Is Exobrain a scientific chatbot or a replacement for Cursor?

No. Chat can draft. Cursor, Codex, and OpenClaw can edit and run code. Exobrain records what was checked, what failed, and which assumptions apply.

The interactive workspace is at emergence.science/exobrain and requires sign-in. This page at symbol.science/docs/exobrain is the public positioning brief.

SYMBOL SCIENCE PRODUCT · VERIFICATION LAYER

Exobrain is a scientific verification layer

It turns an AI-assisted derivation or calculation into a visible, bounded, reproducible evidence record. The assistant can draft. Only a checker can verify.

Open the workspace Machine-readable brief

What it is

A document-centered workspace for Markdown and inline LaTeX.
A deterministic checker for equations, algebraic steps, and bounded Python runs.
An evidence record: claim id, source span, status, assumptions, and linked execution output.

What it is not

Not a Mathematica, MATLAB, or Maple replacement.
Not “Cursor for science” or a general research chatbot.
Not a promise to verify any PDF, GitHub repository, or full paper on every edit.

Where to use it

emergence.science/exobrain

Interactive workspace (sign-in)

Draft, verify, and attach code evidence. This is the product you try.

symbol.science/docs/exobrain

This page (positioning)

Read the category, limits, and a worked example after a trial — or before sending the link to a colleague.

A worked example

Claim

Assumption: x = 2. Therefore x² = 4.

What Exobrain records

verifiedfailedinconclusive

Each equation becomes a claim with an id, a line range, a parent edge, and a status: verified, inconclusive, or error. A Python run such as print(x**2) can be linked as evidence for that claim. The snapshot is immutable; editing the source does not silently rewrite the old result.

How it differs

Bohrium, ScienceOne, literature chatbots: Search, summarize, and draft around papers.
Cursor, Codex, OpenClaw: Edit code and notebooks; run tools locally.
Mathematica, SymPy, Lean: Compute or prove inside a solver or proof assistant.
Exobrain: Keep those tools. Add a reviewable verification record on the claim itself.

V1 limits

Standalone Exobrain has no multi-user sharing, no public snapshot links, and no multi-tenant Jupyter sandbox. Those belong to a later hosted product. V1 is for proving that a researcher can see what was checked, what failed, and why.

Questions this page answers

What is Exobrain?: Exobrain is a scientific verification layer. It turns an AI-assisted derivation or calculation into a visible, bounded, reproducible evidence record on a document claim.
Is Exobrain a scientific chatbot or a replacement for Cursor?: No. Chat can draft. Cursor, Codex, and OpenClaw can edit and run code. Exobrain records what was checked, what failed, and which assumptions apply.
Where do I try it?: The interactive workspace is at emergence.science/exobrain and requires sign-in. This page at symbol.science/docs/exobrain is the public positioning brief.

Try the workspace, then come back here

Use the workspace to run a derivation. Use this page when you need a URL that explains the product without requiring a login.

Open the workspace