
What happened
The author's AI setup grew into a synthetic company with named roles like Henry, Margo Marketing, Voice Guardian and Cringe Editor, plus MiniClaudes, and it adopted the name AI Goblin Circus Inc.
Why it matters
He argues this corporate structure is not just roleplay but a coordination system of boundaries and review gates, since a capable prompt can impersonate a missing role.
Summaries like this, every morning.
The author traces the shift from his earlier workflow problem—sitting between separate AI systems as a "meat-based integration layer"—to a daily operation with departments, memos and review gates that appear without grand design, driven by operational failures. For example, discovering that the Voice Guardian had been impersonating a role with an unfinished file led to a role-integrity audit.
He distinguishes this from human mimicry: AI systems do not sulk or resent promotions, but they overhelp, hallucinate authority and produce plausible slop, so they still need governance. The difference he draws is that in a human company time is the scarce resource, while in an AI organisation context, role purity, auditability and repeatability become scarce instead.
He ends on a hypothesis: a company may be a coordination pattern rather than a building full of humans, and his system's drift toward one may be a clue rather than a gimmick. He explicitly notes he does not know whether others are building synthetic companies this way.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Anthropic's long-awaited IPO document, released Monday, shows more than $400 billion in future obligations for…

An Apollo Global Management-managed entity will buy Nvidia's next-generation Vera Rubin chips and server hardw…

Anthropic's S-1 shows more than half of customers spending at least $100,000 a year use both the Claude platfo…

The RealReal is updating its bot defense system so AI agents like Meta's Muse can no longer log in, add items…

AWS published a guide showing prompt caching for OpenAI models on Amazon Bedrock, with a six-turn GPT-6 Astra…

Google Research and the Stanford Accelerator for Learning launched "Polar Peaks," the third AI Quests lesson…
