Topic hub
AI Coding Workflows
Step-by-step workflows for using AI to plan features, write code, debug issues, refactor systems, generate tests, review pull requests, document projects, migrate codebases, and ship software.
Featured guides
How the OpenAI Codex Team Builds Codex With 10-Bullet Specs
How the OpenAI Codex team builds Codex with 10-bullet specs, eight-week plans, agent-generated code, a lean PM function, and relentless user feedback.
Comet 0.4 Guide: Classic Spec Workflows, /comet-any and Skill Evaluation
A current guide to Comet’s Classic Spec workflow, /comet-any Skill composition and eval platform, with a clear boundary between 0.3.9 and the 0.4 beta.
How to Prototype an AI-Assisted 3D Fight Scene with Blender MCP, Fable 5 and Three.js
A production-minded guide to turning generated textures, image-to-3D meshes and captured motion into a browser fight-scene prototype—and the limits demos hide.
Related topics
All topic guides
How the OpenAI Codex Team Builds Codex With 10-Bullet Specs
How the OpenAI Codex team builds Codex with 10-bullet specs, eight-week plans, agent-generated code, a lean PM function, and relentless user feedback.
Comet 0.4 Guide: Classic Spec Workflows, /comet-any and Skill Evaluation
A current guide to Comet’s Classic Spec workflow, /comet-any Skill composition and eval platform, with a clear boundary between 0.3.9 and the 0.4 beta.
How to Prototype an AI-Assisted 3D Fight Scene with Blender MCP, Fable 5 and Three.js
A production-minded guide to turning generated textures, image-to-3D meshes and captured motion into a browser fight-scene prototype—and the limits demos hide.
Claude Dynamic Workflows on Codex: What This Community Project Actually Does
A fact-checked look at how a community project routes Claude Code workflows through Codex app-server, where it helps, and what its benchmark proves.
Who Gets to Decide? Designing an Authority Control Plane for AI Coding Agents
A research-backed operating model for AI coding agents: authority tiers, least privilege, parameter-bound approvals, sandboxing, rollback, audit trails, and outcome-based evaluation.
Superpowers vs. Grill Me: Match AI Planning to the Cost of Being Wrong
Research shows that clarification, planning, and verification solve different AI coding risks. Compare Superpowers, Grill Me, and a practical hybrid workflow.