I build products. From the interface to the infrastructure.
Co-Founder & CTO at LineCrush. I started the codebase, built the business with Jay, and took native iOS and Android apps to both stores for LineCrush and SureWord. Directing AI agents is my daily work: I brief them, review what they ship, and stop the ones headed the wrong way.
I started the codebase and built LineCrush with Jay. I started the iOS app in SwiftUI and the Android app in Kotlin and Jetpack Compose and wrote their first month myself. I kept building them with the agent fleet I designed, and took both through release and store review.
I built SureWord for the web, then built the iOS and Android apps myself. To keep verse quotations exact, the server supplies the Scripture text while the AI explains it.
I built the engineering agent fleet and our issue tracker, C2. Jay and I operate the fleet together. The LineCrush case study follows one change all the way to the App Store release notes.
10,745completed fleet jobs with a recorded commit link
Commercial game · Steam
Gunmetal Orbit
I built a space-mining roguelite for Steam Deck and PC. It took 67 days to get from the first playable build to Steam. After launch, I caught a shop navigation bug on my own Steam Deck and wrote a regression test that drives the same UI with controller input.
I didn't take a straight path into software. I drove trucks, went back to college, taught web development, and started building products of my own. I'm still learning every day.
Before tech
A different kind of daily driver.
My first career was driving trucks. The work was different, but taking responsibility and following through mattered just as much. I brought that with me when I went back to school.
From the road to the classroom
2021
Learning the foundations.
I earned my A.S. in Information Systems: Web Design at Clovis Community College. I also started doing web development work for Effycient and the college, building client sites and maintaining web pages.
Clovis Community College
Jan 2022 - Jul 2023
Teaching what I was building.
I taught web development at Bitwise Industries and Geekwise Academy. We worked through HTML, CSS, JavaScript, and React. Helping someone get their first app working was definitely one of the best parts.
Web Development Instructor
2023 - 2024
Real businesses, real requirements.
As a paid full-stack developer for Gelateria Del Centro, I built and maintained the business website. I owned the implementation, content updates, and search visibility, and learned what it takes to keep a site useful after launch.
Client development
2024 - today
Starting LineCrush. Building it with Jay.
I started the codebase in 2024 and built LineCrush with my co-founder, Jay. It grew into a paid sports research platform. As CTO, I work across the product, backend, infrastructure, and the things that need fixing once people actually use it.
I built the iOS and Android apps for both LineCrush and SureWord myself. I also brought SureWord to Mac, released Gunmetal Orbit on Steam, and designed the agent fleet our team uses to keep shipping. I run those agents all day from my desktop, laptop, and phone. I know how to brief them properly, how to review their work, and when to stop one that's headed the wrong way.
An agent that looks things up before it answers. It searches this portfolio, reads my live work ledger and public GitHub, and links every answer back to the source.
portfolio-agentReady
gpt-6.2-sol · read-only tools
✻Ready. Ask anything about Austin, or start with one of these.
Portfolio & case studies
Live work ledger
Public GitHub
Verified stats
Stories & decisions
Blog posts
Steam listing
Prefer a command line?help · agent run · ask
Dev Blog
Learning in Public
Notes from the trenches. What I learn building AI-powered systems, leading engineering at LineCrush, and shipping code every day.
Architecture, product intelligence, autonomous delivery, and the infrastructure to run it all.
43K+
Monorepo Commits
70
Agent Skills
725
API Handlers
4
Product Clients
3+
Shipped Game Builds
2+
Years in Production
Agentic AI Engineering
Built autonomous delivery agents and multi-model product systems with durable sessions, provider routing, custom MCP tooling, evidence gates, streaming, vision, and NLP pipelines.
Codex
Claude
OpenAI
Gemini
Perplexity
MCP
Agent Skills
Full-Stack Architecture
Owns a production platform spanning ~2,300 TSX modules, 725 API handlers, Python data services, native iOS and Android clients, and browser extensions.
Next.js 15
React
TypeScript
Python
SwiftUI
Kotlin
PostgreSQL
System Design
Architected real-time data pipelines, distributed cron jobs, multi-tier Redis caching, and 450+ PostgreSQL tables with complex relationships.
Redis
PostgreSQL
Real-time Systems
Caching
Data Pipelines
DevOps & Infrastructure
Managing Ubuntu VPS with automated deployments, Vercel edge functions, AWS services, and production monitoring for 24/7 operations.
AWS S3/SES/CloudFront
Ubuntu VPS
Vercel
Docker
CI/CD
Technical Leadership
As Co-Founder and CTO, I set technical direction, translate product bets into architecture, and created the agent workflows that let a small team ship with serious leverage.
CTO
Architecture
Product Strategy
Agent Operations
Code Review
Data Engineering
Built web scrapers for YouTube, podcasts, news, and social media. Real-time NLP analysis and automated data processing pipelines.
Web Scraping
ETL
NLP
Automation
Cron Jobs
Coding Activity
Live coding stats tracked by WakaTime
Core Tech Stack
Frontend
Next.js, React, TypeScript, Tailwind, shadcn/ui
Backend
Python, Node.js, PostgreSQL, Redis
AI/ML
OpenAI, Anthropic, Gemini, Perplexity, MCP
Infrastructure
AWS, Vercel, Ubuntu VPS, Docker
Measured hourly, updated
Daily Work Log
Measured from real activity across my two machines and updated through the day: when each day starts and ends, how much of it I spend actively prompting my coding agents, and what ships.
Average day, first to last prompt
11.5hours
Clock time from the first prompt to the last, breaks included, averaged over 30 full days in the last 30.
4.9h actively promptingagent runs, reviews, calls and breaks
10+ hour days
74of the last 90
Days with 10 hours or more between the first prompt and the last, breaks included.
10h
90 days agotoday
Days worked
90of the last 90
Days I sent prompts to my coding agents. Current streak: 103 days.
oldesttoday
Today, first to last prompt so far
12.6h
6.1h actively prompting, Nov 17
Active prompting, 7-day average
5.9h
per day, last 7 full days
Current streak
103
consecutive days worked
Prompts sent
3,495
last 30 days
Agent hours, 30 days
615h
349 parent sessions tallied by day
Peak agents at once
24
on Nov 17
Agent actions, 30 days
102,089
file edits, commands, searches
GitHub contributions
33,338
last 365 days, includes fleet work
The last year, day by day
4,344 hours between first and last prompts since Sun, Nov 16, 2025
Use the arrow keys to move between days. Left and right move a week, up and down move a day.
Hours from first to last prompt, per day0under 4h4 to 8h8 to 12h12h+
Last 14 days
Hours of active prompting by project. Gaps of more than 30 minutes between prompts are left out, so long agent runs, reviews and calls don't show here.
0h4h8h12h
8.0h
4
W
5
T
6
F
7
S
8
S
9
M
10
T
11
W
12
T
13
F
14
S
15
S
16
M
17
T
Use the left and right arrow keys to move between days.
LineCrush
LineCrush Games
SureWord
Client sites
Personal & tooling
Where the prompting time went
150 hours of active prompting over the last 30 days
31%
15%
37%
16%
LineCrush47.2h31%
LineCrush Games22.4h15%
SureWord54.8h37%
Client sites1.1h1%
Personal & tooling24.4h16%
Oct 19 to Nov 17
Show the last 30 days as a table
Daily work log, last 30 days, newest first
Date
First prompt
Last prompt
First to last, hours
At the computer, hours
Active prompting, hours
Prompts
Agent hours
Peak agents
Agent actions
Contributions
Tue, Nov 17, 2026
8:29 AM
9:04 PM
12.6
n/a
6.1
270
23.2
24
6,781
227
Mon, Nov 16, 2026
9:50 AM
8:08 PM
10.3
n/a
5.9
146
5.9
8
1,088
283
Sun, Nov 15, 2026
8:36 AM
6:41 PM
10.1
n/a
8.0
323
51.3
14
7,068
395
Sat, Nov 14, 2026
8:37 AM
6:19 PM
9.7
n/a
5.5
117
123
21
12,043
280
Fri, Nov 13, 2026
7:00 AM
8:50 PM
13.8
n/a
6.4
162
65.9
17
9,183
390
Thu, Nov 12, 2026
6:42 AM
6:29 PM
11.8
n/a
4.5
102
8.4
6
1,624
281
Wed, Nov 11, 2026
8:19 AM
9:16 PM
12.9
n/a
3.3
112
9.2
8
2,522
328
Tue, Nov 10, 2026
6:37 AM
6:22 PM
11.7
n/a
7.9
246
36.9
14
8,625
313
Mon, Nov 9, 2026
7:13 AM
3:43 PM
8.5
n/a
3.3
86
18.5
6
3,093
195
Sun, Nov 8, 2026
10:07 AM
6:19 PM
8.2
n/a
4.2
111
83.7
7
11,714
293
Sat, Nov 7, 2026
10:53 AM
5:29 PM
6.6
n/a
1.9
53
5.2
10
1,332
275
Fri, Nov 6, 2026
6:25 AM
6:56 PM
12.5
n/a
3.7
70
7.1
4
1,789
245
Thu, Nov 5, 2026
8:06 AM
5:48 PM
9.7
n/a
4.4
106
12.5
5
2,528
388
Wed, Nov 4, 2026
8:06 AM
10:20 PM
14.2
n/a
2.8
56
1.4
3
365
334
Tue, Nov 3, 2026
6:17 AM
9:50 PM
15.6
n/a
12.1
232
25.4
13
3,971
471
Mon, Nov 2, 2026
4:12 AM
8:02 PM
15.8
n/a
5.2
37
0.1
1
40
542
Sun, Nov 1, 2026
6:01 AM
3:57 AM
21.9
n/a
13.8
93
10.8
5
1,376
605
Sat, Oct 31, 2026
12:01 PM
11:00 PM
11.0
n/a
1.4
8
1.3
4
330
377
Fri, Oct 30, 2026
8:15 AM
7:52 PM
11.6
n/a
8.4
134
49.6
10
8,100
364
Thu, Oct 29, 2026
5:20 AM
6:39 PM
13.3
n/a
5.0
91
44.6
10
10,004
149
Wed, Oct 28, 2026
6:56 AM
10:26 PM
15.5
n/a
5.0
107
22.6
9
6,083
204
Tue, Oct 27, 2026
7:15 AM
5:01 PM
9.8
n/a
2.4
70
0
0
0
158
Mon, Oct 26, 2026
8:47 AM
4:36 PM
7.8
n/a
2.9
69
0.1
1
52
179
Sun, Oct 25, 2026
9:42 AM
9:31 PM
11.8
n/a
3.1
75
0.8
2
239
231
Sat, Oct 24, 2026
11:26 AM
3:37 PM
4.2
n/a
3.6
77
0.4
2
141
253
Fri, Oct 23, 2026
6:12 AM
3:55 PM
9.7
n/a
3.6
126
0.5
1
119
177
Thu, Oct 22, 2026
6:58 AM
6:36 PM
11.6
n/a
6.4
167
1.9
2
624
315
Wed, Oct 21, 2026
9:23 AM
4:20 PM
6.9
n/a
3.6
101
1.9
1
377
203
Tue, Oct 20, 2026
7:10 AM
7:20 PM
12.2
n/a
1.5
60
0.1
1
38
221
Mon, Oct 19, 2026
6:47 AM
9:29 PM
14.7
n/a
3.7
88
2.2
3
840
276
How the work log is measured
First to last prompt is clock time: when I sent the day's first prompt and when I sent its last, breaks included. It is not hours worked. Active prompting only counts the time between prompts sent up to 30 minutes apart, so a long agent run, a review, testing or a call falls outside it and the number reads low. My real working time sits between the two. Time at the computer, measured from keyboard and mouse input with games excluded, fills that gap from Nov 18 on. Agent runtime and GitHub contributions are tracked separately.
C2 and the LineCrush agent fleet
I built the LineCrush agent fleet and C2, the system I use to run it. The fleet investigates issues, builds and verifies changes, and ships them to production.