Questions worth asking before joining an experiment.

Dunbarrios is a framework for small groups that repeatedly share responsibility.

It gives people ways to coordinate recurring work, make decisions, resolve friction, remember what happened, and improve how they function together over time.

DunbarrioSolo begins with private reflection, circles, mutuality, and personal practice. Shared tasks, routines, and sprints become available after a circle graduates.

The larger experiment is whether mutual support can become more dependable without turning into hierarchy, surveillance, or control.

DunbarrioSolo is the current doorway into the wider Dunbarrios framework: a private journaling practice built around relationships.

You write short entries called Embraces, addressed to illustrated mascot characters; organise the people in your life into circles; keep a Ledger of small acts of connection; and generate a Weekly Weave when you want a narrative reflection on recent patterns.

There is no public audience, feed, or follower system for your journal.

The current focus is small groups where people already depend on one another, including:

• households and families
• care circles
• small businesses and small teams
• neighborhood and local groups
• other recurring groups sharing responsibility

There is also a developing stewardship model for financially capable individuals or households interested in making support more dependable without buying control.

No.

Household chores are a useful test case because they expose recurring responsibility, invisible work, memory, fairness, rotation, and coordination very clearly.

The same underlying machinery can apply to care, small-team operations, maintenance, recurring services, onboarding, shared procedures, and other repeated work.

Not in the usual sense.

Project-management software is generally optimized around projects, tasks, deadlines, managers, and throughput.

Dunbarrios is more interested in recurring small-group responsibility: who carries what, how roles rotate, what happened last time, how disagreements are handled, and whether the group gets better at doing something together.

It may overlap with project-management tools in places, but it is not trying to replace every business workflow product.

No.

There is no public feed, followers, likes, audience-building system, or public profile discovery. Your journal entries are not published. Library routine packs can currently be kept private or shared through an unlisted link; a user-facing path to public Library publication is not available.

The app does contain deliberate connections between accounts: invitations, mutual connections, circle governance, and structured repair sessions. Journal entries and private notes about people are not shared through those features.

When someone accepts your invitation, the app may offer them people already in your circles as placement suggestions. A suggestion can include a person’s name, picture, mascot, accent colour, and how long they have been in one of your circles. It does not include birthdays, postal codes, contact details, social accounts, your relationship label, or your private notes.

Direct circles are intentionally small enough that the people inside them can remain recognizable to one another and shared responsibility can stay legible.

The working range is generally 3–9 people.

The number is a design constraint, not a claim that every human group in every situation must have exactly that size.

The larger Dunbarrio concept uses roughly 150 as a practical human-scale ceiling for a wider community of connected circles.

It is not treated as magical scientific precision.

The point is to remain small enough that governance, recognition, responsibility, and recomposition remain plausible rather than building one giant anonymous institution.

DunbarrioSolo is a responsive web app that runs in modern desktop and mobile browsers. Where the browser supports it, the site can also be installed in a standalone, app-like window.

It is not currently presented as an App Store or Play Store app, and browser installation does not make the application work offline.

On-device AI requires WebGPU and a sufficiently capable device. The app checks capability and does not offer that mode where the browser cannot support it.

Use a Google account or an email address and password. You can also enable multi-factor authentication, which is then required for private pages and sensitive account actions.

Your session is stored in first-party browser local storage and persists after you close the browser. Signing out clears that authentication session from the device.

On the free tier, if you opt in on a capable device, a mascot reply can be generated by a model running inside your browser. Your Embrace text is not sent to Google or Anthropic for that reply.

The model must first be downloaded and is then cached on the device.

The AI prompt stays on your device; the app does not become offline.

Your account, circles, Ledger, encrypted entries, artwork, and other application services still use network infrastructure so they can be saved, synced, and displayed. The model itself is also downloaded over the network.

For a mascot reply, the request may include your Embrace, your own recent entries to the same person, the mascot’s voice profile, the person’s display name, your relationship label, and the circle name, type, and tone.

For a Weekly Weave, it may also include counts, moods, energies, and dates of your activity, plus an approaching birthday expressed as a number of days rather than the exact date.

Online requests exclude social handles, postal codes, exact birth dates, favourite colour, private notes about people, Keep note text and tags, and database identifiers. If you name someone in your own writing, that name remains part of the text you asked the feature to process.

Online AI features use Google Gemini or Anthropic Claude depending on the feature and plan. For mascot replies, the free tier tries Gemini first while Companion and Deep End try Claude first.

If the preferred mascot-reply provider is unavailable, the request may be retried at the other provider. Weekly Weaves and Weave follow-up conversations currently remain with their selected provider rather than switching providers. If both mascot-reply providers fail, the app can show a fixed response written without an AI model.

Yes. Turn off “Let the Mascot Answer” on an Embrace. An entry saved with that option off is not sent to an AI provider.

An illustrated character that receives your Embraces and writes back in a consistent voice. Each has a distinct species, temperament, and way of speaking. You choose one to begin with; more become available as you progress, and some are associated with paid tiers.

An Embrace is a journal entry addressed to a person in your circles, a mascot, or yourself. You choose a mood, may add an energy level, write what you want to say, and decide whether the mascot should answer.

Stored Embrace content is encrypted on your device, and the app requires you to protect your vault before it will save sensitive content.

A Weekly Weave is a narrative reflection generated from your Embraces, moods, and Ledger activity. You can ask follow-up questions about it.

You generate it when you want it. The covered week ends on the most recently completed Sunday, but nothing is generated or delivered automatically on Sunday.

Free accounts receive a limited number of full Weaves and then a shorter form. Current limits are shown in the app.

The Ledger is the running record of your practice: Embraces and smaller acts of connection you record. It helps preserve what happened without turning activity into a public performance score.

Sensitive stored content is encrypted on your device with a key generated there. The database stores only wrapped copies of that key in the same database as the ciphertext, locked by a Vault Passphrase you choose and a Recovery Phrase shown once.

The app will not save sensitive content until you choose your own Vault Passphrase. A trusted device can remember an unlock, and trusted devices can be revoked.

If you lose both your Vault Passphrase and Recovery Phrase, your stored content cannot be recovered by you, support, or the operator.

Vault encryption covers stored content, not every processing path. Plaintext passes through the backend when you request online AI, when a Weave or follow-up answer is composed, and when an export is assembled. Circle/contact, account, billing, and support records are protected relational records rather than vault ciphertext.

You can select 30 days, 90 days, 1 year (the default), or Forever for journal text in Profile → Settings. Entries marked Kept and the Forever setting are intended to be exempt from automatic removal. Weekly Weaves are stored separately and are not covered by the journal-text process.

Automatic removal is not switched on yet. Your setting currently records the window you selected; text older than that window may remain stored until removal is enabled.

Yes. Ledger Export produces .txt, .md, and .csv files covering your profile, mascots, circles and people, Embraces and Ledger history, Keep entries, Weekly Weaves, support conversations, growth portraits, and repair sessions.

Your device decrypts vault content and sends the plaintext through the backend for export assembly; the vault key remains on your device. Per-version history behind Keep entries is currently excluded.

Yes, through Profile → Settings → Delete My Account. Deletion is irreversible and removes active-service account data, journal and Ledger entries, Keep history, circles, Weaves, support conversations, repair sessions, and wrapped keys.

If you have a paid plan, cancel first. Deletion attempts Stripe cancellation but does not wait for it to succeed. Cancel through the billing portal before deleting or confirm afterwards through Stripe’s communications that billing stopped.

Non-identifying aggregate records, payment and webhook records, de-identified operational telemetry, backups, and provider logs may remain or age out on provider schedules.

“Preferred” matters most for mascot replies: those requests may fall back to the other provider. Weekly Weaves and Weave follow-up conversations stay with their selected provider. Prices, billing cadence, writing limits, and current Weave allowances are shown in the app rather than repeated here.

No. It is a reflective practice tool, not a therapeutic or medical intervention. Mascot replies are generated by a language model, not a clinician, and are not diagnoses or professional advice.

If you are in crisis or having thoughts of self-harm, contact professional help. In the United States, call or text 988, or text HOME to 741741. Automated distress detection is imperfect and should not be relied on as a safety net.

Use the Support channel in Settings to open a categorised ticket with threaded replies. Support conversations are included in your account export.

You can also email support@dunbarrios.com. Privacy questions can be sent to privacy@dunbarrios.com.

Read the Terms of Service and Privacy Policy for the full current disclosures.

No.

Dunbarrios may surface situations, workload, unfinished work, or other useful signals, but the project explicitly rejects reducing people to a single score.

That is not the design.

Metrics may help people notice something worth discussing. They should not quietly become automatic sanctions.

Not by design.

The stewardship model specifically separates contribution from discretionary control.

Financial support should not buy ideological conformity, permanent authority, ownership of the community, or immunity from disagreement.

The system is designed around rotating responsibilities rather than assuming that one person should permanently coordinate everyone else.

That does not mean every group must distribute every responsibility identically. It means authority should be explicit rather than smuggled in through habit.

Dunbarrios is designed to make disagreement visible and resolvable rather than treating it as system failure.

Governance handles structural decisions after mutual connections are established. Repair is available for relational friction between two accounts with an accepted mutual connection.

The goal is not to automate conflict away.

Repair Mode is a structured way to address relational friction when ordinary interaction is not enough.

It is a mutual, two-account flow that becomes available after the connection between those people has been accepted. It supports repair without turning conflict into public content or an engagement mechanic.

Dunbarrios is designed around private-by-default interaction and bounded sharing.

DunbarrioSolo also distinguishes between on-device and online AI interaction modes.

Because privacy behavior depends on the exact feature and processing mode, the product should explain what is processed where rather than relying on one vague privacy promise.

The beta will keep those explanations explicit as the implementation evolves.

AI can support parts of the private interaction experience, such as mascot responses and reflection.

It is not intended to score people, decide who is trustworthy, impose sanctions, or replace group judgment.

Where processing happens should remain visible to the user.

No.

It is coordination and relational infrastructure for small groups, not a substitute for professional or emergency systems.

DunbarrioSolo is the current working entry point.

Private reflection, circles, invitations, mutual connections, the personal Ledger, and Weekly Weave are part of the current app. Structured repair is available as a mutual, two-account flow.

Governance appears after mutual connections are established. Graduated circles expose shared tasks, routines, sprints, rotating Coordinator and Link surfaces, reusable Library routine packs, and Knowledge Tree, adjacency, and federation machinery for eligible circles.

The Workshop, a complete evidence-to-revision workflow, public Library submission from the app, stewardship, and the broader federated ecosystem remain in development or longer-term direction.

The website should continue to mark those boundaries explicitly rather than presenting every design as shipped.

We are looking primarily for real small groups with repeated shared responsibilities and a genuine reason to use the product more than once.

Households, care circles, small businesses, local teams, and other bounded groups may all fit.

The exact beta terms will be stated on the beta page before applications open.

The general expectation is sustained real-world use, participation by more than one group member where relevant, and candid feedback about what works, fails, confuses, or gets in the way.

No.

The beta is intentionally limited and applications will be reviewed individually.

The final beta access and pricing terms have not yet been published.

Do not assume a future commercial price from current beta access.

No fixed public number has been announced yet.

The limiting factor is the ability to support participants seriously rather than maximizing signup volume.

Yes. Applications are open for the first private-beta cohorts.

Applications are reviewed individually, and invitations are staggered as product-support capacity allows. Applying does not guarantee immediate access.