Production software, built to be owned
I build and maintain production systems, and I help teams decide what to build, buy, or shore up when the economics of each shift quarterly.
Cameron Hightower
Austin, TX — consultant, engineer, production systems
Builder, not just an advisor
Companies bring me in to decide whether to build software or license it. I make that decision with them against LLM and harness capabilities, token costs, and engineering labor, all of which shift quarterly. When the answer is to build, we do it together.
For a telehealth company operating across more than 90 locations in Texas, that meant three connected platforms on a shared backend, built in about two months, after they had spent nine months with other developers and shops attempting to get a working system. Those platforms are now in daily clinical use, and I maintain the system solo with the help of AI agents: roughly 250,000 lines of code.
That experience is the basis for my judgment on these decisions. I know the current and near-term state of the tools, and what maintenance on a production system costs over time.
The system comprises three products for three audiences on one shared backend.
- Onboarded clinics
- 52
- Onboarded clinics
- Telehealth visits per week
- 100+
- Telehealth visits per week
- Incidents since launch
- 0
- Incidents since launch
Also building: Juniper — a multi-agent system that turns plain-language requests into standing automations across any configured service. Instead of re-prompting a model on a schedule, the agent reads the request once and writes a declarative script. That script is validated before it goes live, then run deterministically on schedules and events, calling a right-sized LLM at runtime only for the steps that require intelligence. The architecture is open source.
Build, Buy, Vibe?
Building software has become fast and inexpensive, so the difficult parts are now deciding what to build and maintaining what you have built.
Build
Build the tool deliberately, with a designated owner responsible for maintaining it. That owner may be a member of your team or a contracted engineer.
Buy
License traditional SaaS. You pay a subscription, and the vendor bears the maintenance burden.
Vibe
Vibe-code the tool and deploy it without a designated owner or engineer. No one is responsible for maintaining it.
Offerings
Vibe-Code Health Check
I review the vibe-coded tools your company depends on and report on their risk, maintainability, and operability.
The Build Decision
I cost out maintenance across all three options and help you decide, working from the realities of your organization and evidenced projections of where the tooling is going.
Custom Build
When ownership costs less than licensing, I either advise your team on the build or execute it myself.
How I Engage
Every engagement begins the same way: I learn how your work happens. What follows depends on what I find.
- 01
Embedded Discovery
I embed with your team to learn the workflows, bottlenecks, and undocumented knowledge that process documentation leaves out. The output is a prioritized list of the problems worth solving with software.
- 02
Roadmap & Architecture
I turn that list into a plan: what to build, what to buy, what to shore up, what data and systems work has to come first, and how success is measured.
- 03
Build
The roadmap selects one or more of the following, according to what the situation requires:
Build it for you
Production-grade traditional and agentic software.
Build it with you
I pair with your engineers so that the system's context remains in-house from the outset.
Harden what you have built
I stabilize the vibe-coded tools your company already depends on: tests, documentation, security, and a maintenance plan.
Advise you to buy instead
When a vendor costs less than building and maintaining your own, I recommend the vendor and help you integrate it.
- 04
Ongoing Partnership
A system in production needs a designated owner. That owner may be someone on your team or myself on retainer.
Includes: patching, model migrations, token cost management, monitoring, a defined response window, and new builds on top of what is already running.
Buying a company rather than running one? I run technology and operations diligence for acquirers under LOI→
Looking Ahead
The Three Pillars of Agent Operability
Agent operability is the degree to which your data, systems, and roles are structured for Human ↔ AI collaboration and autonomous agentic work. Each gap in agent operability is a place where your organization does not capture the gains AI can deliver.
Data Organization & Accessibility
I audit how your information is structured, surfaced, and permissioned, then make it discoverable to the humans and agents who need it.
Systems Agents Can Act On
I build the surfaces agents need to act safely: APIs and MCP servers where there were only login screens, permissions scoped to each agent, and audit trails for agent actions.
Human ↔ AI Coordination
I define where agents act autonomously, where humans review, and where the two work together, then design the touchpoints and approval flows to match.