Agents draft the change. You decide what reaches production.
[ 06 — The fleet ]
One engineer. A fleet of branches.
Parallel agents, small diffs, and a merge gate nothing skips.
[ 07 — Pair and ship ]
Learn the habit. Keep the harness.
Teach
The workflow itself: instruction files, tests first, small diffs and a review habit, practiced on your own repository.
Build
The harness around your agents: CI gates, branch rules, checks and monitoring, built beside you so you can change it later.
[ 08 — Changelog ]
Two weeks of release notes.
Under the hood, each thread is a Claude Code cloud session working on its own branch and copy of the repo.
AnthropicClaude blog, “Projects redesigned: from folder to conversation”, September 17, 2026
Source ↗
Claude Code 2.1.277 reads a project’s AGENTS.md when there is no CLAUDE.md; the AGENTS.md site says more than 60,000 open-source projects use the format.
Engineer of One (engineerofone.com) is run by Utlyze for solo engineers and teams of one to five who ship production code with AI agents. It covers the hands-on side: instruction files, tests, review and paying down technical debt.
What is agentic coding?
Coding where an AI agent plans the change, edits files, and runs commands and tests in your repository, while you set the task and review what it produced before anything merges.
What is the difference between agentic coding and vibe coding?
Vibe coding accepts whatever the model writes as long as it seems to work. Agentic coding still hands the typing to an agent, but keeps an engineer’s controls: a clear task, tests that define done, small diffs and a human review of every merge.
Is vibe-coded software ready for production?
Not by default. It needs the same things any code needs before production: tests, review, security checks and monitoring. Who or what wrote it does not change that.
How do I stop agents from adding technical debt?
Keep diffs small, write or approve the tests before the agent implements, run lint and type checks in CI, and read every change before it merges. Paying debt down works the same way: one bounded change at a time, each with a test that proves behavior did not move.
Which AI coding tool should I use?
Choose by how you work (terminal, IDE or cloud agent), then trial the candidates on your own repository with a few real tasks. Results vary by codebase, and most tools now let you pick the model per task.
What security risks come with coding agents?
The common ones are leaked secrets, unsafe or unexpected dependencies, and prompt injection through content the agent reads, such as issues, web pages or files. Attackers use the same tools: TechCrunch reported in September 2026 that a three-person security team, working in OpenAI’s bug-bounty program, used Claude to break into OpenAI and take over an employee account whose Codex was connected to OpenAI’s GitHub organization. Treat this as general education only; it is not security advice for your systems.
How much are Consult and Build?
Consult is $400 an hour and is sold five hours at a time, so one block costs $2,000. Build is $15,000 a month. Each month, pick up to 5 business processes for our build-with-you coaches to work on. They build your AI systems with you and teach your team to run them. If you ever leave, the systems stay with you. Agents work only on your systems, and a shared 24/7 on-call rotation covers incidents, with response times in writing.
How is this different from Business of One?
Business of One covers the business side: strategy, operations and where AI fits in a company. Engineer of One stays in the code: repos, agents, tests, review and debt. Utlyze runs both, and both sell the same two engagements.
How do I start?
The short form under Begin takes a minute. Choose Consult for pairing time or Build for build-with-you coaches who ship with you, and we write back with next steps.
Built on the Of One method: ask the right questions, map what you find, make the best move. of1.ai ↗
