Skip to content

The Method

How we work

Eight short essays on the operating policy of a senior .NET agency working with AI agents. Each one is a rule we enforce. Each one is a rule we have changed our mind about at least once.

  1. 01

    Specs before code

    Every non-trivial change begins with a written spec — the constraints, the failure modes, the acceptance criteria. Code follows the spec, not the other way around.

  2. 02

    No story points

    We do not estimate in story points, T-shirt sizes, or Fibonacci. Estimates are wall-clock days, given by the engineer who will do the work, after the spec is written.

  3. 03

    Two-week cycles

    Sprint-based engagements run on hard two-week cycles. Day 1 plans, Day 14 ships, Day 15 the client decides whether to continue. No fuzzy quarters, no rolling roadmaps.

  4. 04

    CLAUDE.md per project

    Every repo we touch ships with a CLAUDE.md at its root. It is the contract between the codebase and the agents that work on it — conventions, boundaries, what to never do.

  5. 05

    PR descriptions over standups

    We do not run daily standups. The pull request is the standup — what changed, why, what risk, what to test. The team reads PRs in their own time.

  6. 06

    The AI policy

    Agents draft. Humans rewrite. No AI-written code or content ships without a senior engineer rewriting at least 30% of the words. No AI-generated hero images, ever.

  7. 07

    The hiring bar

    Eight years of production .NET, one open-source contribution we can read, one written sample of technical writing. We hire slowly. We do not hire juniors.

  8. 08

    The $80 paid first call

    Discovery calls are eighty dollars. The fee is small enough to not matter to a serious buyer and large enough to filter out the rest. We refund it if we are not the right fit.