BETA PRIVACY NOTICE

Privacy without the mystery.

Worked IRL needs real-world context to do useful research. This page explains what the beta stores, how accounts and optional attachments are handled, why information is retained, how shared cases work, and how product behavior and real-world outcomes can improve the service.

UPDATED SEPTEMBER 26, 2026

What we store

When you use Worked IRL, the beta may store:

The problem, task, question, model or part information, symptoms, and other details you submit in a research hunt.
The researched answer, source receipts, and system metadata needed to restore the case.
Follow-up messages you send in the same hunt and the responses returned.
Your outcome vote such as worked, partly worked, or did not work, plus any detail you choose to add.
For an attachment-assisted hunt, concise observations derived from a photo or PDF when they are useful for preserving research context.
If you create an account, authentication information handled by the account service plus profile details you choose to provide, cross-device hunt links, follows, contributions, and related reputation evidence.
Limited first-party beta product events described below, associated with a random browser identifier rather than problem text or quiz answers.

Research-rate limits are also enforced with a separate signed functional cookie. Hosting and infrastructure providers may maintain ordinary technical request logs as part of operating their services.

Optional photos

If you attach a photo, the browser resizes and re-encodes it before submission and the image is sent to the AI service for visual analysis as part of that hunt.

The raw attached photo is not intentionally written to the Worked IRL case database. After the research request finishes, the saved case may retain a short text description of useful visual observations so later follow-ups can still understand relevant context without retaining the original image.

A photo can contain information you did not intend to share. Check the image before submitting it and avoid including faces, addresses, documents, account numbers, medical records, or other sensitive details unless they are genuinely necessary to the request.

Optional PDF attachments

If you attach a PDF manual, instruction sheet, report, or similar document, the file is sent to the AI service so relevant facts, procedures, specifications, warnings, recommendations, or error information can influence the hunt.

The raw attached PDF is not intentionally written to the Worked IRL case database. The original filename is kept only in the local attachment picker on your device for preview and is replaced with a generic label before the attachment is passed into the hunt. The saved case may retain a short text summary of useful document-derived observations so later follow-ups can preserve research context.

Before attaching a PDF, make sure it does not contain private account information, payment information, medical records, personal addresses, or other sensitive material that is unnecessary to answer the request.

How we use it

Stored case data is used to provide the service: restore a hunt, maintain follow-up context, connect signed-in history across devices, prevent abuse, troubleshoot product problems, and measure whether the research is producing useful results.

Outcome feedback can also become a private first-party signal for future, closely matched hunts. Positive outcomes can support a closely matched approach. Detailed failed outcomes can help avoid blindly repeating the same failed move. A single result is treated as anecdotal evidence, not universal proof.

Prior-user outcome signals are used privately by the research engine. They are not shown as public source receipts, and the product is instructed not to identify or quote the prior user.

First-party beta measurement

During beta, Worked IRL may record a small set of product events so we can tell whether people actually use, finish, share, repeat, return to, and close the loop on parts of the product. Current event types cover IRL IQ category selection, challenge starts and completions, share or friend-challenge actions, real research-hunt starts and completions, Hunt reopens, outcome-prompt views, saved outcomes, detailed outcomes, My Hunts visits, read-only Hunt sharing, and changes between private and community visibility.

These events use a random server-generated browser identifier stored in a server-only first-party cookie for up to 90 days. An event may also reference the internal ID of the related Hunt so Worked IRL can understand the return-to-outcome funnel across visits. The product-event stream does not intentionally store the text of your Hunt, the detail you wrote in an outcome, your IRL IQ answers, your IP address, an advertising identifier, or a device fingerprint.

This measurement exists to answer product questions such as “do people finish the challenge?” and “do challenge users later try a real hunt?” It is not used for cross-site advertising or behavioral ad targeting.

Community visibility

Account-linked Hunts are private by default. If the account owner explicitly opens an individual Hunt to Community, Worked IRL may surface that Hunt inside its community features.

An unresolved Community Hunt may be routed to signed-in members whose prior contributions have relevant owner-confirmed outcomes. A Community Hunt with a reported outcome may appear in the read-only Community Finds library. Community surfaces are marked noindex during the beta.

Community discovery uses a separate privacy-reduced Hunt view. It automatically redacts obvious email addresses, phone numbers, Social Security number patterns, payment-card candidates, IP addresses, and prose URLs, and it does not intentionally show the Hunt owner's handle. This is different from Share by link, which is the fuller read-only copy the owner deliberately sends to someone.

Automatic redaction is a safety layer, not a guarantee. Before opening a Hunt to Community, review the saved problem, answer, follow-up history, outcome detail, and contributions. Your sign-in email and raw attached photos/PDFs are not intentionally shown on Community surfaces.

Shared case links

If you use the Share control, Worked IRL creates a read-only URL with a random sharing token. Anyone who has that active link can view the information contained in the shared case, including any saved outcome you reported for that Hunt. Share by link is separate from Community discovery and does not use the Community privacy-reduced presentation.

Shared case pages are marked for search engines not to index or archive them, but an unlisted link is not the same thing as encrypted private storage. Do not put passwords, Social Security numbers, payment-card information, private medical records, or other highly sensitive information into a case.

You can revoke a Share-by-link URL from My Hunts. Revocation removes that sharing token, so the old URL stops working. Revoking a private share link does not change Community visibility, and leaving Community does not silently revoke a separate link you explicitly shared.

The share link does not carry the private edit capability used by the browser that created a protected case. Raw attached photos and PDFs are not included in shared case pages because the beta does not retain those raw files with the saved case.

Functional and measurement cookies

For newly created protected cases, Worked IRL sets a server-only functional cookie that lets the original browser continue, vote on, or delete that case. A separate signed functional cookie is used to enforce beta research-rate limits.

If first-party beta measurement is enabled, a separate server-only random browser identifier may be stored for up to 90 days so product events from the same browser can be understood as a basic funnel. These cookies are not used for advertising or cross-site tracking.

Service providers

Worked IRL relies on service providers to operate the product. Current infrastructure includes:

OpenAI for AI-assisted research, photo analysis, PDF analysis, and answer generation.
Neon for the application's PostgreSQL database and account authentication infrastructure.
Vercel for application hosting and delivery.

Source receipts may link to independent websites, forums, manufacturers, videos, or other third parties. Their privacy practices are their own.

Access and portability

Signed-in users can download their account-linked Hunt archive from My Hunts. The archive includes the user's saved Hunt problems, researched answers, source receipts, follow-up messages, outcome feedback, dates, and current private/community visibility state.

The Hunt archive does not intentionally include the user's sign-in credentials, email address, authentication user ID, browser measurement identifier, or internal model-usage telemetry.

Retention and deletion

During beta, saved cases remain available until they are deleted or removed as part of product maintenance. The original browser for a newly protected case, or the account owner when a hunt is account-linked, can remove the case from the application database through supported controls. Stored follow-ups, derived attachment observations, and outcome feedback tied to that case are removed with it according to the database relationships.

Deleting a case also makes its read-only share link stop working because the shared page reads the same underlying case. Anonymous product-event rows are a separate measurement stream; deleting a hunt removes the database link from any event that referenced that hunt but does not necessarily remove the anonymous aggregate event itself.

What we do not do today

Worked IRL does not currently sell user case data or conversation content to advertisers.
Worked IRL does not use third-party advertising pixels or cross-site behavioral tracking as part of the beta measurement layer described above.
The beta does not intentionally retain raw attached photos or PDFs in the Worked IRL case database.
The beta does not intentionally retain the original PDF filename as hunt data; the attachment picker replaces it with a generic label before submission.
Real-world outcome feedback is not treated as independently verified evidence merely because a user submitted it.

Children

Worked IRL is a general real-world research and problem-solving service and is not designed for children under 13. Do not submit a child's personal information through the service.

Changes to this notice

The beta is evolving quickly. If the way Worked IRL collects or uses case data changes materially, this notice should be updated to match the product rather than relying on outdated boilerplate.