If you know me, you know I’ve been focused on how to make AI better at maintaining requirements and developing features. This workflow is the best practice that crystallized out of my latest round of practice.

Over the past two years I’ve written a lot with AI, from one-line scripts to entire systems. After stepping on enough landmines, I settled on a path I actually trust: don’t let AI write code directly — first have it “translate” the requirement into a tens-of-thousands-of-words design document, read it yourself, pass blind review, then touch code. It sounds roundabout, but it’s actually much faster.

The Core Proposition

Can 3-4 fast, cheap models working as a team outperform 1 expensive, slow model?

The answer is yes — and it’s not just intuition. A growing body of research from 2024–2025 backs this up. pi-multi-agent is the engineering realization of this idea: break a big task into pieces, assign each to an agent with a distinct perspective, let them debate, collide, and cross-review. The collective output consistently exceeds what any single model produces alone.

The Problem: AI Writes Code Too Fast

Vibe-Coding is changing the rhythm of software development. AI coding tools have accelerated code generation by 10x or even 100x. But this speed brings an obvious problem: code out of control.

The Sweet Trap of Vibe-Coding

The Vibe-Coding experience goes like this: You tell the AI — “Help me build an order system that supports placing orders, payments, and refunds. Payments use webhook callbacks, refunds need an approval workflow, and all operations must have audit logs.”

Five million tokens a day, I stroll with ease; fifty million tokens a day, I push with vigor; half a billion tokens a day, I’m drenched in sweat.

Background and Problem

The Context Cost of a Single Session

Every LLM call requires passing the full conversation history, meaning context grows linearly with the number of tool calls. After 50 tool calls in a complex task, the 51st call must carry the preceding 50 entries of history — not only is this costly, but the LLM also tends to lose focus and miss critical information in an overly long context.

Overview

Since the start of this year, I’ve been using AI intensively (Claude Code in particular) to help with many projects and write a lot of software. Over this time, my understanding of how a programmer should work with AI has gone through stage after stage. There’s a lot in my head, and I want to write it down to help myself reflect and synthesize.

Dimensions of the Journey

Several parallel threads have been evolving in my understanding of AI:

One-Line Summary

A complete requirement-driven development framework that turns the full software lifecycle — requirement analysis, technical design, coding, security audit, cleanup, compliance review, verification, and archival — into 8 executable skills for Claude Code, with checkpoint recovery and formal change management.

Why This Exists

The biggest problem with AI-generated code isn’t that it “can’t write code” — it’s that it writes without discipline:

• Starts coding before understanding requirements
• Makes a bunch of changes but can’t tell if they’re correct
• Interruptions mean starting over
• Security vulnerabilities go unchecked
• Requirements drift silently

This framework turns software development best practices into executable AI skills. Claude Code follows a structured process instead of winging it.