Why the Best AI Agents Are Built Without Frameworks (Primitives over Frameworks)
summary
Claims that frameworks slow iteration and primitives win, backed by concrete results: thousands of agents scaled to 100 million monthly runs, eight patterns described as eight agent architectures, and live examples built from memory primitive, parser, chunker, tool integrations, orchestrators, routers, evaluators, and serverless deployments. Code-level approaches use plain JavaScript/TypeScript, call specific LLMs and tools (e g Gemini llama 70B GPT variants, Mistl OCR), and demonstrate what was actually built and how reliably it runs.
AI-written summary · about
chapters (11)
- 0:00Introduction to AI Agents
- 1:36Why Use AI Primitives
- 4:56Building a PDF Chatbot
- 6:12Agent Architectures: Augmented
- 12:56Prompt Chaining and Composition
- 14:01Agent Routing Technique
- 16:45Parallel Agent Execution
- 17:32Orchestrator-Worker Model
- 20:09Evaluator-Optimizer Pattern
- 21:44Real-world Agent Examples
- 25:06Conclusion and Resources
related talks
40mBuilding Effective AI Agents Simple, Composable Patterns
27mEmerging Patterns Agentic AI
32mProgramming language evolution in the AI era
41mThree agents, three frameworks, one talk
24mAI Agents in Production: Why the Platform Matters
19mDon't Build Slop (4 Levels of AI Agent Maturity)
31mAgentic AI architecture patterns
34mProgramming Languages for AI Programing Agents