Notes from the software factory
Build the system that builds the software.
Real patterns from real projects. 338 sessions, 256 commits, and Dec 31 – Feb 27 of building software with AI as an engineering partner. The look borrows from the compound engineering talk because the thesis is the same: the system is the thing that compounds.
Longer reads
Featured essays
ReviewPing Is an Experiment in the Part of AI Startups Most Builders Still Ignore
I built a small SaaS to learn GTM strategy and agentic engineering in the same live loop.
I One-Shotted an Entire SaaS App with OpenAI's 5.4 Model
One prompt. One run. A fully functional link-in-bio app with Stripe billing, SMS delivery, and Supabase auth — built in Codex.
I Stopped Prompting and Started Orchestrating
306 sessions taught me that the next job in tech isn't prompt engineer. It's orchestrator of agents.
Why We Rolled Our Own
On evaluating SaaS options, building with Supabase + Stripe + Resend, and the 4-hour experiment that changed everything
The Two-Slice Team
Block just cut 4,000 people. Jack Dorsey says AI + smaller teams do more. Every runs 4 products with single-person teams. This is what comes next.
I Run Multiple Claudes at Once (and Other Surprises From My Data)
81 parallel sessions. A 72-second thinking loop. 1,140 messages after dark. The numbers behind 924 hours.
Building a Second Brain with Claude Code and Obsidian
28 sessions using an AI coding tool for something that isn't coding — scaffolding a professional knowledge system for a new Director role
What Will My Kids Do for Work?
I have three boys entering the workforce in the next 3-10 years. The honest answer is I have no idea what's waiting for them.
Blueprint for an AI-Driven Software Factory
What happens when you stop thinking of AI as a coding assistant and start thinking of it as a production line
Start here
Featured post
What It's Actually Like to Use Claude Code for Everything
338 sessions, 256 commits, 924 hours — an honest account of treating AI as an engineering partner
Archive
All posts
The Workflows That Actually Work
Concrete patterns for shipping features, fixing bugs, and running code reviews with Claude Code
Where Things Go Wrong
An honest friction log from 338 sessions — buggy code, planning paralysis, and deployment gotchas
Claude Code Power User Guide
Battle-tested prompts, CLAUDE.md rules, and workflow tricks from 338 sessions
The Best Story From 924+ Hours of AI Coding
User asked Claude to build a dev blog from their Claude Code session data — then realized it was about to publish all their raw transcripts for the world to see
Where AI Coding Is Actually Heading
Parallel agents, self-healing deploys, and autonomous development pipelines — based on patterns already working
5 Projects, 338 Sessions: What I Built
From SaaS platforms to infrastructure to marketing sites — what Claude handles well and where it struggles
Built by Matt Hartman
A logbook for working in public with AI.
Director of Intelligent Systems at Monti, Inc. 20+ years in enterprise architecture, IoT platforms, and product leadership. Previously at P&G and the University of Cincinnati.