Privacy Policy

open·intuition is a platform where people publish explanatory artifacts (videos and interactive explainer sites) about scientific papers. This policy covers the website at openintuition.org, its API, and the open-intuition browser extension. Open Intuition is the service name; the legal operator will be identified before this draft takes effect. Privacy requests go to hello@openintuition.org.

• Account data: your Google account identifier, verified email address, profile information supplied during sign-in, and your Open Intuition handle. Google sign-in requests your basic identity, email and profile; it does not request access to your Gmail messages or Google Drive files.
• Access: account access and any existing waitlist or approval records. With open access enabled, a verified Google account lets you participate without founder approval; signed-out visitors can browse. If private beta is enabled, requests and approval-email status support admission.
• Your contributions: drafts, uploaded files, immutable artifact revisions, votes, comments and remix notes. Comments include notes you or your connected agent record on a revision (your feedback, decisions and preferences while creating it), and feedback you give on an explanation’s page or in the extension. Feedback is private to you and the explanation’s creator unless you choose to post it publicly; notes on your own work are private to you unless you make them public. Saved and recently visited papers are associated with your account so you and your connected agent can resume work.
• Reports and filed papers: when you report a paper or explanation, we store your account, the reason and your note so the administrator can review it; other members don’t see who reported. When your agent files a PDF that no registry lists, the paper record notes your account as the one that filed it, which the administrator sees when reviewing reports.
• Video playback activity: while signed-in members watch a video on the website or in a kept-open extension preview, we record each visit’s playback log, even with Feedback closed: play, pause, seek, speed changes, buffering stalls and whether the tab was in the foreground, each with the time, time zone and video position. Turn off “Share playback patterns” in the player’s reaction-settings menu to stop future recording on that surface in that browser; the website and extension have separate settings. Muted hover previews and signed-out visits are excluded. Visit logs and optional “I’m confused” / “This clicked” markers are linked to your account and the exact revision, kept for a rolling 30 days, and deleted in periodic batches after expiry. The creator and their agent receive only aggregate counts in ten-second sections (watched, replayed and sped-up time, rewinds, pauses, speed changes); viewer-facing counts and timelines are currently hidden. Individual visit logs are never shared with creators or other viewers. We do not send this data to AI labs through this feature.
• Feedback activity: while the Feedback box is open, we also record play, pause, seek, speed, dictated text and typing changes with the time, time zone and video position alongside submitted feedback. This detailed log is available to you, the creator and their authorized agent, never publicly. Optional research use of eligible records would require the separate permission described below. Closing the box stops this detailed log; the separate playback-sharing switch controls the always-on playback log.
• Dictation uses your browser’s speech recognition. When you press Dictate, your browser turns speech into text; Chrome sends the audio to Google’s speech service for this, and Safari may do it on your device. Open Intuition receives only the resulting text and never receives or stores audio.
• X account (optional): if you link your X account on your account page, we keep your X user ID and username, nothing else from X, so a request you make by tagging or messaging @open_intuition can start your agent and credit you. You can unlink it at any time.
• Mentions on X: when anyone tags @open_intuition on X, we read that post and the posts it replies to or quotes to find the paper, keep a record of the request (the post, its author and the paper), and reply publicly on X with the explanation video.
• Direct messages to @open_intuition: we act only on DMs from members who linked their X account. From those we keep the first post the DM shares and any paper links in it, never the message text, and post the explanation video publicly from @open_intuition once it is made. From anyone else we keep only the sender’s X ID and username, to show that a DM arrived.
• Authentication: session cookies and the credential and grant records needed to sign you in and authorize connected agents. We do not ask you to create or paste an Open Intuition API key.
• Operational data: request information needed for security, rate limits and diagnosing failures. Sign-in records may include your IP address and browser information. Video visit summaries are described above; we do not set an anonymous view-tracking cookie.
• Activity: for each day you use the website or a connected agent, we record that your account was active that day (the first and last time, and whether it was through the website or an agent) so the administrator can see how many members use Open Intuition and who was recently active. This is one record per account per day, not a log of pages, and it is deleted after about 13 months.
• Aggregate page views: the website counts visits with Cloudflare Web Analytics, which sets no cookies, does not identify you and does not track you across sites. It reports totals such as page views, referring sites, countries and devices.

• Sign-in tokens are stored locally. When you connect your account, OAuth tokens are kept in your browser’s extension storage on your device. The extension never sees your password; account sign-in and connection consent happen in your browser.
• Paper lookups are explicit. While the extension’s side panel is open on a recognized scholarly page, or when you ask it to look up another page, the extension sends that page’s URL (and the page’s own title, author, date and abstract tags) to openintuition.org to look up the paper’s gallery. On other pages it sends nothing until you ask.
• Auto-tracking is whitelisted, toggleable, and minimal. If you leave “Remember papers I read” on, the extension sends the URLs of pages you visit only on a fixed whitelist of scholarly article pages (preprint servers such as arXiv, bioRxiv, SSRN and NBER; machine learning venues such as OpenReview, ICML, NeurIPS, ICLR, CVPR, PMLR and the ACL Anthology; PubMed and PubMed Central; and journal publishers such as Nature, Science, PNAS, Cell, PLOS, eLife, Elsevier, Springer, Wiley, Oxford, Cambridge, ACM and IEEE; and Substack posts; the full list is in the extension), together with that page's own title, author, date and abstract tags so papers the server can't read itself are still identified, to power “the paper I was just reading” in your AI agent. Nothing outside that whitelist is ever sent. Repeat visits are deduplicated for six hours. You can turn this off on your account page at any time; while it is off, nothing is recorded, and the extension sends at most one whitelisted URL an hour to learn the setting, which the server discards.
• Sharing on X happens only when you click. When you click Share on X on an explanation, the extension opens X’s post page in a new tab with the post text and link. In that tab only, it downloads the explanation’s video from Open Intuition and adds it to the post, which is why it asks for access to x.com. It never presses Post, reads nothing else on X, and sends nothing about your X account to Open Intuition.
• Paper context stays limited. The extension sends paper URLs, not the full contents of pages, to Open Intuition. It does not inject ads. Its interface may load fonts from Google Fonts.

We use account information and contributions to provide sign-in, paper lookup, hosting, review, publication, attribution, saved papers, notifications and authorized agent connections. We use activity and feedback to provide the disclosed viewing and feedback features, understand product use, improve explanations and the service, investigate reports, prevent abuse and diagnose failures.

We also intend to use eligible contributions to research how people understand information and to develop, evaluate and improve products, services and technologies that help explain ideas, including AI systems. This work may serve research and commercial purposes. Published explanations may be reused under their applicable licenses. Those licenses do not grant permission over private feedback or personal viewing histories.

Use of private or account-linked contributions beyond the disclosed platform features, including selected preference notes, votes, feedback and eligible playback activity for research, would require separate participation under the Research Contribution Agreement. Enrollment and research exports are not currently available. The enrollment notice must identify the data, purpose, recipients, retention and withdrawal choices before use begins.

That optional program may include paid licensing to AI developers or research partners for the stated purposes. External sharing would require a separate affirmative choice. Such a transfer may count as a sale under some privacy laws even if direct identifiers are removed; applicable notice, consent and opt-out rights still apply. Service use or accepting the Terms alone is not research enrollment, and removing identifiers does not automatically make data anonymous.

The current platform has no advertising or data-broker integration. A future new purpose or recipient must be disclosed and have an appropriate legal basis before the processing starts; this policy is not blanket permission for unspecified uses.

Our use of information obtained through the extension complies with the Chrome Web Store User Data Policy, including its Limited Use requirements. We limit that data to providing or improving the extension’s disclosed purpose and permitted related operations. Browsing activity is used only for the disclosed user-facing paper features. Transfers and human access must satisfy those rules.

These restrictions also apply to data derived from extension records, including aggregated, anonymized or de-identified forms. Extension data is excluded from the proposed external research dataset program; an opt-in cannot waive Chrome’s restrictions. Data from other services, including X, remains subject to its source restrictions. We do not treat it as freely licensable OI data.

Public visitors receive live published explanations and public profile information. Signed-in participants receive the comments and feedback their permissions allow. Creators and their authorized agents receive feedback and viewing summaries as described above. Service providers process information to run the service, and your connected agents receive the information you authorize them to access.

Any future research recipients are subject to the separate enrollment notice and agreement. We may disclose information when legally required, to address security or rights violations, or in a lawful business transfer with required notices and permissions. Source restrictions, including Chrome Limited Use, continue to apply; a business transfer is not a waiver of them.

The application runs on Cloudflare Workers, with account and application data in D1, OAuth protocol state in KV, files in R2, and transactional notices sent through Cloudflare Email Sending. Google provides sign-in and the website loads fonts from Google Fonts. Your chosen agent and video tools have their own data policies; they receive the information you provide or authorize them to access. When you ask us to import a finished video, our server retrieves that file from its delivery provider and stores a copy in Open Intuition storage.

When open access is enabled, anyone can browse the homepage, Explore, papers and public profiles. Google sign-in lets anyone participate without approval. When open access is disabled, these pages require Google sign-in and founder approval. A published explanation’s live revision is always viewable by anyone with its link. Voting, publishing, bookmarking and requesting explanations require sign-in. Comments are for members; earlier revisions and private feedback remain private. Published work and every file bundled into its live revision are shared under CC-BY, with downloading and remixing enabled. Drafts and saved papers are not gallery entries. Reviewing a draft requires membership and an unexpired review link, or ownership. Playback and downloads use temporary signed file links; anyone holding one can use it until it expires. Share review and file links only with people you want to see the work. Neither link grants permission to publish.

Playback logs and section markers have a rolling 30-day window and are deleted in periodic batches after expiry. Daily member-activity records are deleted after about 13 months. Drafts, revisions, comments, feedback, profile information and saved papers are retained while your account and content remain on the service, subject to deletion requests. Authentication and temporary capabilities expire or can be revoked. Limited records may be retained where needed for security, legal obligations or resolving disputes. A separate research notice must specify any retention of an eligible research copy; this draft grants no indefinite research retention.

Where data-protection law requires a legal basis, core account and hosting processing is based on providing the requested service; security and proportionate service analytics may rely on legitimate interests subject to the applicable balancing requirements; legal compliance relies on legal obligations. Optional research processing relies on the separate informed choice described above. Other processing requiring consent will not be justified merely by accepting general Terms.

Providers and recipients may process information outside your country. Any transfer requiring safeguards under applicable law must have them in place. This draft does not claim that a research recipient or international research transfer has already been approved.

You can disconnect the extension, revoke connected agents in account settings, and turn off extension auto-tracking. The player’s playback-sharing switch stops future playback recording on that surface in that browser; it does not remove previously saved summaries or explicit markers. You can undo a section marker. To request account or content deletion, or ask a privacy question, email hello@openintuition.org. Published work may already have been downloaded or remixed by others; removing it from this service does not remove those independent copies.

Depending on the law that applies, you may have rights to access, correct, delete or obtain a copy of your personal data, object to or restrict processing, withdraw consent, opt out of certain sharing, or complain to a data-protection regulator. We may need to verify your request and will respond within applicable legal deadlines. Exercising these rights does not waive other rights.

A future research program must support withdrawal and deletion as described in its separate agreement. We cannot promise to reverse completed model training or retrieve independent public copies, but this does not override applicable personal-data deletion obligations.

When this draft is activated, we will identify its operator and effective date. Material changes will be communicated through an appropriate service notice, with a new effective date and renewed permission where required. Publishing new wording alone does not expand permissions over earlier private data.