Open Access Research

We built 10 production systems with AI. Then we published every number.

Defect rates. Build costs. Speed. Quality. Contractor spend. Every meaningful metric — tracked, verified, and open. The first dataset of its kind.

63 publications — 17 open access, full library via sign-up
Git-verified, financially audited
Every number independently verifiable
60 45 30 15 0 S1 S3 S6 S8 S10 SYSTEMS SHIPPED
13.4×
Output acceleration
12.1%
Defect rate held
Linear
596,903
Lines of production code
22/day
Average output rate
12.1%
Defect rate
$65K
Total build cost
What's In The Dataset

Six categories. Real numbers. Every one sourced.

Each category contains independently verifiable data — pulled from git logs, financial records, and transaction data.

4.6 → 61.5
Changes/day · 13.4× acceleration
Speed & Output
12+ data points
12.1%
vs. 20–50% industry norm
Quality & Error Rates
10+ data points
23–84×
Return on $34,473 build spend
Cost & ROI
8+ data points
69% → 7%
Outside dependency · 4 months
Independence & Control
8+ data points
194,954
Lines in one system · WordPress: 160K
Scale & Complexity
6+ data points
10 of 10
Validated against published research
Methodology Validation
6+ data points
11×median developer output
31×avg Meta engineer output
3–6×faster than agency timeline
½ to ¼industry defect rate
79%contractor cost reduction
6–20×larger than typical SaaS MVP
$0outside cost on fastest build
4 monthsvs 12–18 months typical
11×median developer output
31×avg Meta engineer output
3–6×faster than agency timeline
½ to ¼industry defect rate
79%contractor cost reduction
6–20×larger than typical SaaS MVP
$0outside cost on fastest build
4 monthsvs 12–18 months typical
The Methodology

Technical Papers

Thirty-eight papers. Each isolates one mechanism, grounds it in published research, and validates it against the production data above. Seventeen are open access — sign up to unlock the full library.

Core Mechanism
TP-01
Foundation
Why every project ships faster than the last

Why does every project ship faster than the one before it? Foundation is the reusable asset library that stores everything built across every project — so the next project starts with all the work from the last one already done. This paper documents how that compounding effect works and what it produces.

Read paper
Key Data
Output acceleration 4.6 → 61.5 changes/day (13.4×)
Total code built on Foundation 596,903 lines
Defect rate on reused work 3.7–3.9% (norm: 20–50%)
Theoretical basis Wright's Law (1936)
Proven in Flagship Build · Five Days · Full Portfolio
TP-02
Nested Cycles
How work gets structured without sprints
Extended output days 17 days at 20+ hrs
Zero burnout across 118 days
Proven in: Flagship Build · Full Portfolio
 TP-03
The Pendulum
How every decision gets made — instantly
Decisions without a backlog 15,000+
Industry: 60–80% never completed
Proven in: Independence Curve · Full Portfolio
 TP-04
The Governor
Why quality doesn't collapse at speed
Defect rate at 14.5× speed 12.1%
Industry norm: 20–50%
Proven in: Recovery Build · Full Portfolio
 TP-05
The 80% Premise
Why "done enough" beats "done perfectly"
Fastest system to production 5 days
10 of 10 systems revenue-ready
Proven in: Five Days · Full Portfolio
 TP-06
The Drift Tax
What AI actually gets wrong — measured
AI error rate (structural) 12–15%
2,561 changes measured · managed, not eliminated
Proven in: Flagship Build · Recovery Build
 TP-07
Scaffold
How new products start with 100K+ lines already done
Defect rate on reused code 3.7%
Industry: 20–50% · 79% cost reduction
Proven in: One Scaffold · Same Product, New Country
 TP-08
Bridge
Why connecting beats building
Systems cross-connected 10 across 7 verticals
2 countries · shared infrastructure
Proven in: One Scaffold · Same Product, New Country
 TP-09
Stop, Pause, Reset
What happens when things break
Worst defect spike → final 41.7% → 0%
2 breakages · 2 recoveries · on time
Proven in: Recovery Build
 TP-10
No Backlog
Why there's no to-do list — by design
Backlog items across 10 systems 0
Industry: 60–80% never completed
Proven in: Full Portfolio
The Evidence

Case Studies

Twenty-five real builds — tracked from day one through deployment. Every number verified against version control and financial records. Seven are open access — sign up for the complete set.

CS-01
The Flagship Build
One platform replaced six paid tools
194,954
Lines of code
13.4×
Speed increase
$82,640
Annual savings
CS-02
Five Days to Production
Zero to revenue-ready in 120 hours
5 days
To production
100%
Solo execution
$0
Outside cost
CS-03
One Scaffold, Four Products
One codebase deployed into four markets
4
Products shipped
3.8%
Avg defect rate
79%
Cost reduction
CS-04
The Recovery Build
Two breakages. Two recoveries. Shipped on time.
0%
Final defect rate
113
Changes in peak week
7
External integrations
CS-05
Same Product, New Country
South Africa → United States in 16 days
2
Countries served
16 days
US market entry
~$330
Expansion cost
CS-07
The Independence Curve
From directing others to doing it all — in 4 months
69→7%
Outside dependency
$7,995→$0
Contractor cost
4 mo
To full independence
CS-06
The Full Portfolio
The complete picture — every system, every metric
Total changes logged 2,561
Calendar days 116
Replacement value $795K–$2.9M
Ownership retained 100%
596,903
Lines of code
10
Systems shipped
23–84×
Return on investment

See What the Data Built

The research documents the methodology. The methodology produced the systems. The systems are live and generating revenue.

The Method
Compounding Execution Method

The methodology behind these results — how it works, what it produces, and how to apply it.

Explore CEM →
Solutions
Build With Us

Custom builds, proven systems, and training — powered by the same methodology documented here.

See Solutions →
Contact
Get in Touch

Questions about the dataset, methodology, licensing, or working together.

Send Email →