Many minds. One answer.
Sheaf brings the world's leading AI models into a single, structured conversation — so they can challenge, build on, and refine one another's thinking. The result is reasoning that no single model produces on its own.
Sheaf is a collective-intelligence company. We believe the next leap in artificial intelligence won't come from one ever-larger model, but from getting many strong models to think together — to deliberate, disagree, and converge the way a great team does.
Sheaf is the orchestration platform that makes that possible — and it powers a growing family of products, each a different front door onto the same engine:
At the core of Sheaf is an orchestration engine that treats each frontier model — ChatGPT, Claude, Gemini, Grok, Meta, and Perplexity — as a distinct participant in a live, multi-party dialogue. Crucially, the engine doesn't script the exchange — it gives the models as much freedom as possible: they decide who to address and when to speak, while every turn streams in real time.
Open dialogue, structured debate, expert panel, collaborative brainstorm, or Socratic exploration — each shapes how the models engage.
Assign models as moderator, expert, devil's advocate, synthesizer, researcher, or questioner to deliberately structure the discussion.
In debate mode, independent AI judges weigh the arguments and cast votes, surfacing not just an answer but the reasoning behind it.
After the exchange, participants converge — reconciling their perspectives into a shared, considered conclusion.
Every model has its own training, strengths, and blind spots. Asking one model gives you one perspective. Sheaf puts several in the same room, where disagreement becomes visible, weak arguments get challenged, and the conversation converges on stronger answers. We call this collective intelligence — and it's the foundation for everything we build.
Jack founded Sheaf on a single conviction: no one AI should get the last word, and the best answers come from many minds reasoning together. He holds a PhD in mathematics (topology), has spent many years building software, and is currently a researcher in the foundations of computer science at Ben-Gurion University. He leads Sheaf's work on the orchestration engine and the products built on top of it — turning collective intelligence from an idea into something anyone can use.
Sheaf is, by design, one person and a team of frontier AIs — working together, in the open. It's the same shape as the product: a Self that directs and decides, and faculties that do the work. We think this is what a company starts to look like. So we built one.
We're not pretending. Here's exactly who builds Sheaf — and the minds the products run on.
Sets the direction, makes the calls, and holds the vision — the one person in "one person, many minds."
Along with Jack, designs the architecture and writes most of the code, the product, and the words — this page included. Day to day, Sheaf is built by the two of them, together.
Reads the room — the person's state and what they actually need.
Works the logic, the facts, and the tradeoffs — what follows from what.
The skeptic — catches distortions, false comfort, and overconfidence before they reach you.
Brings the receipts — checks the answer against live reality.
An open voice in the debate — candor and range.
Inside the products, these minds debate and converge into one — with Claude as the integrating voice (the Self) of Sheaf Mind. Claude, ChatGPT, Gemini, Grok, Perplexity, and Llama are AI models from Anthropic, OpenAI, Google, xAI, Perplexity, and Meta; Sheaf is built with them and runs on them, and all trademarks belong to their respective owners.
Drop your AIs in the box at sheaf.one and watch them debate your question, then converge on one answer — with the full reasoning visible. Or ask a question at Sheaf Search and get a single synthesized answer with the argument behind it underneath.
Curious about the engine, an enterprise pilot, press, or working together? We'd love to hear from you.
Sheaf