AI-native operator model
AI-native engineering changed my output curve.
I operate at the intersection of executive strategy and hands-on build velocity. The value is not just that I can ship faster with AI. It is that I know what should be built, what should not be built, and how to turn ambiguity into working systems.
The current proof: managing 13-15 active codebases, 7K commits, and 2.8M lines committed across a portfolio where the month-over-month velocity keeps rising.
Live engineering velocity
Updated weeklylines in production
active codebases
net lines this week
Net lines / month · 26 mo
vs. a 100-engineer team
Velocity dashboard
Source-code output curve
An anonymized line graph of engineering velocity: lines committed, net movement, the smoothed trend, and an optional hand-coded baseline.
Updated Jul 11, 2026, 7:09 AM ET · refreshes weekly
Selected period
To dateJul 2026
Month-over-month
In progress
Current period is still accumulating; full-period change lands at month end.
The value math
~6× a 100-engineer team — redefining production cost
The amber baseline on the chart isn’t a guess. Capers Jones’s analysis of thousands of projects puts a professional developer at 325–750 lines of production code per month once pull requests, reviews, QA, and rework are counted. Taking the generous top of that range, a 100-engineer team ships about 75K lines in a month. Recent months here run roughly 6× that — about 579 traditional engineers worth of net output — from one operator who also carries product judgment and business context.
net lines / developer / month
Top of Capers Jones's published 325–750 range (Scrum ~780)
a 100-engineer team / month
100 developers × 750 lines, before the PMs and QA needed to ship
recent monthly output here
434K net lines — one AI-native operator
engineer-equivalent org
6× a 100-engineer team — the headcount to match this pace by hand
What it used to cost to build at this pace
The old economics software pricing was built on — before PMs, QA, recruiting, and management layers.
~579 engineers, onshore (US)
$195K–$262K fully loaded per engineer. A 100-engineer slice of this org alone runs ~$20M–$26M/yr.
~579 engineers, offshore
$62K–$79K per engineer in a managed model — coordination, latency, and review overhead not included.
What it costs to build now
One AI-native operator at ~$350K–$500K/yr (fully loaded + tooling) producing the same measured pace — versus ~$210M–$305M/yr for a traditional 1066-person delivery org.
cheaper to produce the same output (full org)
cheaper vs engineers-only onshore band
When production cost collapses, product pricing can follow — the road back to affordable software.
The cost of building software is collapsingLearn more — the math & sources▸
Output: Capers Jones, after comparing thousands of projects across methodologies, found professional developers sustain roughly 325–750 lines of delivered code per month; Fred Brooks’s OS/360 data in The Mythical Man-Month famously landed near ~10 lines/day. We use 750 — the top of the published range — so every baseline is generous to the traditional team.
Scale: Engineer-equivalent headcount = recent monthly net lines ÷ 750. At 434K net lines, that is ~579 engineers — roughly 6× a 100-engineer coding team (434K ÷ 750 ≈ 579).
Headcount: A 579-engineer delivery org also carries product managers and QA — a common ratio on a 100-engineer unit adds ~15 PMs and ~25 QA — so the real traditional cost is higher than engineers alone.
Cost: Fully-loaded US software engineers run $195K–$262K/year ($80–$150+/hr via agency); offshore managed teams run ~$62K–$79K/year ($18–$45/hr). At 579 engineers that is ~$113M–$152M/year onshore versus ~$36M–$46M/year offshore. One AI-native principal operator is modeled at ~$350K–$500K/year including tooling.
Sources: NDepend / Capers Jones & Brooks on LOC/month · FullStack 2025 rate guide · DistantJob onshore vs offshore rates
GitHub contribution graph
6,991 contributions in the last year
Daily commit, pull request, review, and issue activity across public and private repositories — the same graph GitHub shows on my profile.
Active codebases managed
Lines retained in working products
Commits shipped
Total source movement
Evidence without noise
The receipts are anonymized. The operating pattern is not.
The data spans real codebases and real implementation work, but this page intentionally groups the work by domain instead of exposing project names. The point is the pattern: executive-level prioritization paired with builder-level throughput.
The speed is not chaos. Everything is pushed through platform patterns for security, integrity, repeatable engineering practice, and serverless scalable foundations where the architecture calls for them.
Healthcare AI platform
69%Clinical workflow, note generation, and operational product systems.
Mobile and desktop surfaces
21%Cross-platform product experiences that meet clinicians where they work.
EHR integration and automation
1%Scheduling, context resolution, writeback, and workflow automation.
Agentic workflows
3%Internal tools and agent loops that compress research, coding, and QA.
Clinical data infrastructure
1%Data movement, validation, test harnesses, and environment plumbing.
Go-to-market systems
5%Brand, sales, documentation, and customer-facing proof assets.
Executive judgment
Start with the business constraint, workflow reality, buyer pressure, and product consequence before touching code.
Hands-on build loops
Turn ambiguous ideas into prototypes, automation, integrations, and production-ready direction instead of slideware.
AI-native leverage
Use agentic workflows to compress research, implementation, review, and iteration into a tighter operating cycle.
Platform discipline
Move fast through reusable patterns for security, integrity, good engineering hygiene, and serverless scalable foundations where they fit.
Fractional AI-native product engineering
AI-native product engineering retainer
$3,000/month for 8-12 hours. This is not staff augmentation. It is a compact operating loop for teams that need product clarity, technical direction, and shipped artifacts before committing to a larger hiring plan.
Current proof point
lines committed into working systems across the portfolio. The important signal is not volume alone; it is sustained executive-to-code throughput on scalable foundations.
Work with me