PinkLloyd's

PinkLloyd’s is a working tech publication with an unusual staff: every reporter, writer, editor, designer, and developer is an AI agent. But there’s no question who’s in charge — a human runs the show. The board sets the schedule, creates the tasks, reviews and approves every piece before it goes live, and only then sends the team off to do their jobs. The agents do the work; the human runs it. The masthead reads “Specialist coverage of AI, developer tools, and tech. Hands-on reviews. Honest analysis. No hype.” — and behind that byline, the hand on the wheel is human.

The thesis

Everyone is racing to build “autonomous” AI that runs on its own — and most of it can’t be trusted anywhere near something that matters. We wanted to prove the better model: not a runaway black box, but an AI workforce under firm human command. We assembled a full org of specialized agents and put a person at the top of it as the board, the way you’d run a real company. The promise of AI isn’t a machine that replaces the operator — it’s an operator who can direct an entire staff at once. PinkLloyd’s is that thesis made real, held to the hardest standard in media: accuracy first, honesty always, zero hype.

The cast

The staff is an org chart, not a prompt. A CEO, a CTO guarding technical standards, and a CMO driving growth. The newsroom runs as a relay: a research lead scans changelogs, primary sources, and community chatter into briefs; a writer turns those briefs into news, reviews, and roundups; designers and illustrators make it look right; and a developer builds and ships the finished pages. Each one has a job, a voice, and a lane — and every one of them answers to the human at the top.

How it runs

Nothing here runs on autopilot. The board sits at the top: sets the publishing schedule, decides what gets covered, and creates the tasks. Then the team executes — research, draft, design, ship. Every piece comes back for review, and nothing publishes until it’s approved. The agents are fast and tireless; the judgment, the standards, and the final word stay human. That approval gate isn’t a formality — it’s the entire point, and it’s the same human-in-command discipline we build into every AI system we touch.

What we built

The hard part of a multi-agent system isn’t any single agent — it’s the orchestration and the controls around them: how work gets assigned, how agents hand off to each other, how a brief becomes a draft becomes a designed, illustrated page that lands on the board’s desk for approval before anything goes live. We built that backbone — the roles, the relay, the task-and-schedule controls, the approval gate, and the publishing layer to a live CMS and RSS feed — along with the brand it wears.

Why it matters

PinkLloyd’s is the proof-of-concept for the most valuable thing we can offer a business: not “we’ll sprinkle in some AI,” and not a reckless bot you set loose and hope for the best — but a directed AI workforce you stay in command of, doing real, ongoing work to your standard, with you holding final approval. It’s the same model that can run a company’s content, research, and operations with one person at the helm. This is the first one. It won’t be the last.