Twenty years of engineering, with AI as a very fast pair of hands.
I use AI to build your site. I design the system, direct the work and read every line, and AI types faster than any team could. You pay for the result.
How AI is used on a project
- Research and draftingGathering context, drafting content and documentation, exploring options.
- Scaffolding and boilerplateThe repetitive code that used to take hours of typing.
- TestsFirst drafts of tests that I then review and extend.
- Review supportA second pair of eyes that points to things worth checking.
What is never delegated
- Architecture and trade-offs
- Security-sensitive code, without line-by-line review
- The decision about what ships
- Conversations with you
Why experience matters
Every language I have worked in shares the same bones: control flow, data structures, state, input and output, and the same ways of failing. That is why a senior engineer can read an unfamiliar codebase in an afternoon, and why AI output is easy to judge once you know what correct looks like. AI is fast and fluent and sometimes confidently wrong. Knowing the difference takes years of watching software fail. For example, a draft might apply a change before checking that a person approved it. Catching that is the job.
Stack-agnostic by design
After twenty years across PHP, JavaScript, TypeScript, Python and SQL, a new framework is a dialect, and reading what code does is the easy part. AI's job is the dialect: the exact signature, the idiom, the boilerplate. Mine is the system: how it is built, where it will fail and what it should do. So I work across TypeScript, Next.js, Node, PHP and WordPress, Python, SQL, Astro and Cloudflare, and pick the tool that fits the problem.
Where this comes from
I led the adoption of AI-assisted development across an agency web team: the tooling, the standards, and the review that keeps it honest. This site, ZoneSteward and CookieSteward are all built this way.
Why it costs less and moves faster
An agency bills for a team: strategists, designers, developers, project managers and account managers, each adding hours and handoffs. Here, one engineer covers the chain, with AI removing the typing and lookup that used to fill a developer's day. There are fewer handoffs, no team overhead and no relay of messages, and I give direct attention to every decision.
Related
Is this just copy and paste from a chatbot?
No. AI is one tool inside a process with design gates, staging, tests and review. The output is read and judged by a person, and the architecture decisions are mine.
Does it lower quality?
It should not, and the process is built to prevent it: tests, review, accessibility and security checks run on every project.
Who is responsible when something is wrong?
I am. Nothing ships that I have not reviewed, and you have one accountable person to talk to.