Senior Laravel/PHP engineer building AI-augmented development workflows.

PHP since 1999 · Laravel since the v4 beta era · Dayton, Ohio

I build production web applications, data systems, and the operating layers that make software agents useful without asking anyone to trust an agent's claim that it is “done.” My current work combines deep Laravel experience with agent orchestration, deterministic verification, durable state, and explicit boundaries on autonomy.

A private R&D codebase with a public product and architecture portal. ContextStack turns engineering work into bounded, verifiable runs: a Laravel control plane coordinates work, Go services handle distribution and execution, isolated workspaces contain each run, and append-only traces make outcomes reviewable.

The implementation is active and private. The public site shows the thesis, architecture, current measurements, and the exit gates that must be met before a workflow earns more autonomy.

An Agency Intentional research publication documenting the evolution of autonomous software production and the practical work of building a dark software factory. I create and edit the publication; its research informs the agency's client work, while lessons from that work feed back into the corpus without exposing client-confidential details.

As of an August 2026 audit, I operate a first-party estate of 24 scheduled tasks, 22 enabled, producing roughly 41 invocations per day. It spans research monitors, publishing support, job and event discovery, and personal operations; these are not all coding agents.

The patterns are intentionally unglamorous:

• durable JSON ledgers instead of conversational memory;
• thin schedulers over versioned operating manuals;
• shared configuration read at runtime;
• stable deduplication keys and delta-focused reporting;
• direct source verification rather than snippets or cached summaries;
• independent output-freshness checks, because a scheduler timestamp proves an attempt—not a successful result; and
• an explicit rule that agents may propose changes outside the state they own, but may not silently apply them.

Claude Code is my daily driver. I use agents to plan, implement, refactor, test, and investigate across parallel workstreams, while keeping acceptance outside the model: tests, static analysis, linters, browser checks, source inspection, and human review decide whether the work is complete.

The other half is framework judgment. Generated Eloquent can hide an N+1 behind clean-looking code. A generated migration can lock a production table. A generated test can prove the mock instead of the behavior. Speed comes from letting agents do more work—not from lowering the standard for accepting it.

• 25+ years building production web applications
• PHP since version 4; Laravel since version 4 beta
• Six years at Life Happens, progressing from senior developer and architect to CTO
• Founder of Agency Intentional, the continuation of Vaught Consulting (est. 2010)
• Hands-on experience across Laravel modernization and greenfield work, database architecture, APIs, AWS operations, CI/CD, and distributed engineering teams
• Previous open-source contributions include merged pull requests to laravel/docs and cartalyst/sentry

• Public: this profile, the sites linked above, older repositories, and selected open-source contributions.
• Private: the active ContextStack implementation and most client code.
• Running in production: the scheduled-agent estate described above.
• Experimental: higher-autonomy ContextStack workflows remain gated by published measurements; they are not presented as solved.

LinkedIn · Personal site · Dark Factory Dev · ContextStack