Presentation Material for VanJS talk on Vibe Code Like An Engineer

Presentation Material for VanJS Talk: Vibe Code Like an Engineer

This page gathers everything related to the talk, including the slide deck and deeper reference material for the workflows demonstrated live.

Slides

Use the following link to view the presentation: Google Slides.

Supplementary Material

The talk introduces a practical approach to AI‑augmented engineering. You can enhance your productivity by parallelizing your work using work trees and stacked PRs.

Git worktrees

Git worktrees let you check out multiple branches of the same repository into separate directories so you can work on features in parallel without stashing or switching branches. They give you isolated workspaces that share the same git history, which makes context switching cleaner, feature development safer, and code reviews easier. Independent features shine in worktrees because each workspace stays focused and conflict-free, while dependent features work well when you use contract-first development: define interfaces and mocks upfront so both sides can progress without blocking each other. Worktrees fit naturally into professional workflows since they support parallel development, clean isolation, smooth PR reviews, and predictable integration paths.

Stacked PRs

Stacked PRs let you split a large, interdependent feature into a series of small pull requests that build on one another, so each piece stays focused and reviewable while the whole feature moves forward in parallel. Instead of waiting for one PR to merge before starting the next, you create a chain of branches where each branch depends on the previous one. Reviewers can look at all PRs simultaneously, and you merge them sequentially once approved. This approach works well for multi-week features, complex systems with layered dependencies, and teams that want cleaner histories and faster feedback. Tools like git-town or Graphite automate stacking, syncing with main, rebasing dependent branches, and shipping merged PRs. The workflow starts with creating the foundation PR, then appending new branches for backend, frontend, or other layers, each with its own PR based on the one before it. When main updates, you sync the whole stack so every PR stays up to date. When a PR is approved, you ship it, which rebases the rest of the stack onto main and keeps everything clean. Stacked PRs also power parallel agent workflows: one agent builds the schema, another builds the backend on top of it, another builds the UI, and all PRs go up together and merge in order. They speed up development, maintain clear single-responsibility changes, keep reviews small and focused, and avoid blocking progress across dependent parts of a feature.

Link for the Github repo for KAI

KAI (Knowledge Augmented Intelligence) is an open‑source meta‑framework focused on practical AI‑augmented engineering. The repository is available here: github.com/MAnfal/kai.

This project is in progress. The repository currently contains a placeholder README, and code will be added as testing completes. If you find the project valuable, consider starring the repo to show your support!