We want the engineer whose workflow already looks slightly illegal to normal software teams.
You are running your own paid AI plans because waiting on a company seat would be embarrassing. Claude Code, Codex, model routers, local scripts, custom commands, repo agents, throwaway scaffolds, eval harnesses, whatever helps you move. You are not dabbling. You are using LLMs to generate essentially all of your code and spending your time directing, reviewing, testing, and tightening the system.
You do not need to be as client-facing as an Operator-Engineer. You do need enough business sense to understand why the thing exists, where it will be used, and when the technically interesting answer is the wrong answer.
What you will do
- Build internal tools, prototypes, automations, integrations, and production features with an AI-native workflow.
- Use Claude Code, Codex, and similar tools as your normal development environment, not as autocomplete decoration.
- Turn rough requirements into working software quickly, then harden what deserves to live.
- Review AI-generated code with actual engineering judgment around architecture, security, maintainability, and tests.
- Collaborate with Operator-Engineers and client-facing leads when the work needs business context.
What we need
- You have shipped production software and can explain the tradeoffs behind it.
- You are already paying for serious AI tooling, likely around $200 per month or more, because it makes you faster.
- You are an active Claude Code and Codex user with opinions earned from real usage.
- You can generate code with LLMs without becoming a passenger.
- You understand basic business concepts well enough to avoid building beautiful useless machinery.
Strong signals
If any of these describe you, the conversation will move quickly.
- Your GitHub, scripts folder, or local setup tells a story before you do.
- You have custom prompts, commands, agents, or harnesses because the defaults were too slow.
- You can ship a first version in a day and still care about the second-order consequences.
- You have strong taste about what AI coding tools are good at and where they still produce nonsense.