Terms of Use
Version 2026-08-24. Accepting these is a condition of holding an account.
The short version
- 1.This is a research tool for faculty, not a product and not for students. No warranty, and it may change or break.
- 2.The library is not yours or ours to redistribute. It is summarised from other people's repositories and papers. Study it, cite the originals, do not republish or bulk-export it, do not train models on it.
- 3.Everything a model writes here can be wrong, including confidently. Treat it as a starting point for your judgment, never as a finding.
- 4.Your account is yours alone, and what it can do depends on its role. Running models costs real money — ours by default, yours if you supply a key.
- 5.Your consultations are private unless you agree to share one. An administrator can read any of them.
What this site is
GeWu.ai is a research environment for faculty studying and designing AI education systems, operated by the TZStats aiX Lab at Columbia University. It is provided for academic use, without warranty, and it is not a product. Features may change or be withdrawn, and periods of unavailability should be expected.
It is not a student-facing tool. Do not use it to deliver instruction to students, and do not enter student work, grades, or other education records into it.
Your account
It belongs to one person
Yours, named. Do not share it or pass on your credentials.
What it can do depends on its role
Signing in makes you a reader: browse the library, read your own consultations. A user can also run Elements Lab runs, probes and analyses of new repositories — all of which call commercial models and are billed to this project. An administrator can also run consultations, run the pipelines, and delete material.
Use it in proportion to your purpose
If your role can run models, keep the volume reasonable. Automated or bulk querying needs asking first. Access can be suspended where use falls outside what is described here.
Who pays, and for what
By default, everything you run is billed to this project. It is funded by the lab, without external support, which is why running things needs a role and reading does not.
If you supply your own API key in Settings, the Simulator and the probes run on it instead. Four things follow from that, and they are the reason this clause exists:
- That usage is governed by your agreement with that provider, at their prices — not by these terms.
- We do not cap, monitor or estimate your spending. A long run costs what it costs, and we cannot see the bill.
- The key stays in your browser and we never store it, so we cannot revoke or rotate it for you.
- If the key belongs to your institution rather than to you, their rules on its use apply and are yours to observe.
We will not fall back to our keys if yours is missing or rejected — the request is refused instead.
Your consultations
Private by default
Nobody else sees a session of yours unless you consent to sharing it and an administrator then publishes it. Withdrawing consent un-publishes it.
An administrator can read any session
Support, moderation and evaluating the research all depend on it.
Deleting moves it to a trash
It leaves your list at once, but an administrator can still see and restore it until it is purged. Ask if you want one gone for good.
Starting one needs the role, or the word
The Workshop is intended for live sessions with Tian Zheng at this time. Starting or continuing a consultation needs the administrator role, or the spoken word issued for a session in progress.
The material in the library
The library holds structured summaries and extracted design knowledge derived from third-party repositories, papers and documentation. That underlying material belongs to its authors. It is made available to you here for scholarly study, and is not yours or ours to redistribute.
So: do not republish, mirror, or bulk-export the contents of this library, and do not use it to train or fine-tune models. Quote it in your own scholarship as you would any source — attributing the original work rather than this site. Access is controlled for exactly this reason.
What the models produce
Much of what this site shows you is generated by large language models. It can be wrong, and it can be confidently wrong. Analyses, critiques, simulated conversations and design plans are all starting points for your judgment, not findings.
Nothing here is an evaluation of the quality of any project, tool or vendor, and nothing here should be represented as one. Where the interface states what a number does and does not mean, that statement is part of these terms.
This is a research project under continuous development. It changes without notice, parts of it appear and are withdrawn, and the library is re-curated as it grows. Reviewing what you take from here — a source, a summary, a generated plan, a claim about somebody else’s system — for accuracy and validity before you teach from it or cite it is your responsibility. It is not something this site has done on your behalf, and nothing in the interface should be read as saying otherwise.
What you write here
You keep whatever you write. By entering it you allow us to store and process it in order to run the site and evaluate the research — the privacy note sets out which parts are kept and for how long. Consultation transcripts, design plans and any tags you save are kept; Simulator sessions are not.
Do not enter personal data about third parties, anything confidential, or anything you are not free to share with a commercial model provider. What you type is sent to OpenAI or Anthropic to be answered — including what you type into the search box, which is matched by meaning and so has to be sent out to be understood.
Changes, and getting in touch
These terms carry a version. If they change materially you will be asked to accept the new version before continuing, and a record of which version you accepted is kept. Questions, objections, or a request to have your material removed: gewu.z2.ai@gmail.com.