← Articles
CJ

Charlin Joe

Author

Articles by Charlin Joe(13)

The Codebase Your Successor Will Thank You For

Step 10 closes the loop: amend the ADRs, update the metrics, write the close-out note. Why the final 45 minutes of a feature determine whether the next one is harder or easier

ai-sdlcclose-outdocumentationcodebase-health

25 July 2026

What "Done" Actually Means

83% test coverage isn't done. Step 6 verification is a structured checklist — requirements, ADR consequences, GDPR constraints — not a feeling

ai-sdlcverificationdefinition-of-donequality-gates

25 July 2026

Three Gates. Three Different Failures

One review gate catches one class of failures. The ANN SDLC uses three gates — each one designed to catch what the other two structurally cannot

ai-sdlcquality-gatescode-reviewprocess

25 July 2026

The Plan Is the Contract

Why bounded context turns an 847-line PR into a 298-line PR — and how the implementation plan creates the boundary that keeps AI generation predictable

ai-sdlcimplementation-planbounded-contextcode-generation

25 July 2026

Why ADRs Are the Most Important File Nobody Writes

The cost of writing an ADR is 45 minutes. The cost of not writing one is three days of production investigation, six months later. Here's the math

ai-sdlcarchitectural-decisionsadrdocumentation

25 July 2026

The Acceptance Criteria That Saved the Sprint

How five verifiable acceptance criteria constrain what the AI builds — and why a single 15-second criterion prevented a two-day bug investigation

ai-sdlcrequirementsacceptance-criteriagdpr

25 July 2026

Write the Problem Before You Write the Code

The business intent document is the foundation everything else builds on. What goes in it, why the 'so that' clause matters, and what it costs to skip it

ai-sdlcrequirementsplanningbusiness-intent

24 July 2026

The Test Passed. The Feature Was Wrong

CI green doesn't mean the feature is correct. Why AI self-testing creates a blind spot that only human verification at Step 6 can close

ai-sdlctestingverificationquality-gates

24 July 2026

The Demo Worked. The Codebase Didn't

Why AI-generated codebases look perfect at demo and collapse at handoff — and the 14 documentation gaps that cause every preventable failure

ai-sdlcdocumentationonboardingcodebase-health

24 July 2026

The AI Made a Decision Nobody Asked It To

How an AI's silent backward-compatibility decision turned into a 40-hour production investigation — and why ADRs are the only reliable defense

ai-sdlcarchitectural-decisionsadrincident-analysis

24 July 2026

The Commit Is a Paper Trail

Every commit is a message to a future engineer debugging your code at 2am. How writing commit messages from the implementation plan makes that message useful

ai-sdlcgitcommit-messagespull-requests

23 July 2026

You Can't Review What You Can't Read

A 623-line PR with no context is unreviable. How writing PR descriptions from the implementation plan — not the diff — makes every review tractable

ai-sdlccode-reviewpull-requestsdocumentation

23 July 2026