One person, with agents, on the record.
I design, build and run software end to end: product, design, engineering, infrastructure, support. The agents do the work a crew would, under gates I decide. This page is the process in plain words, so a buyer can check it before the first call.
01 · Scope and proposal
Nothing starts without a written scope every party has accepted. The scope says what ships, in what order, by when, and what is out. It is the contract and the handoff: the same document the build is run from, kept in the repository, versioned with the code.
Change is normal. A change to scope is a new written line, accepted the same way. Hours are logged against the line they belong to and visible to the client while the work is live.
02 · Agents under gates
Agents read, draft, test and propose freely. Every send, every payment, every write to production and every deletion passes a gate a person decides, and the decision is written to a log with who decided and when. Listening is free; acting needs a yes. The first piece on the front page is that rule running; it is also how this site and every client system I run are operated.
No agent holds a credential a person could not revoke in one place. Agents act through the same accounts, roles and audit trails a contractor would, never through a shared secret.
03 · Data handling
Client data lives in the client's own systems and accounts, or in infrastructure the client owns and I administer. It is not copied to this site, to social media, to a demo, or to a personal machine. Every demo on this site runs on synthetic data and says so beside it.
- Access is least-privilege, per person and per agent, removed at handoff or on request the same day.
- Every account that can reach production has multi-factor authentication. Secrets live in the platform's secret store, never in code, chat or documents.
- Data is encrypted in transit and at rest on the platforms in use (Vercel, Neon, Supabase, GitHub), and each client gets the list of which platforms hold what.
- If something goes wrong, the client hears from me within 24 hours with what happened, what was affected, what was done, and what changes.
- Nothing about a client's contract, program, facilities or systems is published without written consent. Case studies here are shown for the engineering: industry, scale, problem, measured result.
- For federal work, the fifteen basic safeguards in FAR 52.204-21 are the floor I work to for contract information. I do not hold a CMMC certification; I take no work involving controlled unclassified information until the assessment that work requires exists.
04 · How a change reaches production
A branch, a pull request, a type check, Lighthouse three times on every page in the contract, a crawl of every internal link, a preview deployment for review, then production. A push that misses the contract is marked red and does not merge. Rollback is one action on the platform and is tested, not assumed.
The result of that process is counted, not claimed: 1,777 production deploys since Sep 1, about 29 a day, read from the deployment API each hour. This site is held to the same contract as the client work.
05 · Testing
What is covered and what is not is written in the scope. Type checks on every push. Behavior tests on the paths money or data move through. Contract checks (performance, accessibility, links) on every build. Regressions are caught by the same checks the first version passed; a bug found in production becomes a test before it becomes a fix.
06 · Handoff
The client owns the repository, the infrastructure accounts and the domains from day one, or receives them at handoff with a written runbook: how to deploy, how to roll back, where the secrets live, what runs on a schedule, who to call. A dependency manifest ships with the code. Support terms are written in the scope, not implied.
07 · Working with contractors and primes
For defense and aerospace primes, tribal 8(a) firms and task-order vehicles in the Oklahoma City corridor, I work as a subcontractor or as a fractional senior engineer on a team, not as a prime. I bring product, design and full-stack engineering (TypeScript, Next.js, Postgres, document and data pipelines, agent systems with approval gates), the process above, and a public record of shipping. Registration and vehicle details are available on request.
08 · References
References from current clients are available on request, by email. Every live system on the front page has a public link beside it; the ones that do not say why.