Vibe Engineering with AI Coding Agents
A project-based path through vibe coding, vibe engineering, and agentic engineering, given room to breathe across a full eight weeks.
Currently teaching2 instructional days per week, plus a weekly hands-on build session
What this course is
Vibe Engineering with AI Coding Agents takes the core ideas of agentic software development and gives them room to breathe across a full two months. Instead of rushing through tools in a single sitting, each stage gets dedicated time: more practice with every tool, more variations on every project, and full weeks set aside for topics a faster course would otherwise compress into a single afternoon.
The course follows a simple three-part arc. Learners begin with vibe coding, move into the more disciplined practice of vibe engineering, and finish with agentic engineering at an expert level. Each stage spans two to three weeks rather than a handful of days, with an added capstone week and a closing session on career paths and monetization.
Three stages, from vibe coding to a team of agents
Each stage gets two to three weeks, so nothing gets compressed into a single afternoon.
Vibe coding
Mental models, prompt and context engineering, and hands-on fluency with IDE-based agents such as Cursor, Copilot, Codex, and Antigravity.
Vibe engineering
The more disciplined practice: professional CLI workflows with Claude Code, MCP, Skills, Plugins, and real engineering-tool integrations.
Agentic engineering
Expert level: sub-agents, hooks, sandboxing, and multi-agent orchestration, culminating in a full team-of-agents capstone build.
Eight weeks, week by week
Two months, split into foundations and vibe coding, then vibe engineering and agentic mastery. Expand any week for the full breakdown.
Month One · Foundations and Vibe Coding
W1Welcome to the Agentic Landscape
Building the mental models everything else relies on. No project this week, the focus is purely conceptual so later weeks move faster.
Welcome to the Agentic Landscape
Building the mental models everything else relies on. No project this week, the focus is purely conceptual so later weeks move faster.
Orientation and the Eight Levels of AI Adoption
- A roadmap of how vibe coding, vibe engineering, and agentic engineering relate to one another
- The origin of vibe coding and how it grew into a professional discipline
- The eight levels of AI adoption, from basic prompting through autonomous multi-agent systems
- The three surfaces for AI coding tools: IDEs, plugins, and CLIs
How Large Language Models Actually Work
- Tokens, context windows, and memory, explained from the ground up
- Why agent output varies from run to run, and how to design around that variability
- Hands-on: generate a working app from a single prompt to feel the unpredictability firsthand
W2Prompt and Context Engineering
Giving agents the information they need to succeed.
Prompt and Context Engineering
Giving agents the information they need to succeed.
Context Engineering Fundamentals
- Writing effective prompts: specificity, constraints, and breaking work into small testable steps
- Introducing AGENTS.md and CLAUDE.md as project-level instructions an agent reads first
- Cross-checking agent answers across models to catch mistakes early
Workshop: Building Your AI Digital Twin
- Scaffold a personal website project with a clear AGENTS.md file
- Iterate with an agent through cycles of planning, generation, and feedback
- Review common failure modes and how targeted feedback resolves them
W3Cursor and YOLO Mode
Working quickly inside an AI-native IDE.
Cursor and YOLO Mode
Working quickly inside an AI-native IDE.
Cursor in Depth
- Cursor settings: autorun versus YOLO mode, and sandboxed versus unsandboxed execution
- Using planning mode and iterative feedback loops to steer an agent toward a working build
Workshop: Build the Kanban App
- Build a complete drag-and-drop Kanban app from scratch using Cursor in YOLO mode
- Debug and refine the codebase using specific, targeted feedback to the agent
W4Copilot, Codex, and Antigravity
Comparing agents across IDEs and providers.
Copilot, Codex, and Antigravity
Comparing agents across IDEs and providers.
Touring the Alternatives
- GitHub Copilot's agent mode compared with Cursor's workflow
- OpenAI Codex for autonomous task execution
- Google Antigravity: rules, workflows, and parallel agent tasks
Workshop: Ship a Commercial MVP
- Choose a tool and scope a real MVP idea with a clear specification
- Build, test, and polish the MVP end to end as a portfolio piece
- A Month One retrospective comparing strengths and weaknesses across all four tools
Month Two · Vibe Engineering and Agentic Mastery
W5Claude Code Fundamentals
Moving from IDE agents to professional CLI workflows.
Claude Code Fundamentals
Moving from IDE agents to professional CLI workflows.
Claude Code Core Mechanics
- Why CLI-based agentic workflows differ from IDE-based ones
- Slash commands, checkpoints, and session management
- Writing a precise CLAUDE.md for a real, multi-file project
Workshop: Scaffold the SaaS Legal Assistant
- Plan the legal assistant's architecture together with Claude Code
- Generate a first working version of document drafting and PDF export
W6MCP, Skills, Plugins, and Integrations
Extending what an agent can reach and do.
MCP, Skills, Plugins, and Integrations
Extending what an agent can reach and do.
Model Context Protocol and Skills
- MCP fundamentals: how agents connect to external tools and data
- Packaging reusable Skills and Plugins for specialized, repeatable agent behavior
Workshop: Jira and GitHub in the Loop
- Raise a Jira issue and have Claude Code pick it up, build it, test it, and push the change to GitHub
- Ralph Loops and other iterative delivery workflows for unattended progress
- Debugging strategies suited to non-deterministic, agent-driven workflows
W7Multi-Agent Systems and Sandboxing
Moving from a single agent to a coordinated team.
Multi-Agent Systems and Sandboxing
Moving from a single agent to a coordinated team.
Sub-Agents, Hooks, and Sandboxes
- Delegating work to sub-agents, and observing and controlling behavior with Hooks
- Sandboxing: running agents in YOLO mode safely, without risk to the host system
- Working with large, pre-existing codebases rather than greenfield projects
Workshop: Orchestration Frameworks
- Agent swarms and orchestration patterns, including Claude Agent Teams and Gas Town
- A look at the frontier: Claude Agent SDK, Cowork, and OpenClaw
- Design the multi-agent architecture for the capstone project
W8Capstone Build and Career Paths
Bringing everything together.
Capstone Build and Career Paths
Bringing everything together.
Workshop: Build the Trading Workstation
- Stand up the realtime data feed, virtual trade engine, and AI assistant using a multi-agent team
- Apply the sandboxing, hooks, and orchestration patterns from Week 7 to a real, complex build
Capstone Review and What Comes Next
- Present and review completed capstone projects
- Turning agentic coding projects into a portfolio, freelance offerings, or a SaaS product
- Where the field is heading, and how to keep pace with new tools after the course ends
A trading workstation run by a team of agents
A realtime trading workstation with virtual trades on live market data and an AI strategy assistant, built and operated by a coordinated team of AI agents.
- 1An AI digital twin personal website
- 2A drag-and-drop Kanban app
- 3A commercial-grade MVP
- 4A SaaS legal assistant with PDF export
- 5A realtime trading workstation
- Practical fluency across both IDE-based agents (Cursor, Copilot, Codex, Antigravity) and CLI-based agents (Claude Code, OpenCode, Amp)
- A working mental model of how language-model agents reason, use context, and vary between runs
- Five complete, portfolio-ready projects, from an AI digital twin site to a realtime trading workstation
- Production-level practices: MCP, Skills, Plugins, Hooks, sub-agents, sandboxing, and multi-agent orchestration
- The ability to scale from working solo with one agent to directing a coordinated team of agents on a single codebase
- A concrete plan for turning these skills into freelance work, a job-ready portfolio, or a SaaS product of your own
Everything is open
The full course, code, and projects live in the open. Browse the repo or the course website.
Let's build something that ships.
Hiring for a senior Data/AI role, or need a data platform that actually holds up in production? Let's talk.
or email me directly at muhammaduzairkhan329@gmail.com