Vibe coding and AI coding agents have made it possible to turn an idea into working code faster than ever. Developers can ask an agent to explore a codebase, implement a feature, write tests, fix issues, and even create a merge request. But when code becomes easier to produce, code review becomes the new bottleneck.

In this talk, we will look at a practical workflow for working with AI coding agents without wasting tokens, time, or engineering attention. We will explore how to provide focused context, break work into smaller tasks, reduce unnecessary token usage, and use AI effectively for implementation and repetitive development work.

The second part focuses on code review. When multiple merge requests arrive quickly, reviewing every line with the same level of attention does not scale. We will look at how to prioritize reviews based on risk, complexity, and impact, use automation and AI for the first layer of checks, and focus human review on architecture, business logic, security, concurrency, performance, and edge cases.

The goal is not simply to generate code faster. It is to let AI help with coding while engineers focus on what requires human thinking, review, and decision-making.

Key Takeaways

• Use vibe coding and AI agents effectively without blindly accepting generated code.

• Save tokens and cost by giving agents focused context and breaking large tasks into smaller steps.

• Use AI and automation to handle repetitive implementation and first level code checks.

• Prioritize merge requests based on risk, complexity, and impact instead of reviewing everything equally.

• Focus human review on architecture, business logic, security, performance, concurrency, and edge cases.