2LDW / THE SMALL PRINT, MADE READABLE
Privacy.
No account is required. Submitted videos and finished summaries are not private.
Updated October 19, 2026
What we save
We save the video ID, title, creator, thumbnail URL, generated summary, source type, caption language, dates, and a source hash. Older records may include claim assessments and citations. We also save the full transcript, caption timestamps and source label in a private database cache so retries and follow-up analysis do not need another download. The raw transcript is not published or exposed through public endpoints. Processing jobs also temporarily store their title and stage history. Finished summaries have public links and are kept until updated or removed.
AI processing
A private processing computer retrieves information from YouTube and sends the transcript to OpenAI through Codex for transcript analysis, using our existing ChatGPT sign-in. OpenAI processes this material under the policies and data controls applicable to that account. Local runs use ephemeral sessions and disabled transcript history; this does not promise zero retention by OpenAI. Do not submit private or sensitive information.
Earlier saved summaries used a model running on the processing computer. New requests use the service described above. See OpenAI’s privacy policy for its data practices.
Reader- and browser-provided transcripts
Pasted text and extension-imported text/metadata initially wait in server memory for up to one hour. Once the worker validates it, the full text is saved in the private transcript cache and sent to OpenAI for analysis. Saved supplied text is kept separate from captions retrieved from YouTube. The generated summary is public, appears in the library, Trending, related summaries, and sitemaps, and can be indexed by search engines. Do not submit private or sensitive information. Reader- and browser-provided summaries are labeled because the supplied text has not been verified against the video. A restart before the input has been saved can interrupt a pasted-text request.
Usage limits and security
A keyed, one-way hash of your network address or IPv6 network supports rate limits. Usage records are removed after 31 days when new requests are admitted. Server and Cloudflare logs may include IP addresses and request information for security and operations. Those logs are separate from summary data. We do not install advertising trackers.
Site analytics
We use Fluxly analytics on ordinary public pages to understand visits and activity. The extension handoff omits this analytics script. The script sends page paths and titles, referring URLs, campaign parameters, screen and viewport sizes, and page-view and activity events to analytics.fluxly.net. It creates random visitor and session identifiers using your browser’s local storage and session storage. Like other web requests, analytics requests also disclose your network address to the receiving service. The installed tracker does not read form fields or send pasted transcripts.
You can clear these identifiers by clearing site storage, or prevent analytics requests by blocking analytics.fluxly.net. Read our storage information for more details.
Summary views and Trending
We count approximate summary reads to order the Trending page. We use a keyed hash that changes each UTC day to limit repeat counts from a network, without setting a tracking cookie. The view-count table does not store your raw IP address or browser identifier. These daily hashes are removed once they are more than one UTC day old when another counted page is opened, so deletion can take longer while the site is idle. Aggregate counts remain with each summary.
Services that receive information
Our existing hosting provider runs the website and database. Cloudflare provides traffic protection and Turnstile verification, which receives browser and network information to check new requests. Thumbnails load from YouTube, which receives a normal image request from your browser. Clicking a video thumbnail loads a YouTube player in a popup using its privacy-enhanced domain. The player connects to YouTube only when opened, and YouTube’s rules apply to playback. A separate link lets you open the video on YouTube.
Read Cloudflare’s Turnstile privacy notice and Cloudflare’s privacy policy. Provider processing locations and retention rules can differ from ours.
Retention and your choices
Public summaries and privately cached transcripts are retained until removed. The cache can also retain a successfully retrieved transcript when its analysis fails, allowing a later retry without downloading it again. Completed and failed job records are eligible for cleanup after one day during subsequent admissions. Daily database backups are retained on the same server for 14 days, so a removed record can remain in a backup until it expires. The site does not have off-server disaster recovery configured.
You can browse saved text without submitting a video or transcript. Your browser can restrict cookies and external resources, though doing so may affect security verification or thumbnails. Search engines and other sites may retain copies of public pages outside our control.
Extension reminder
Updated extensions signal their presence locally to hide the install reminder, without reading page content, storing data or making a network request. For older releases, the website remembers the date of a successful handoff for 90 days; this contains no video ID or transcript and is not sent to the server. A live signal clears that old preference. See our storage information.