Orchestrating AI Coding Assistants with Actor-Based Specifications

Most developers struggle with AI coding assistants because they jump straight to code generation without proper specifications. The result? Inconsistent architecture, unpredictable outputs, and endless debugging.

In this workshop, you'll learn a systematic multi-agent approach: first use architecture agents to create actor-based system models and specifications, then deploy language-specific implementation agents to generate code that matches your design. This specs-first approach to vibe coding gives you the benefits of AI acceleration while maintaining architectural control.

We'll work with real examples, showing how to go from business requirements to visual system models to working code - all within a 2.5 hour session.

What attendees will learn

• Actor-Based System Modeling: Design systems with Actors, Actions, and Conceptual Entities
• Modeling Systems with AI: Use AI agents to create and validate system specifications
• Multi-Agent AI Development: Orchestrate specialized agents for architecture, implementation, and testing
• Keeping AI Agents on the Leash: Control and guide AI output through precise specifications
• Specs-Driven Workflow: Transform vague requirements into implementable blueprints
• Quality Through Structure: Ensure consistent, maintainable code via systematic modeling

Target audience

• Developers who want to use AI assistants more effectively
• Team Leads looking to standardize AI-assisted development
• Architects interested in maintaining design control with AI
• Anyone frustrated with inconsistent AI code generation

Perfect for developers who want to use AI tools professionally without sacrificing code quality or architectural integrity. Ideal participants have programming experience and want to learn how to leverage AI without losing control over their architecture.

Requirements for the workshop

Technical Requirements:

Attendees should come prepared with:

• Laptop with Internet access
• Claude Code installed (we'll provide API keys)
• VS Code installed (for consistency, we can't guarantee other IDEs will work during the workshop)
• Git and GitHub account

Knowledge Prerequisites:

Attendees should have:

• Intermediate programming experience (workshop examples use Java, but you're welcome to try your preferred language, we'll support you as best as we can)
• No AI/ML experience required
• No previous experience with AI-assistant coding is needed

Companies using this tech

Specs-driven AI development:

Many companies like Thoughtworks are already adopting structured approaches to AI coding, moving from ad-hoc prompting to systematic specifications with specialized agents.

AgentGuild platform:

Currently in beta with ~50 testers from startups to enterprises (specific names not yet public).

Agenda

• Introduction to Specs-Driven AI Development
• Actor-Based System Modeling (MOSY Framework)
• Setting Up AI Coding Assistants and Architecture Agents
• Hands-On: From Requirements to Working Code
• Advanced Patterns and Best Practices
• Q&A

Key Takeaways

Participants will leave with:

• Multi-Agent AI Skills: Orchestrating specialized AI agents for different development phases
• Actor-Based Modeling: Creating clear system specifications before coding
• Practical Control: Keeping AI agents focused and productive through structured specs
• Continued Learning: Resources and templates for applying these techniques