CREATE SOMETHING: Systematic Evaluation of AI-Native Development

Micah Johnson

PAPER-2025-001

Subtractive Triad Audit: Kickstand

Applying the Subtractive Triad framework (DRY → Rams → Heidegger) to audit a production venue intelligence system—demonstrating that creation is the discipline of removing what obscures.

Case Study • 12 min read • Intermediate

Abstract

This paper documents the application of the Subtractive Triad framework to Kickstand, a venue intelligence automation system serving Half Dozen (a CREATE SOMETHING client). The system had evolved through multiple architectural phases (Node.js → Railway → Cloudflare Workers), accumulating significant technical debt. Through systematic application of three subtractive disciplines—DRY (Unify), Rams (Remove), and Heidegger (Reconnect)—we achieved: 92% reduction in active scripts (155 → 13), 100% reduction in TypeScript errors (30 → 0), and 48% improvement in overall health score (6.2 → 9.2). The case study validates the Subtractive Triad as an effective framework for production system audits.

9.2

Final Health Score

-92%

Active Scripts

TypeScript Errors

153

Scripts Archived

1. The Subtractive Triad Framework

Meta-principle: Creation is the discipline of removing what obscures.

The Subtractive Triad provides three lenses for evaluating any codebase, each operating at a different level of abstraction:

Level	Discipline	Question	Action
Implementation	DRY	"Have I built this before?"	Unify
Artifact	Rams	"Does this earn its existence?"	Remove
System	Heidegger	"Does this serve the whole?"	Reconnect

The triad is coherent because it's one principle—subtractive revelation—applied at three scales. Truth emerges through disciplined removal at every level of abstraction.

2. System Context: Kickstand

Kickstand is a venue intelligence automation system that monitors music venues' social media and websites to extract artist performance data. It serves Half Dozen, which is a client of CREATE SOMETHING.

The system produces daily intelligence reports, artist extractions, and venue monitoring data. It had evolved through multiple deployment phases:

Phase 1: Node.js + local development
Phase 2: Railway deployment
Phase 3: Cloudflare Workers (current production)

Each migration left artifacts behind, creating the debt that this audit addresses.

3. Level 1: DRY (Implementation) — Unify

Question: "Have I built this before?"

Score: 5/10 — Critical duplication found

Finding: Parallel Runtime Implementations

The codebase maintained two complete implementations of core services—one in Node.js and one in Cloudflare Workers TypeScript:

Component	Node.js	Workers	Duplicated
Artist Extractor	1,594 lines	644 lines	~500 lines
Website Monitor	~800 lines	~600 lines	~400 lines
Instagram Monitor	~900 lines	~700 lines	~600 lines
Total			~3,200 lines

Action: Unified to Workers

Marked Node.js services with @deprecated notices
Fixed 30 TypeScript errors in Workers implementation
Updated Cloudflare Workflow API usage (event.payload not event.params)
Added proper type annotations throughout

4. Level 2: Rams (Artifact) — Remove

Question: "Does this earn its existence?"

Score: 6/10 — Significant excess found

Finding: Script Proliferation

155 JavaScript files in the scripts directory, with only ~20 actively needed:

Before

• 155 scripts total
• 35 explicitly archived
• ~70 likely obsolete
• ~30 one-time migrations
• ~20 actively needed

After

• 13 scripts active
• 153 scripts archived
• Organized into categories:
- migrations/ (38 scripts)
- tests/ (24 scripts)
- one-time/ (19 scripts)

Action: Archived Obsolete Artifacts

Moved 153 scripts to organized archive directories
Archived Railway configuration to config/archive/
Moved Railway docs to docs/archive/railway/
Created archive README documenting restoration process

5. Level 3: Heidegger (System) — Reconnect

Question: "Does this serve the whole?"

Score: 7/10 — Minor disconnection found

Finding: Architectural Confusion

The README described three different deployment targets, creating systemic incoherence:

// README claimed:

1. Node.js + Railway (documented as primary)

2. n8n (mentioned as future target)

3. Cloudflare Workers (actual production)

Additionally, the relationship between Kickstand → Half Dozen → CREATE SOMETHING was undocumented within the system itself.

Action: Unified Documentation

Rewrote README.md for Cloudflare Workers architecture
Created docs/ARCHITECTURE.md documenting system context
Added services/LEGACY.md deprecation guide
Added monitoring/LEGACY.md deprecation guide
Documented the hermeneutic circle: how Kickstand fits into the larger system

6. Results

Metric	Before	After	Change
Health Score	6.2/10	9.2/10	+48%
Active Scripts	155	13	-92%
Archived Scripts	35	153	+337%
TypeScript Errors	~30	0	-100%
Build Status	Failing	Passing	Fixed
Documentation	Outdated	Current	Unified

TypeScript Fixes Applied

Key fixes to achieve zero build errors:

Added warn() method to Logger class
Fixed Cloudflare Workflow API (event.payload not event.params)
Fixed Logger.error() call signatures throughout
Added type assertions for API responses
Updated MonitorResult interface
Removed @types/node from tsconfig to resolve conflicts

7. Conclusion

This case study validates the Subtractive Triad as an effective framework for production system audits. The three levels complement each other:

DRY catches mechanical duplication (parallel implementations)
Rams catches functional obsolescence (155 → 13 scripts)
Heidegger catches systemic disconnection (documentation drift)

Kickstand is now a functional and coherent system. Its core value proposition—venue intelligence through automated monitoring and artist extraction—works well, and the codebase now reflects this clarity.

The Subtractive Path Forward

Unify

the implementation

Remove

the obsolete artifacts

Reconnect

the documentation

"Creation is the discipline of removing what obscures."

This case study is part of the CREATE SOMETHING research program exploring AI-native development patterns. View the Kickstand repository or read more about our methodology.