Houses of Thought
WORKING DRAFT · This document is under legal review. Its substance reflects how the product works today; entity, jurisdiction, and effective date will be finalized before launch.

Privacy Policy — Houses of Thought

Status: Working draft — under legal review. An effective date will be set when this policy is finalized.

This Privacy Policy describes how Houses of Thought ("we," "us," "our") collects, uses, and shares information when you use Houses of Thought (the "Service"), a web application for building structured reasoning ("houses"), individually or in a classroom setting.

1. Information We Collect

1.1 Account information

When you create an account, we collect through our authentication provider (Supabase):

  • Email address and a password (the password is handled by Supabase Auth; we never store it in plaintext).
  • Account type — you choose one of *Standard*, *Student*, or *Teacher* at signup.

We do not collect your name, date of birth, or age at signup, and the Service does not currently verify age (see Section 7 on children's privacy).

1.2 Profile information (optional, user-provided)

On your profile page you may choose to provide:

  • A username (publicly unique within the Service).
  • "About me", current project, role, and location free-text fields.
  • Four "perspectives" fields (*biological, social, familial, individual*) describing your personal background and viewpoint. These are reflective, free-text fields; depending on what you write, they may contain sensitive personal information. You control what, if anything, you enter here.

1.3 Your houses (reasoning content)

The core of the Service is content you create in the house builder. For each house we store: a title, the question it addresses, concepts, watchpoints, perspectives (name, summary, strength), evidence items (text, source, and optionally a citation URL), assumptions, implications, the house's mode (Learn/Decide), status and completion progress, timestamps, and — if you use the AI context interview — a distilled AI context (a short summary plus a list of facts you shared; see Section 1.5).

If you use the builder without an account (the open /house and /try pages), your house content is stored only in your browser's local storage on your device, not on our servers — except when you invoke an AI feature, in which case the content is sent to our servers and AI providers for processing (Section 3) but is not saved to our database.

1.4 Classroom and roster information

If you use classroom features:

  • Teachers create classes (class name) and receive a short shareable join code. Teachers also create assignments (question text, mode, due date, optional "strawman" generation settings such as an intended grade level and audience age — these describe the assignment, not any specific student) and may record written feedback and a grade on each student's submission.
  • Students join a class by entering or following a join code. We store the class membership (which class, which user, role, join date).
  • Teacher visibility: a teacher of your class can see the class roster (your username, email address, and join date) and has read-only access to the houses you own, including houses created for that teacher's assignments and self-started houses. Teachers cannot edit your house content; they can only add feedback and a grade in a separate record.
  • Students in a class cannot see each other's roster entries or houses.

1.5 AI interview transcripts and AI interactions

The AI co-pilot's context interview asks you questions to understand your situation. The running transcript of that conversation is processed ephemerally: it is sent to our AI providers to generate the next question and, at the end, a distilled context (summary + facts). Only the distilled context is saved to your house; the full transcript is not stored in our database. Other AI features (suggestions, critique, research, strawman generation) similarly send your current house content for processing and return results without storing the request.

1.6 Usage limits and technical data

  • AI usage counters: to enforce daily AI usage limits, we store a daily count keyed to your user ID (if signed in), an anonymous random cookie identifier (if not signed in — see Section 6), or, as a fallback only, a one-way hash of your IP address. We do not store raw IP addresses in this system.
  • Analytics: we use Vercel Analytics to collect aggregated page-view and usage statistics (see Section 4).
  • Server logs and AI service health: our administrative monitoring records the operational status of our AI providers (success/failure/rate-limit events per provider). This monitoring is about provider health, not about the content of your requests.

2. How We Use Information

We use the information above to:

  • Create and operate your account and authenticate you.
  • Store and display your houses and profile.
  • Operate classroom features: rosters, assignments, teacher read-only review, feedback, and grades.
  • Provide AI features (question suggestions, critique, research with web citations, the context interview, and teacher-directed strawman generation).
  • Enforce fair-use daily limits on AI features and prevent abuse.
  • Understand aggregate usage of the Service (analytics) and keep it running.

We do not sell personal information, and we do not use your content or personal information for advertising.

3. AI Processing and Third-Party AI Providers

When you use an AI feature, the relevant content (your house content, your question, and — for the interview — the conversation so far) is sent from our servers to one or more third-party AI providers. Our routing system selects a provider per request based on availability and request size. The providers configured in the Service are:

  • Cerebras (api.cerebras.ai)
  • Mistral AI (api.mistral.ai)
  • Groq (api.groq.com)
  • Google (Gemini API, generativelanguage.googleapis.com)
  • OpenRouter (openrouter.ai) — used only as a fallback of last resort

In addition, the Research Mode feature sends a search query (generated from your house's question and content, not your identity) to Brave Search (api.search.brave.com) to retrieve real web sources for citations.

AI requests are made server-side using our API keys; we do not send your email address, user ID, or other account identifiers to these providers as part of the request content. However, the content you write in a house or say in the interview is, by necessity, shared with the provider that processes that request, and their handling of API data is governed by their own terms and privacy policies.

4. Other Third Parties (Subprocessors)

  • Supabase — authentication, database, and storage of all account, profile, house, and classroom data described above.
  • Vercel — application hosting and Vercel Analytics. Analytics collects page views and related technical signals; Vercel Analytics is designed to work without third-party cookies or persistent client-side identifiers.
  • The AI providers and Brave Search listed in Section 3.

We do not use advertising networks or social media trackers.

5. Data Retention and Deletion

We retain your account data, profile, houses, and classroom records for as long as your account exists. The Service does not currently implement automatic retention periods or scheduled deletion of any stored data.

What the current system does and does not do:

  • AI interview transcripts are not stored (only the distilled summary you approve into your house is saved).
  • Deleting a house (where the interface allows it) removes the house and all of its layers from the database.
  • Deleting an assignment or class does not delete student work: student houses are preserved and merely unlinked.
  • Account deletion is not yet self-serve. The "Delete account" control in the profile page's Danger Zone does not currently delete anything; it informs you that deletion is not yet available. To delete your account, contact us (Section 10) and we will delete it manually. Our database is structured so that deleting an account cascades to delete its profile, houses and their contents, and class memberships.
  • AI usage counters are stored per calendar day; the code does not currently purge old daily rows.

6. Cookies and Local Storage

  • Authentication cookies (Supabase): first-party cookies that keep you signed in. Strictly necessary for the Service to function.
  • hot_aid — an anonymous, random, first-party, httpOnly cookie set when a signed-out visitor uses an AI feature. It exists solely to enforce the anonymous daily AI usage limit. It is not readable by page scripts, is not used for analytics or advertising, and is not linked to your identity; if you later sign in, we stop using it and rate-limit by your account instead. It expires after approximately 13 months.
  • Local storage: if you use the builder without an account, your house content is kept in your browser's local storage on your device under a single key. It stays on your device unless you use an AI feature (Section 1.3) and can be cleared through your browser settings.

We do not use third-party advertising or cross-site tracking cookies.

7. Children's Privacy (COPPA) and Student Records (FERPA)

The Service includes classroom features designed for use by students, and we have written this section from what the product actually does today.

How students get accounts. Students create their own accounts with an email address and password, selecting the "Student" account type themselves. Teachers do not create accounts on students' behalf; they only share a class join code that a student redeems after signing up. The Service does not ask for a date of birth, does not verify age, and has no built-in parental-consent flow.

COPPA. Because there is no age gate or parental/school consent mechanism in the product, the Service is not currently designed to accept self-registered users under 13. Children under 13 may not create accounts. Where a school wishes to use the Service with students under 13, COPPA permits a school to consent on parents' behalf for educational services, but that consent must be established between us and the school outside the product (e.g., in a written agreement), because the product itself has no mechanism to record it.

What student data exists. For a student account, the personal information involved is: email address, chosen username and any optional profile fields the student fills in, class memberships, houses (including assignment submissions and any personal context the student shares with the AI interview, of which only a distilled summary is stored), and teacher feedback and grades on their submissions.

FERPA. Where the Service is used by a school, student submissions, teacher feedback, and grades may constitute part of a student's education records. In that setting we act as a provider to the school: teachers — not we — control class creation, assignments, visibility, and grading, and we use student data only to provide the Service (we do not sell it or use it for advertising, per Section 2). Access controls in the product enforce that only the student and their class's teacher(s) can see the student's work, and students cannot see one another's rosters or houses.

A note on AI processing of student content (Section 3): student house content and interview messages are processed by the third-party AI providers listed above. Schools evaluating the Service should review that list.

8. Your Rights and Choices

  • Access and correction: you can view and edit your profile fields and your houses directly in the Service at any time.
  • Deletion: house-level deletion is available in the product where shown; account deletion currently requires contacting us (see Sections 5 and 10).
  • Leaving a class: a student can remove themselves from a class, which ends the teacher's roster visibility going forward.
  • Anonymous use: you can use the core builder without an account; see Sections 1.3 and 6 for what that does and does not send to us.
  • Depending on where you live, you may have additional legal rights (such as access, portability, deletion, or objection) under laws like the GDPR or state privacy laws. To exercise any right, contact us at the address in Section 10.

9. Security

Data is stored in Supabase with row-level security policies enforcing the access rules described in this policy (owner-only access to houses, read-only teacher visibility, roster privacy between students). AI provider API keys are held server-side only. Passwords are managed by Supabase Auth. No method of transmission or storage is completely secure, and we cannot guarantee absolute security.

10. Contact

Questions, requests (including account deletion), or complaints:

  • Address: a registered address will be added when the operating entity is finalized

11. Changes to This Policy

We may update this policy from time to time. We will post the updated version here with a new effective date. For material changes affecting classroom or student data, we will make reasonable efforts to notify account holders (and, for school deployments, the school).