Ask me anything. Here is a head start.

I rebuild how a company is understood on the web and by AI. That covers the positioning, the content and evidence behind it, the website itself, the machine-readable layer AI systems consume, and the research that measures whether any of it moved. Three ways to engage: fractional AI visibility leadership, a complete web presence transformation, or a custom AI system build.

A head-to-head AI visibility assessment. I run real buying questions across ChatGPT, Claude, Perplexity, Gemini, and Grok and report how they describe, cite, and recommend you against a competitor field you name. You get the findings, the gaps, and a prioritized plan. Start one from the contact page.

Leaders who suspect AI answers are sending buyers somewhere else and want one owner across strategy, content, website, and measurement. I have done that class of work for Fortune 5 enterprises, state and federal programs, and more than 40 universities.

No. I work as a consultant or fractionally, which means a defined slice of my week is yours for as long as you need it. That covers head, director, and principal-level AI leadership without the employment. Austin, TX or remote US. This site is the working sample; test it before we talk.

Because the models read your public evidence and found a better answer somewhere else. I measure the choice: locked prompts across five platforms, official-domain verification, and grounded citations, normalized so the same test can run again after the work ships. Then I change the evidence and retest.

SEO optimizes for a ranked list of links. AI visibility is about being the answer: whether ChatGPT, Claude, Perplexity, Gemini, and Grok describe you accurately, cite your pages, and recommend you when a buyer asks. The measurement, the content work, and the site architecture are all different jobs.

Soapbox Bulletin went from a 6.2 percent to a 36.2 percent AI recommendation win rate in 11 weeks. A financial services firm I keep anonymous took four out of four head-to-head wins in a strict URL-grounded series. The evidence is directional rather than a controlled study, and I will walk any number line by line.

Human Agentic Lifecycle Orchestration, the method under all of my work. One responsible human directs bounded AI workstreams: agents research, draft, and verify inside explicit stages, a person approves everything that ships, and knowledge lives in durable files instead of a model's memory. Orchestration, not autonomy.

My working thesis: content is the algorithmic substrate AI systems read when they decide what to explain, cite, and recommend. So treat it like code. Measure how the models answer, decode the signals behind the preferred answers, engineer content that carries them, publish under human review, and retest.

AI removed the bottleneck the CMS existed to solve. A disciplined human can now direct AI to change production code under version control and review, which is faster and safer than a plugin stack. The CMS era is sunsetting. This site is the argument: written straight to code, deployed on a git push. The full essay is on the blog.

Since 1994, before HTML was a standard. I founded AzTech Interactive when fewer than 20,000 websites existed and have shipped through every era since: streaming, enterprise process systems, government, higher education, and now AI.

In late 1995, the moment RealAudio 2.0 made live streaming possible, I put the first live music internet stream from Austin on the air from the Steamboat nightclub on 6th Street. That drew Mark Cuban's AudioNet, later Broadcast.com: more than 100 live broadcasts, artists including Willie Nelson and Barenaked Ladies, and over $600K raised for the infrastructure. In 1998 to 1999 we delivered the quality first: a VHS-resolution, CD-quality live stream of Jimmie Vaughan on Microsoft streaming technology with Broadcast.com.

The biggest was Dell's Worldwide Trade Compliance Profiling System in the early 2000s. It mapped federal denied-party and denied-entity regulations, gave the sales, credit, and compliance divisions the reporting they needed, and plugged into DOMS, Dell's order system at the time. Dell Americas took it over and rolled it out to the salesforce worldwide. I also delivered systems for Compaq, Tivoli Systems, and Los Alamos National Laboratory. Making separate divisions run as one controlled process is the same problem multi-agent AI solves today.

Eleven years at the Texas General Land Office. I built every public-facing .gov property, including glo.texas.gov, recovery.texas.gov, vlb.texas.gov, and cleancoast.texas.gov, and served as Web Accessibility Officer holding them to Section 508 and ADA. After Hurricane Harvey I built the first CDBG-DR federal disaster recovery website with HUD, and HUD shared our Best Known Practices as a development model.

U.S. Patent 7,373,346, Methods and apparatus for improved security services. It covers networked software systems for security management and internet-accessible databases for security data collection and retrieval, granted on my early-2000s architecture work. The full PDF is on this site.

B.A. in Radio-Television-Film from the University of Texas at Austin, with a specialization in audio production and sound engineering. The live-sound years are why the streaming first happened. I still play lead guitar.

Trinzik is the Austin AI company my brother John runs. I co-founded it and architected and built its entire production platform: the Research Engine that measures how five AI platforms describe, cite, and recommend a category, Agentic Websites, and Compliance-Grade Chatbots. I built the systems; the client results Trinzik reports are Trinzik engagements delivered on that platform.

Most tools stop at the report. Trinzik pairs the measurement with the content and website work that changes the result, and every consequential output passes a human before it ships. John leads the client engagements; I build and run the technology underneath them.

Yes. The site publishes an agent card at /.well-known/agent-card.json and a callable JSON endpoint, so another system can discover it and ask questions programmatically. It is a discoverable and callable machine identity, live right now. Details and a test call are on the agent page.

It answers from a knowledge base compiled from this site's approved content, and it captures a meeting request only when you ask for one. The deeper doctrine runs in the research systems I build: grounded means cited or declined.