From Vibe Coding to Agentic Engineering
Engineer automated software factories and agent harnesses that direct autonomous coding agents to build, verify, and ship software.
- Software Engineers moving from interactive prompting to engineering automated software factories with autonomous coding agents
- Tech Leads & Architects designing specification standards, verification loops, and review gates for agent-built software
- Cloud, Data & Platform Engineers building the repository infrastructure, CI pipelines, and sandboxes that run autonomous coding agents
Guided take-home project: engineer an event-driven software factory using Devin or OpenHands that turns backlog issues into reviewed pull requests, triaging GitHub issues, dispatching autonomous coding agents for implementation, running independent code-review loops, and gating merges on severity.
Move from ad-hoc vibe coding to engineering repeatable systems that direct autonomous coding agents across the software lifecycle. Examine foundation model failure modes, comprehension debt, and how engineering focus shifts to specifications, verification loops, and factory design.
Examine coding-agent harness design (Agent = Model + Harness): control loops, context assembly and compaction, tool execution, sandboxing and permission policies, state persistence, and configuration surfaces (AGENTS.md, hooks, skills, plugins, and MCP).
Structure how engineering intent reaches autonomous coding agents through three repository artefacts: specifications with machine-checkable acceptance criteria, persistent project instructions, and maintained architecture knowledge bases.
Engineer bounded execution, verification, and recovery loops across autonomous coding-agent runs. Separate implementation and review roles, coordinate isolated worktrees, and wire repository events and CI checks into automated software factories.
Extend the coding-agent harness (filesystem, shell, code execution, skills, and MCP) beyond software repositories as a general-purpose runtime for knowledge work and business automation, covering agent SDKs, planning and memory mechanisms, personal assistants, and workflow orchestration.
- Working proficiency in Python or TypeScript (comfortable reading, writing, and debugging multi-file applications)
- Professional software engineering experience (mid-to-senior recommended) and a working grasp of system design
- Git and CI/CD fluency: command-line workflows, branching, pull requests, automated testing, and code review
- Access to a coding agent (such as Claude Code, Codex, or Cursor)