Ellev8

// How I work

How the work moves, and where I stop it.

I work with a system of AI agents. Each one owns one part of the work: research, strategy, build, copy, or review. I brief them, and the work stops twice for me. Nothing passes either stop without my call.

A dirt trail switchbacking up a grassy foothill at dawn, mist in the valley below

The agents

Do the research, the drafts, the build, and the checks. They verify that the work is correct.

JD McCulley smiling at a MacBook on a wooden desk against a white brick wall

Me

Make every call that matters: which problem is real, which direction, and whether it's good enough to ship.

You

Get a fixed scope, dated weeks, and one person accountable for every decision.

// The roster

Eight agents, four stages, two gates.

Each line under “On this site” is what that agent did on the first build of this website, September 2026.

  1. Runs the engagementConductor
    OwnsThe shared engagement state, routing, and the run log.
    On this siteOpened the engagement, loaded the brief and the proof points, ran Scout and Analyst in parallel, and wrote the Gate 1 package.
  2. DiscoverScout
    OwnsMarket, competitor, and pattern research.
    On this siteRead 44 sources and found the judgment thesis already shared by four others.
  3. DiscoverAnalyst
    OwnsSynthesis with citations and the buyer’s decision criteria.
    On this siteRanked how each buyer decides, moved the thesis out of the hero, and put the numbers first.
  4. DefineStrategist
    OwnsDirections, site maps, and the buyer’s journey.
    On this siteWrote three directions with site maps and recommended Receipts First.
  5. Gate 1I pick the direction.On this site: I picked Direction 2.
  6. BuildBuilder
    OwnsTokens and front end, responsive and accessible from the start.
    On this siteBuilt the homepage, Scope, and How I work pages, phone width first, then rebuilt the homepage results first.
  7. BuildEditor
    OwnsEvery word, in the client’s voice, every claim traced.
    On this siteWrote every word on the first build and kept a source for every number.
  8. VerifyCritic
    OwnsReview against the brief, through the people who will judge it.
    On this siteReviewed the pages as three buyers and a craft reviewer, scored them, and returned a ranked list of fixes.
  9. VerifyAuditor
    OwnsAccessibility, heuristics, token discipline, and reduced motion.
    On this siteReturned twelve blocking findings in the first two rounds, all fixed, then passed the rebuild with zero.
  10. Gate 2My taste pass and sign-off before anything ships.On this site: I signed off before launch.

The agents verify that work is correct.
Whether it is good is my job.

The same pattern, every weekday · The Morning Scan Desk

Design the decisions, then automate around them.

My own mornings run on it: scheduled agents read 18 sources, rules screen out 63% with a logged reason, and three decisions stay mine. About two minutes of my attention a day.

See it in the work

See the system on your problem.

Scope runs on the same agents and the same two gates. Two weeks, a fixed fee, credited if you proceed.