automated email campaigns written by AI.

Community Support and Training Resources

Accessing Training Materials

• Comprehensive training resources are available in the profit boardroom, including tutorials on mixture of agents and lead generation with Hermes.

• Community support is robust with 3,900 members providing assistance around the clock; personal responses from trainers are common.

Learning Pathways

• A structured learning path exists for beginners to advance to expert levels within six weeks through guided coaching calls and community interaction.

Using Fable 5 for AI SEO

Performance Insights

• Fable 5 has shown significant potential in building tools for AI SEO but is not recommended for direct content creation due to token limitations.

• It excels at automating processes but runs out of tokens quickly; thus it’s better suited for backend operations rather than front-end content generation.

Content Creation Strategy

• Combining Fable 5's capabilities with other models like Claude or Hermes enhances unique content generation while managing resource constraints effectively.

Benchmarking Different Models

Comparative Analysis: Fable 5 vs. Others

• Side-by-side tests were conducted comparing outputs from Fable 5, Hermes Mixture of Agents, and Open Rooters Fusion across various applications.

Game Development Outputs

• In game output tests, Fable 5 produced superior results compared to both Fusion and Hermes Mixture of Agents in terms of detail and navigation ease.

UI Design Tests

• For landing page design tests, Fusion outperformed Fable 5 significantly; however, Hermes Mixture produced visually appealing outputs despite minor alignment issues.

Conclusion on Model Performance

• The collaborative nature of Hermes Mixture may enhance output quality due to its panel approach that critiques individual contributions before finalizing results.

Overview of Operating Systems and Performance Tests

Comparison of Operating Systems

• The operating systems being compared are similar, with the only notable difference being a more appealing introduction in one version. Overall, they are considered to be on par with each other.

Testing Web OS Applications

• A web operating system created for Fable 5 is introduced, integrating various applications into a single system.

• Initial tests reveal that the notes app in Fusion does not auto-save, despite its attractive interface. Users cannot access the save function effectively.

• The terminal and calculator functions in Fusion show limited responsiveness; navigation is challenging.

Performance Insights from Hermes Mixture of Agents

• The Hermes version exhibits some bugs, such as erratic movement of the notes application. However, it has a visually appealing UI but lacks responsiveness.

Evaluation of Fable 5's Functionality

• Fable 5 performs well across various tests, including paint and terminal functionalities. It is deemed superior to both Hermes Mixture of Agents and Fusion based on performance metrics.

Vauil Craft Testing Results

• The Vauil Craft version from Hermes Mixture of Agents is highly buggy and difficult to navigate.

• In contrast, Fusion’s version works well overall but has minor bugs like flying through floors during gameplay.

Game Development Comparisons

• Fable 5's game output is slightly less impressive than Fusion's during testing phases.

• Frost Fell shows varying performance across versions; while Fusion's output looks good, Hermes' feels smoother and less buggy.

Analysis of Doom Version Outputs

• Both Mixture of Agents and Fusion provide enjoyable gameplay experiences; however, Fable 5 struggles with basic movement controls.

Summary of Test Results

• After multiple rounds of testing:

• Fable 5 wins three tests,

• Mixture of Agents also wins three,

• Fusion does not win any tests.

Recommendations for Using AI Models

Subscription Considerations for Claude Users

• Users on Claude’s subscription plan may quickly exhaust tokens when using Fable 5; thus, having Hermes Mixture of Agents as a backup could be beneficial.

Workflow Efficiency Tips

• Utilizing Fable 5 allows easy back-and-forth interactions via CLI within an organized workspace setup.

Future Developments in AI Models

Upcoming Features from Hermes Bench

• Anticipation builds around upcoming features from Hermes Bench which will test model combinations against benchmarks to determine optimal performance setups.

Community Engagement and Resources

Accessing AI Profit Boardroom Resources

• Comprehensive resources available include SEO agents, video agents, memory systems integrated into the AI profit boardroom platform.

Importance of Personal Testing Over Benchmarks

• Emphasis placed on personal experimentation over relying solely on benchmarks for evaluating model effectiveness.

Daily Updates and Support Structure

Continuous Learning Opportunities

• Regular updates provided through community engagement include tutorials on new features like Mixer of Agents along with step-by-step guides for effective usage.

Building Projects with Fable 5

Showcasing Recent Creations

• A showcase highlights various projects built using Fable 5 demonstrating its capabilities in creating solid outputs efficiently.

Best Practices for Using Fable 5

• Recommendations suggest utilizing Claude to select models effectively while noting that access may change post-July deadline due to subscription limitations.

Limitations Noted During Usage

• Observations indicate that while powerful at coding tasks, Fable 5 struggles with UI design quality compared to other models like Opus 4.8 which adheres better to guidelines set by users.

Agent Orchestration and Collaboration

Systems for Agent Communication

• The discussion begins with the importance of orchestrating agents to facilitate communication and collaboration among them, highlighting a group chat feature in the agent mastermind section.

• Agents can read from an Obsidian vault, allowing them to share ideas and quickly implement successful concepts into a pipeline for creation.

• A simple challenge can prompt agents to generate ideas autonomously, exemplified by asking for suggestions on an SEO app.

Building Teams of Agents

• The speaker emphasizes the uniqueness of their system compared to others like Hermes, which limits customization and collaboration among different agents.

• Using Hermes restricts users to its ecosystem without the flexibility needed for building diverse teams of agents tailored to specific tasks.

• The custom-built system allows full control over how agents work together, enhancing productivity and adaptability.

Progress Tracking and Flexibility

• A dashboard provides visibility into all projects completed by the team of agents, including an inbox feature for direct communication with the user.

• Users can easily track progress over time, referencing past creations and their timelines for better project management.

• New features like Ultra Code enable teams of sub-agents (e.g., 12 SEO sub-agents) to collaborate effectively within separate tabs.

Model Management and Workflow Efficiency

• The system's design allows seamless swapping between different models (e.g., Fable 5 or Opus 4.8), ensuring workflows remain uninterrupted even if one model becomes unavailable.

• This flexibility minimizes downtime; users can replace models in seconds rather than spending extensive time rebuilding systems after a change.

Automation and Self-Sustaining Processes

• An orchestration system enables collaborative efforts among agents on various projects, such as video generation using Hermes agents.

• Recent innovations allow users to create educational videos effortlessly through newly integrated systems that enhance content production capabilities.

Autonomous Agent Operations

Loop Engineering Concepts

• The speaker discusses transitioning from manually prompting agents to enabling them to self-prompt on loops, improving efficiency without constant oversight.

• Systems are designed so that they refresh automatically (e.g., every 24 hours), reducing the need for human intervention in quality control processes.

Memory System Integration

• A memory system is established where thousands of notes help AI agents provide personalized responses based on historical data stored within the system.

• Agents utilize this memory framework not only for generating new content but also for updating existing information dynamically.

Community Engagement and Resources

Supportive Community Framework

• Testimonials highlight positive experiences from community members who have successfully utilized the agent operating system despite varying technical skills.

• Continuous sharing of wins fosters a collaborative environment where members learn from each other’s successes with AI tools.

Educational Opportunities

• Members have access to daily tutorials addressing common questions about using the agent operating system effectively.

• Weekly coaching calls offer personalized guidance while connecting individuals working with AI technologies locally.