
Learn to balance vibe coding ideas with disciplined agentic engineering, emphasizing safety, ownership, and fundamentals for orchestrating agents without manual coding.
Explore AI coding tools like Cloud Code by Anthropic, Codex, and OpenCode, and learn terminal setup across macOS, Linux, and Windows, plus Cursor and OpenClaw options.
Demonstrate how to delegate sub agents, manage context windows, and use claude.md with XML prompts to optimize token use and memory in cloud environments.
Explore setting up local agent instructions with CLAUDE.md to keep project data private while building a to-do list app, and compare cloud prompts with project-local agents.md.
Master agent delegation and testing best practices for artificial intelligence projects, including using UV for fast Python package management, enforcing tests and linting, and applying absolute and relative tolerances.
Master how to use spec.md as the project north star, benchmark precisely, and accelerate your workflow with voice typing to minimize guessing and maximize accuracy.
Explore voice typing tools like Super Whisper, Just Speak, Whisper Flow, and AquaVoice, emphasizing offline transcription, screen-context OCR, and fast local processing for fluent transcription.
Explore how Claude Code's harness layers tools, memory, and prompts to power long-running agents, enabling task execution across local context, terminal commands, and MCPs.
Profile Minecraft 1.21.11 on a MacBook M4 to identify hotspots and FPS bottlenecks, then automate optimization with a Fabric mod and benchmark workflow to patch performance.
See AI agents autonomously run benchmarks, optimize hardware-specific kernels, and integrate MCPs within Cloud Code, exploring Context 7, NVIDIA Cutlass, and plugin workflows for performance tuning.
Profile the bottlenecks to reveal driver overhead from OpenGL to Metal; isolate GPU kernels, test with baseline inputs, compare outputs to ensure correct pixels, and implement faster raw Metal calls.
Update the spec and CLAUDE.md to reflect progress and prevent data loss before compaction, using the edit tool to add a new header and finalize path B.
Achieve a 6.3x performance boost by moving entity processing to the gpu, dramatically reducing lag compared to cpu, and integrate the final optimization to launch the game.
Move beyond basic vibe coding and step into the era of Disciplined Agentic Engineering. This course is a comprehensive guide to mastering the next generation of AI development tools, including Claude Code, Codex, OpenCode, and Cursor. Across 7 sections and 17 focused lectures, you will learn how to transition from simple prompting to managing autonomous AI agents that can handle complex, real-world tasks.
What You Will Master:
The Agentic Workflow: Learn the difference between standard AI coding and disciplined engineering using CLAUDE md, context windows, and sub-agent delegation.
Advanced Project Documentation: Set up local and global project instructions to maintain code quality and testing standards across your entire workflow.
Next Gen Efficiency: Supercharge your output using Spec md combined with professional voice typing workflows to speak your code into existence.
Deep Tech Mechanics: Understand how the Claude Code Agent Harness works, utilizing memory, tools, and the Model Context Protocol (MCP).
Hands-On Project:
The course culminates in a massive, real-world project where you will use AI agents to optimize Minecraft performance. You will witness a live demo of agents running autonomous benchmarks and writing GPU rendering code to achieve a 6.3x performance boost.
Whether you are looking to automate bottlenecks or build fully autonomous systems, this course provides the blueprint for the modern AI-augmented engineer.