A developer's reference

Working With AI
as a Developer

Field notes on shipping with AI — models, workflows, and tooling. What actually works in production, kept up to date as the space moves.

Over the past year I've been using AI inside real projects — specs, tickets, production code. What started as curiosity turned into a set of habits I now run every day. This site is the distilled version of what I've learned: which models for which tasks, how to build workflows that don't spiral, and how to configure the tooling without fighting it.

It's not theory. Every pattern here has been tested inside a real codebase, on real tickets, under real deadlines. Each chapter includes an interactive tool so you can apply it immediately.

Part I

Understanding the Models

Reasoning vs Fast ModelsWhen to choose reasoning models vs fast models for different coding tasks

Model PersonalitiesHow different models approach the same task, and when to lean into that

Prompting and PitfallsThe mindset shift that makes AI coding actually work

Part II

Shipping Workflows

Spec to PRThe full AI-assisted development loop, from spec file to merged pull request

Bug Report to FixA repeatable AI-assisted debugging flow, centered on minimal reproduction

Design Token to StorybookFrom Figma tokens to production component to documented Storybook stories

Jira Ticket to Cursor to PRTurning a Jira ticket into a structured Cursor task with Figma design context

AI-Powered Code ReviewUsing BugBot and Codex as PR reviewers - what they catch and what they miss

Spec FilesAI-readable documentation that survives context window resets

Building Blocks FirstWhy starting from atoms changes how AI writes code

Part III

Review & Quality

Diff Review LoopsAvoiding the looks-fine trap in AI-generated code

Agent GuardrailsScope control and failure modes for autonomous coding agents

Part IV

Tools & Infrastructure

Cursor Max ModeWhen the 1M token context window earns its cost

Agents and SkillsHow reusable skill files encode domain knowledge beyond a single session

Figma MCP WorkflowsDesign context in the coding loop - from interpretation to direct translation

Code to CanvasPushing live rendered UI back into Figma as editable frames

Design to Code and BackThe full design and code feedback loop with Figma MCP

On My RadarWhat is moving fast in agent tooling right now