diff --git a/TODO.md b/TODO.md
index 714dfec0..f39ae009 100644
--- a/TODO.md
+++ b/TODO.md
@@ -6,6 +6,8 @@ If you have stuff that needs human attention, you can put it in [Adam's inbox](/
 
 Commonality also keeps its own product/architecture backlog in [`commonality-ui/TODO.md`](./commonality-ui/TODO.md) (open incompleteness allowed at merge). Prefer filing Commonality-specific follow-ups there when they are package-local; use this root list for cross-cutting work or items that should be visible to any LLM picking up the project inbox.
 
+Civility still has a product gap between the copy and the screens. The details are in [`civility/TODO.md`](./civility/TODO.md). Content-funding follow-ups that are not Civility-specific are in [`content-funding/TODO.md`](./content-funding/TODO.md).
+
 When an item from this page is done and no longer needs an LLM implementor's attention, don't mark it "done", just delete it. I don't want this file to get cluttered with already-completed items.
 
 Fake data / seed content is a **standing plan**, not a pile of one-shots: read [`fake-data-generation/PLAN.md`](fake-data-generation/PLAN.md) and do the next unchecked item there (tiny UI world vs real statements vs stress traffic). Do not invent a parallel seed pipeline.
@@ -14,6 +16,8 @@ Getting **testnet to a two-person shared lab** is also a standing plan, not a pi
 
 ----
 
+- **(Ask)** Generalize content funding into extensible structured work types: domain-defined canonical work-item IDs, explicit links to funding requests with duplicate-claim rules, structured creation forms, and item/sequence alignment attestations feeding fundable-projects boards. Capture and refine the design before implementation; see [Structured fundable work](specs/product/structured-fundable-work.md).
+
 - **(Tell)** Next fake-data/seed-data step lives in [`fake-data-generation/PLAN.md`](fake-data-generation/PLAN.md). Abortion, immigration, crime, and LGBT-schools triples are accepted; next is the demo-seed live UI pass.
 
 - **(Tell)** Next testnet-lab step lives in [`workflow/testnet-working-plan.md`](workflow/testnet-working-plan.md). Items 1–10 are checked. The nightly wrapper exports the mutation flag itself. The job still exits red until the local deep-stack item below passes. Indexer ingest is in [`inbox.md`](inbox.md) only if GraphQL goes red again. Do not mix with mass fake activity or mainnet.
diff --git a/civility/TODO.md b/civility/TODO.md
new file mode 100644
index 00000000..421e5d2e
--- /dev/null
+++ b/civility/TODO.md
@@ -0,0 +1,43 @@
+# Civility backlog
+
+Civility is the noninflammatory vertical on top of content funding. The site
+today is mostly that machinery with Civility copy. These items are the gap
+between that and a board of posts that passed the civility check, plus a way
+to put money on the property itself.
+
+Untagged items are **Ask**. See [task autonomy tiers](../workflow/task-tiers.md).
+When an item is done, delete it.
+
+The general "pick a property, see its board" surface lives in
+[`content-funding/TODO.md`](../content-funding/TODO.md). Civility should be one
+property on that surface, not a one-off list.
+
+----
+
+- **Browse attested items, not only contracts.** Cause boards list assurance
+  contracts. A content item can be *included* in a contract, but nothing shows
+  the items themselves: posts that passed a check, whether or not someone has
+  already wrapped them in a contract. Build a browsable item list. Useful
+  scopes:
+  - Items that are not already in a contract. Cap what is shown (age or
+    another bound) so the list stays finite. Each viewer needs a way to hide
+    items they do not want to see again.
+  - Items from one channel.
+  - Items from a set of channels the viewer follows.
+  Civility's first cut is that list filtered to items a trusted attester has
+  marked noninflammatory. The same surface is what Content Funding needs for
+  any property, and later for work items in general. Content items have not
+  been generalized to work items yet; when they are, this UI is the browser
+  for them too. See
+  [Structured fundable work](../specs/product/structured-fundable-work.md).
+
+- **Finish "Explore fundable content."** The landing-page link currently opens
+  the same X / YouTube / Substack creator directory as Content Funding. It
+  should open the attested-item board above (the civility property), not a
+  channel index.
+
+- **Finish "put money on the property."** Donors can fund a specific channel
+  or piece. There is no Civility screen for pledging toward noninflammatory
+  content as a kind — the cause-pool path the docs describe — including
+  delegating the picks. Build that screen so the landing-page promise is a
+  flow, not a paragraph.
diff --git a/content-funding/TODO.md b/content-funding/TODO.md
new file mode 100644
index 00000000..23dfc75b
--- /dev/null
+++ b/content-funding/TODO.md
@@ -0,0 +1,18 @@
+# Content Funding backlog
+
+Cross-cutting content-funding follow-ups. Civility-specific screens stay in
+[`civility/TODO.md`](../civility/TODO.md).
+
+Untagged items are **Ask**. See [task autonomy tiers](../workflow/task-tiers.md).
+When an item is done, delete it.
+
+----
+
+- **Pick a property, see its board.** An attester profile is a configurable
+  prompt (Civility's noninflammatory check is one profile). The UI has no way
+  to choose a property and see the posts a trusted attester has marked as
+  having it. Add that: a property is a configured content check, and its board
+  is the item list described in the Civility backlog (not-yet-in-a-contract,
+  one channel, or a set of channels). Civility is the first property, not a
+  separate product. "Explore kinds of content" is four static labels today;
+  this replaces that.
diff --git a/docs/end-user/commonality/vision-and-strategy/credible-solution/README.md b/docs/end-user/commonality/vision-and-strategy/credible-solution/README.md
index 31f1c2db..8db38853 100644
--- a/docs/end-user/commonality/vision-and-strategy/credible-solution/README.md
+++ b/docs/end-user/commonality/vision-and-strategy/credible-solution/README.md
@@ -6,6 +6,7 @@ TL;DR:
   - And the rest — [getting stuff done effectively](./effectiveness.md), [matching supply and demand](./matching.md), [discovery and vetting](./discovery.md) — is handled by mechanisms that are less novel but still concretely better than what we have.
   - The same primitive also powers [matching funds](./matching-funds.md) — the friendly, cooperative twin of the [credible threat](../hard-to-stop/credible-threat.md).
   - The reason this adds up to a *government-scale* solution rather than a nicer Kickstarter: [delegation and at-cost reimbursement are what let you fund a whole ecosystem](./beyond-one-project.md), not just one project at a time.
+  - One picture of what that ecosystem can be aimed at: [something like a network society, without the society](./network-societies.md).
 
 ## Why is funding public goods hard?
 
diff --git a/docs/end-user/commonality/vision-and-strategy/credible-solution/network-societies.md b/docs/end-user/commonality/vision-and-strategy/credible-solution/network-societies.md
new file mode 100644
index 00000000..c8017de4
--- /dev/null
+++ b/docs/end-user/commonality/vision-and-strategy/credible-solution/network-societies.md
@@ -0,0 +1,156 @@
+# Something like a network society, without the society
+
+This page does two things.
+
+It describes the vision: a population of aligned people who can find each other, meet, keep a memory, and fund concrete work, without first agreeing to join one group. And it lists jobs that population might want to fund, as a starting menu for anyone trying to build something like this.
+
+Balaji Srinivasan's network state starts from a real observation: a lot of people who want the same concrete thing (a sugar-free food environment, medicine outside the FDA's menu, a school that teaches a particular curriculum) cannot get it from the institutions they already live under. The path I pictured while reading the book was a startup society: one moral innovation, one founder, one commandment, a shared community, a shared treasury, in-person gatherings, and eventually a census large enough to bargain with existing states. That picture may be more specific than what he is actually arguing. The rest of this page is about that picture, and about what to keep from it.
+
+The part worth keeping is the observation, the physical layer, and the point that a movement of aligned people needs a way to fund things together. A big Discord with a treasury gets that last part right. What it gets wrong is the premise that the funding has to sit in one pot. The same effect can come from a more decentralized mechanism: many statements, many boards, many assurance contracts, and [organic coalitions](../why-its-better/organic-coalitions.md) that show up where people agree on a concrete outcome and nowhere else. The mechanism is more complicated than a treasury. The hope is that the extra complexity is worth it, because it sharply reduces how often anyone has to say "now let's all agree to join this one group."
+
+People with a shared preference still need a front door, a way to find each other, gatherings at a few scales, a memory, a way to trust or drop providers, people who keep the local nodes alive, and a small interface to landlords, venues, and insurers. The purpose changes the directory and the events. Those jobs stay the same. [Easier than politics](../so-what/easier-than-politics.md) already says why they should not be stuffed into a million-member group with a constitution and a membership boundary: that group is a new government, with the usual fight over who holds the pot.
+
+Two things follow.
+
+A playbook is possible. The same list of jobs can be written as example projects. Funded together, they produce something that *behaves* like a functioning network society. Nobody ratifies it, and nobody can capture it by taking the treasury, because there isn't one.
+
+And a lot of the central roles in that startup-society picture shrink or split. They do not disappear. Someone still writes a good statement. Someone still draws a map. The difference is that a second statement and a second map are normal, not a schism.
+
+## Front door
+
+A stranger still needs somewhere to start. They do not need *the* door.
+
+**Canon.** The one-commandment picture wants a single text, revised on purpose so the loudest chapter does not silently replace it. Commonality wants many texts. Anyone can write a statement and center a cause board on the projects aligned with it. Leaders who can write a statement other people actually sign are still valuable. Writing it is usually not its own funded project. It is closer to a one-time task — not exactly once, but close — that a cause-founder does alongside the rest of the work of getting a preference started. What drops out is the requirement that everyone else coordinate on that one text. A chapter that rewrites the wording in its own words has not broken the society. There was no single society to break. Where two wordings still point at the same project, the implication graph can put that project on both boards. See [organic coalitions](../why-its-better/organic-coalitions.md).
+
+**The site.** The one-front-door picture wants the map of chapters, the next events, and the canon, in one place. Here a chapter publishes its own existence if it wants to be found: a place, a time, a statement it cares about. Anyone can publish a map over the chapters that opted in. Tools can build directories off those publications. The job remains — a map nobody maintains is not a map — but "the" map is a product someone can fund, not an office someone has to hold.
+
+**Onboarding.** The path is from "I read a post" to "I know which statement I am willing to sign, which projects that puts in front of me, and who meets near me." There is no society-wide committee that rejects people for wanting the aesthetic. Filtering happens in two smaller places. A cause board only shows projects aligned with what you signed. A local chapter, which is just a small group with a room, can decide who it invites back. That second filter is ordinary hosting, not a membership roll.
+
+## Finding each other
+
+Introductions do not scale by adding more well-connected Alices unless the introductions leave a trail.
+
+**Local matching.** Point a new person at a few people nearby who will actually reply. The fundable version is a month of that work in one city, with a public count of introductions made and the private details left off-chain.
+
+**A directory of providers.** Doctors, restaurants, tutors, employers, hosts, landlords — whoever a given statement's signers might want. The list rots in months. Paying someone to rebuild it is one project. Paying someone else to check it is a second project, and the checker should not be the person who listed the providers. Competing directories are fine. An audit that says which listings are real is what keeps them honest.
+
+**Routing.** "Who do I ask about X" is private judgment. The artifact worth funding is that judgment remaining available after Alice is busy: a maintained page, a standing offer to answer for a month, not a vibe.
+
+## Gatherings
+
+Events are how a pile of signers becomes a population. These jobs look like the in-person layer of a network state, because a room is a room.
+
+**Local recurring meetings.** A host, a room, a rhythm, and enough facilitation that a newcomer is not stranded. The failure mode is month six, when the first host is tired. Fund the quarter, not "the community."
+
+**Regional gatherings.** Worth traveling for. Programming, ticketing, housing, and a reason to come besides "meet strangers who agree with us."
+
+**The big event.** Once a year, at whatever scale the signers actually support. Norms get demonstrated in public there, so the program and the code of conduct are the product. One event can have one voice for a weekend. That does not make that voice the society's voice.
+
+**Pop-up presence.** A street, a building, or a week in one city where people actually live together. An archipelago is this job repeated until it has an address. Each repetition can be its own contract and its own lessee.
+
+## Memory and voice
+
+**Internal record.** What a chapter tried, which gatherings happened, which experiments failed. A page with an editor, not a chat scrollback. Each chapter can keep its own. A cross-chapter record is just another publication someone chose to maintain.
+
+**Outbound explanation.** Newsletters, talks, a standing answer to a hostile question. Many people will do this, in different words, and that is the ordinary case. Humans do not agree on everything. The point of the organic-coalition machinery is to let them fund a shared outcome where they agree on *something*, which is the situation the usual institutions handle badly: those institutions often demand agreement on *everything* before anyone can move. A contradictory explainer is not evidence that the arrangement has failed. It is evidence that nobody was forced onto one platform.
+
+What still wants a single voice is a specific artifact: one event, one chapter's house rules, one map's methodology. Fund those. Do not fund a minister of information.
+
+**Translation and local adaptation.** Translate a statement if it helps. Rewrite it if it helps. A city that words things differently has not failed to be a chapter of the original. If the rewrite still implies the same concrete project, both wordings can fund it. If it does not, they are two preferences, and both can exist.
+
+## Trust
+
+Any cluster organized around a preference attracts people extracting status, money, or cover.
+
+**Vouching.** Who will introduce someone, and what they are willing to be on the hook for. This is a local reputation fact. Publishing "I vouch for this host" is enough structure. A global vouching office is not required.
+
+**Dispute handling.** Two people who share a kitchen or a chapter need a person and a procedure, or the fight becomes the chapter's main activity. That person is a local role. It does not generalize to "the society's court."
+
+**Removal.** A small group that meets in person needs the ability to stop inviting someone. Above that scale, the equivalent of removal is unfollowing them, not co-signing their projects, and not listing them. There is no central discussion group to moderate, and no central treasury to take over. That absence is the point of doing this with assurance contracts instead of a pot.
+
+**Directory quality.** A listed clinic, school, or host that is fraudulent is a failure of that listing. It is not a stain on a society, because the listing was never an official endorsement unless someone attested it. The fundable job is the independent check, published where the directory's readers will see it.
+
+## Keeping the doers
+
+A local habit dies when the host burns out and the next person has to invent the first month from scratch.
+
+**Host kits.** How to start in a new city: first ten people, first venue, first three months, and when to stop. One public artifact. Every later city is a use of it, not a new ideology.
+
+**The next host.** Recruit them before the current one quits. This is a role inside one city for one season.
+
+**Succession.** Founder-shaped roles still exist, but they attach to a product, not to a people. The person who edits a particular map, leads a particular annual event, or is the phone number a venue already has. When they leave, fund a handoff of that product. There is no office called "the person outsiders call about the society," and inventing one recreates the single founder the startup-society picture needs.
+
+## Census
+
+An auditable count of people, income, and real estate matters in the network-state picture because a society that cannot count itself cannot bargain with a state. Diplomatic recognition is not the goal here. Knowing which efforts are real still is. So is being able to say, roughly, how many people are in this overlapping cluster.
+
+**Headcount that is hard to fake.** Tally counts people who signed a statement. LazyGiving counts money pledged and work that actually cleared a threshold. Neither counts a chat membership. Both are harder to inflate than a follower number, and neither pretends to be a citizenship roll.
+
+Once signers can be deduped against a linked unique-human id, the same records support a set-union across many more-or-less-aligned statements: each person counts once, even if they signed several of them. That is closer to "here's how many of us there are" than any single statement's signature count, and it still is not a citizenship roll. Choosing which statements belong in the set is itself a job — not a huge one, but real work — that this population might want to fund. The counting rules are sketched in [durable participation](../../../../../specs/product/durable-participation.md). A union is only as meaningful as the set someone published; the union of "build this road" and "do not build this road" is not a constituency for either.
+
+**Liveness.** Which cities met this month, and which providers are still practicing. Signatures do not show this. A host who publishes "we met" or an auditor who checks a listing does.
+
+**Outcomes of the purpose.** Sugar metrics, case series, school results. Designing the measurement is meta-work. The activity being measured is the meat of the preference, and it gets its own projects.
+
+What this census will not do is produce "the population of the society" as a single official number. It produces overlap: these people signed this, these contracts paid out, these gatherings were published, and — where someone has done the work — this published set of statements has this many distinct people. That is the number a funder can use.
+
+## Interface with the outside
+
+Gatherings, houses, and clinics still sit inside a country.
+
+**Legal wrappers.** Someone has to be able to sign a venue contract, buy insurance, and hold a lease. That someone should be scoped to the event or the house, not to the whole preference. A reusable kit — "here is how a weekend incorporates and gets insured" — is a public good. A single foundation that all chapters must use is the treasury problem in legal costume.
+
+**Permits, insurance, safety** for anything bigger than a dinner. Part of the event's budget, not a standing ministry.
+
+**Deals.** Employers, schools, clinics, landlords. Someone negotiates, then publishes the result into a directory other people can audit. Adjacent preferences can share a provider without merging. A sugar-free dinner and an unwoke school do not need a common commandment to use the same insurance kit.
+
+**Crisis logistics, and helping one person.** A person is stranded, doxxed, or sick in a city where a parallel service exists. The public good is the response path: who to call, what is already arranged. The transfer of money to the person in trouble is not a public good. It can still be something aligned people want to crowdfund together — the "network union" half of the picture, people sticking with each other. That transfer fits an ordinary unconditional gift better than an assurance contract, because a partial amount already helps.
+
+## What to fund
+
+| Job | What the project actually is | How to fund it |
+| --- | --- | --- |
+| A host kit, a one-time map, a legal-wrapper kit | A public artifact. Partial funding publishes nothing useful | Assurance contract |
+| A city's dinners for one quarter, a regional weekend, one annual event, one pop-up week | Hosting an encounter, budgeted with its own insurance and its own lessee | Assurance contract. The gathering happens only if the threshold is met |
+| An audit of a directory, a count of which chapters met, a check on a provider's claims | Evidence, published by someone other than the person being checked | Assurance contract. An unchecked claim is not worth buying |
+| A published set of statements, and the deduped headcount over that set | Choosing which statements count as "us," then publishing the union once identities can be deduped | Assurance contract. The artifact is the set and the count |
+| A month of introductions in one city, a month on call as the local host, mediation inside one chapter | A named person's availability for a bounded period, with a public log of what was done | Assurance contract. Same shape as a month of software maintenance |
+| Help for one stranded, doxxed, or sick person | A private transfer. Aligned people sticking together, not a good everyone can use | Ordinary unconditional gifts. Partial funding already helps |
+| A statement people can sign | Wording a cause-founder produces while doing the rest of the startup work | Usually not its own project |
+| "People should feel welcomed," a membership committee, a society spokesperson, a shared treasury | Not a project | There is nothing to deliver, or delivering it rebuilds the organization this setup exists to avoid |
+
+## A starter playbook
+
+These are examples for one preference in one city. Swap the purpose and the city. None of the funded items requires the others to exist first, which is the property that makes this a protocol rather than an organization. The lumpy public goods are assurance contracts. The standing local roles are repeating monthly or quarterly contracts. The mutual-aid item is an ordinary gift.
+
+1. **Write the statement,** as part of founding the rest of this, not as a project of its own. Specific enough that a project can align with it. Not a manifesto that settles every neighboring question.
+2. **Publish the host kit.** First ten people, first venue, first three months, when to stop calling it a chapter.
+3. **Host this city for this quarter.** Room, rhythm, facilitation, a published note that the meetings happened.
+4. **Be the introducer for this city for this month.** A public count of introductions. No dossier on the people introduced.
+5. **Publish a map** of gatherings and providers whose organizers opted in. Say what was left out on purpose.
+6. **Audit that map.** An independent pass over a stated sample. Publish which listings answered and which did not.
+7. **Run one regional weekend,** with ticket threshold, housing, program, and insurance inside the same contract.
+8. **Publish the one-event legal kit** used for that weekend, so the next city does not hire the same lawyer from scratch.
+9. **Hand off a role.** One contract whose deliverable is that a named next host, editor, or event lead has everything the last one had.
+10. **Publish a headcount set.** Name the statements that count as this cluster, and publish the deduped union once unique-human ids exist.
+11. **Measure the purpose,** separately. Whatever the preference claims to change in the world. That project cites the gatherings and the directory. It does not define who is a member.
+12. **Help one person in trouble,** when the cluster wants to, as ordinary gifts toward a named need. Not an assurance contract.
+
+Items 3 through 6 are the ones a treasury-Discord usually extracts from unpaid status. They are also the ones that decide whether two people in the same city ever meet. The purpose-specific meat — the clinic, the school, the sugar program — hangs off that, and can be funded by anyone whose statements reach it, including people who would never join a society together.
+
+## Jobs worth funding
+
+Concrete projects one city can start from. Each line is something a person could pledge to. Statement-writing is omitted on purpose: a founder usually does that alongside the rest of the work, rather than as its own project.
+
+- Host kit for starting a chapter in a new city. A duplicated doc is enough.
+- This city's dinners for one quarter, including the room and a published note that they happened. Luma or Meetup for the room; the note is the record.
+- Introducer on call for this city for one month, with a public count of introductions. A calendar link and a count sheet. No dossier of who met whom.
+- Map of opted-in gatherings and providers, with an explicit "left out on purpose" section. A published spreadsheet, or uMap if it needs pins.
+- Independent audit of that map over a stated sample. Resample the same spreadsheet.
+- One regional weekend: program, tickets, housing, and insurance in one contract. Luma or Meetup for tickets and housing.
+- Legal kit for running a weekend like that one: incorporation and insurance, reusable by the next city. A duplicated doc.
+- Handoff of one named role (host, map editor, or event lead) to a named successor. The doc, plus a shared password vault.
+- A published set of statements for "how many of us," plus the deduped union count. A cause board already computes that union over its planks.
+- A liveness note: which of these cities actually met this month. Collect the "we met" notes.
+- An outcome measurement for the preference itself (the sugar metric, the case series, the school result), citing the gatherings and the directory. Use whatever instrument fits — a survey, a lab series — and cite it.
+- One person's emergency costs, funded as ordinary gifts, with the response path (who to call, what is already arranged) funded separately as a public artifact. An ordinary gift page, not an assurance contract.
diff --git a/docs/end-user/commonality/vision-and-strategy/so-what/easier-than-politics.md b/docs/end-user/commonality/vision-and-strategy/so-what/easier-than-politics.md
index ef2242bc..a0db8812 100644
--- a/docs/end-user/commonality/vision-and-strategy/so-what/easier-than-politics.md
+++ b/docs/end-user/commonality/vision-and-strategy/so-what/easier-than-politics.md
@@ -38,3 +38,5 @@ Commonality achieves the same goals without requiring a monolithic organization:
   - **No membership boundary.** Anyone can sign a statement, pledge to a project, or become a delegate. The system discovers commonality among people who happen to agree, without requiring them to formally join anything.
 
 The result is a *protocol* rather than an *organization*. No one needs to run it, no one can capture it, and it scales without governance headaches. It grows organically because each individual action is useful on its own.
+
+What that looks like as a set of fundable jobs — gatherings, directories, local hosts, and the rest of the network-state stack, minus the shared treasury — is in [something like a network society, without the society](../credible-solution/network-societies.md).
diff --git a/docs/end-user/commonality/vision-and-strategy/why-its-better/README.md b/docs/end-user/commonality/vision-and-strategy/why-its-better/README.md
index fa734d9f..cebf78ad 100644
--- a/docs/end-user/commonality/vision-and-strategy/why-its-better/README.md
+++ b/docs/end-user/commonality/vision-and-strategy/why-its-better/README.md
@@ -20,4 +20,6 @@ First, a scoping question worth answering head-on: we already *have* governments
 
   - **Infrastructure as a public good:** The [rails](../ease-of-adoption/rails.md) are built once, available to everyone, auditable by anyone. Established orgs stop spending energy on infrastructure; new orgs get world-class infrastructure instantly without needing to establish their brand.
 
+For the underlying coordination problem, see [Why benefits fail to become incentives](./why-benefits-fail-to-become-incentives.md): sometimes a desired kind of value has never had a concrete funding destination, and sometimes it needs many contributors paying many independent producers through a shared funding purpose.
+
 Also better than private charity — see [charity.md](../legacy-frustrations/charity.md) for a detailed comparison.
diff --git a/docs/end-user/commonality/vision-and-strategy/why-its-better/what-its-better-for.md b/docs/end-user/commonality/vision-and-strategy/why-its-better/what-its-better-for.md
index 6a1caacd..2a9a68ab 100644
--- a/docs/end-user/commonality/vision-and-strategy/why-its-better/what-its-better-for.md
+++ b/docs/end-user/commonality/vision-and-strategy/why-its-better/what-its-better-for.md
@@ -19,6 +19,11 @@ advantage for Commonality.
 
 ## The kinds it's better for
 
+For a complementary catalogue of where the connection between beneficiaries and
+decision-makers breaks, see [Why benefits fail to become incentives](./why-benefits-fail-to-become-incentives.md).
+It maps those failures to Commonality mechanisms, existing fake-data examples,
+and proposed demonstrations of how funding could change someone's choice.
+
 Organized not by topic but by *what's blocking them today* — each a specific failure mode of legacy institutions, paired with the specific Commonality mechanism that addresses it. These categories overlap rather than partition: a given good can sit in several at once.
 
 **1. Self-financed minority goods.** A dedicated minority cares intensely and is willing to pay for the thing itself — niche cultural or scholarly preservation, enthusiast infrastructure, specific community or religious goods, unconventional research.
diff --git a/docs/end-user/commonality/vision-and-strategy/why-its-better/why-benefits-fail-to-become-incentives.md b/docs/end-user/commonality/vision-and-strategy/why-its-better/why-benefits-fail-to-become-incentives.md
new file mode 100644
index 00000000..0d5990ed
--- /dev/null
+++ b/docs/end-user/commonality/vision-and-strategy/why-its-better/why-benefits-fail-to-become-incentives.md
@@ -0,0 +1,291 @@
+# Why benefits fail to become incentives
+
+A writer could make a careful argument that people on the other side would
+actually consider, or publish another dunk. Many people would benefit from the
+careful argument: readers who want to understand, people who want their views
+represented fairly, and everyone who would prefer a less hostile public
+conversation. But those benefits do not necessarily pay the writer. If engagement
+rewards the dunk, producing the dunk may be the individually sensible choice.
+
+[Civility](../../../civility/index.md) tries to supply the missing reward. People
+who want better discourse can fund the kind of writing they want to exist,
+including writing addressed to someone else. The writer can respond to that
+demand without first persuading a social-media platform to change its business
+model.
+
+This is an instance of a much broader problem:
+
+> People would benefit from better decisions or useful work, but their willingness
+> to pay never becomes a credible reward that reaches the people who could deliver it.
+
+That describes both missing work and the wrong mix of work: maintenance versus
+another feature, replication versus another novel finding, or a public reference
+versus knowledge kept inside one business. The person doing the work need not be
+uncaring or irrational. The benefit to others simply fails to reach their choice.
+
+Commonality's opportunity is to let dispersed beneficiaries turn some of that
+benefit into funding. The practical test is: **Whose choice changes once this
+money can reach them—and why?**
+
+## Sometimes the funding opportunity has not been imagined
+
+“I wish social media were less inflammatory” is a familiar desire. “There should
+be money available for anyone who produces a good, noninflammatory explanation”
+is a further step. Someone can have the first thought without ever having the
+second. They may think of better discourse as something to wish for, ask platforms
+to encourage, or practice themselves, without thinking of it as something they
+could help pay others to produce.
+
+In that case, the obstacle comes before finding contributors or creators. There
+is no concrete funding opportunity in people's heads, and no shared place to
+point to. Naming the desired quality, spelling out what would count, and attaching
+actual funding makes an otherwise vague wish actionable:
+
+- “Here is where you put money to make more of this happen.”
+- “Here is where you find out how your work could get paid for making it happen.”
+
+A page alone cannot do that. It needs a connection between the stated purpose,
+available money, judgments about qualifying work, and a route to payment. But
+giving that connection a visible home matters: people can share it, add to it,
+and plan around it. Creating the funding destination is itself organizing work.
+
+This differs from **missing discovery**. Sometimes an opportunity already exists
+and people cannot find it. Here, someone must first formulate the opportunity:
+turn a quality people value into something they recognize as fundable. A useful
+founder question is: **What do people wish existed that they have never thought
+of attaching a funding pool to?**
+
+## A crowd paying a crowd for a kind of value
+
+Ordinary crowdfunding makes it easy to imagine many people paying for one
+specified project. Supporting a creator makes it easy to imagine many people
+paying one person for an ongoing body of work. There is another useful pattern:
+many people funding a kind of value, with many independent people able to produce
+qualifying work. The contributors need not select every producer in advance,
+and each producer need not assemble a separate paying audience.
+
+[Content Funding](../../../content-funding/index.md) makes this concrete. People
+can fund content with explicitly stated qualities, and different creators can
+produce work that serves that demand. A contributor might care about careful
+explanations addressed to a particular audience without already knowing who
+should write them. A creator might be able to write one without already having
+fans willing to support their career.
+
+The organizing commitment is “more work of this kind,” which can outlast any
+particular piece or producer. The pool gives both sides a common reference point:
+contributors can see what they are helping make possible, and producers can see
+what people are willing to fund. Individual payments still go to particular
+recipients; the crowd need not act as a single contractor.
+
+Here, “pool” describes a shared funding purpose, not a requirement that everyone
+surrender their money to one administrator or accept one judge. In Commonality,
+people can choose whose evaluations and spending decisions to trust. Nor does
+qualifying automatically entitle every submission to payment: a credible
+opportunity must explain how limited funds are allocated, when payment happens,
+and which decisions remain discretionary. A visible budget is a signal of demand;
+a commitment to pay for specified work is a stronger promise.
+
+The general pattern also appears in prizes, bounties, and grant programs. The
+opportunity is to make it easier to establish and sustain for neglected kinds of
+value, including repeated contributions by many producers.
+
+### Beyond social-media content
+
+Possible applications of the same coordination pattern include:
+
+| Shared funding purpose | Work different people could contribute | What evaluation would need to establish |
+|---|---|---|
+| Make public information accessible in a neglected language | Translations of useful guides, supplied by different translators | Accuracy, usefulness to the intended readers, and permission to publish |
+| Keep a shared collection of open-source tools usable | Reproduced bugs, accepted fixes, or maintained documentation across many tools | That the work solves a real problem and is accepted or useful, rather than manufactured activity |
+| Improve an open map's accessibility information | Verified updates about entrances, lifts, and step-free routes | Accuracy, freshness, and whether an update adds information rather than duplicating a paid claim |
+| Make published research easier to check | Independent reproductions of analyses, with data and methods available | Methodological quality and independence, including when results fail to replicate |
+
+These are possible extensions of the funding pattern, not claims that Content
+Funding already verifies or pays for these activities. Some work needs funding
+before it begins; some can be rewarded after delivery. Each domain needs its own
+evidence and allocation rules.
+
+“Post proof that you planted a tree” exposes the same distinction. A post is
+evidence to assess, not the value itself: someone must establish that the tree
+was appropriately planted, survives, and has not been claimed repeatedly. Making
+a pool visible addresses a missing coordination point; it does not make real-world
+outcomes easy to verify.
+
+## How this fits the existing strategy
+
+[What kinds of public goods is Commonality better for?](./what-its-better-for.md)
+catalogues many types of institutional blockers. This page supplies a complementary
+diagnosis: where does the connection between benefit and incentive break?
+[Unbundle the organization](./unbundled-organization.md) explains how different
+people can supply money, judgment, trust, distribution, and execution. The
+[cause taxonomy](/specs/product/cause-taxonomy.md) turns these considerations into
+varied, concrete projects.
+
+The distinction matters because a good cause title is not enough. “Science” does
+not tell us which choice needs to change. “Pay a research team to perform and
+publish a replication even when the result is negative” does.
+
+One source of broader inspiration is [Moloch's Toolbox](https://equilibriabook.com/molochs-toolbox/) by Eliezer Yudkowsky,
+which distinguishes decision-makers who are not beneficiaries, asymmetric
+information, and coordination traps. Those distinctions also set limits on the
+funding thesis. Sometimes the missing ingredient is trustworthy knowledge;
+sometimes several actors must change together. More money reaching one person
+does not automatically resolve either problem.
+
+## Where the connection breaks
+
+These categories overlap. A project can encounter several at once.
+
+| Why the incentive fails to arrive | How Commonality might help | What remains unresolved |
+|---|---|---|
+| **The value has no funding destination.** People wish for more of something without imagining it as a kind of work they could pay for. | Give the desired value a name, explicit criteria, and a visible connection to actual funding and a route to payment. | A page and a good label do not supply contributors, workable criteria, or credible payment terms. |
+| **No single producer to commission.** The desired value consists of many contributions from people who may not yet be known. | Let a shared funding purpose direct money to qualifying work from many independent producers. | Limited funds need allocation rules; eligibility alone cannot promise payment to everyone. |
+| **Dispersed benefits, concentrated costs.** Many people benefit, but none can justify commissioning the whole thing. | Aggregate contributions toward a concrete deliverable, such as a maintained shared dependency. | Beneficiaries can still choose to free-ride. Total social benefit is not the same as money people will contribute. |
+| **Uncertainty about others contributing.** People would help if enough others did; a partially funded attempt may buy nothing useful. | Assurance contracts make contributions conditional on reaching a funding threshold. | Reaching the threshold does not itself guarantee delivery or quality. |
+| **Missing discovery.** Beneficiaries cannot find each other or someone willing to do the work; producers cannot see demand. | Visible funding intentions and fundable-projects boards help supply and demand find each other. | A founder still needs to reach people. Publishing a board does not create an audience. |
+| **Costly evaluation.** Each person's benefit is too small to justify independently researching every proposal or result. | Chosen delegates and attesters let people rely on others' judgment; evaluation work can itself be funded. | Evaluators can be wrong or captured. A public record is not proof of truth. |
+| **Competing rewards favor the worse choice.** Attention, prestige, or sales reward behavior that benefits the wider constituency less. | Add a revenue source for the neglected behavior: civil argument, replication, maintenance, or open documentation. | The additional reward must be credible and large enough to change the relevant choice. Money may not compensate for lost status or institutional access. |
+| **Value is recognizable only afterward.** Early contributors bear uncertainty while later beneficiaries enjoy the result. | Retroactive donations reimburse early contributors at cost, allowing the same giving budgets to fund another attempt. | Later donations are uncertain. Reimbursement offers no profit and cannot by itself attract people seeking an investment return. |
+| **No shared institution.** Beneficiaries cross jurisdictions or distrust one another's organizations. | People can co-fund a specific deliverable while choosing their own delegates and sources of judgment. | The shared deliverable must actually be acceptable to them; discovering overlap cannot manufacture agreement. |
+| **Several actors must move together.** Going first is costly or useless without complementary changes. | Finance shared transition costs, enabling infrastructure, or a viable alternative. | A money threshold is not a binding agreement that several institutions will adopt a new practice together. |
+| **Beneficiaries cannot pay or participate.** The benefit accrues to future people, people without resources, or other absent stakeholders. | Motivated sponsors can contribute on their behalf. | The system still depends on someone present being willing and able to pay. |
+
+Assurance contracts address a particular obstacle to cooperation: “I would
+contribute if enough others did.” They do not eliminate every reason to free-ride.
+Similarly, delegation lowers the cost of participating but does not make judgment
+infallible, and retroactive reimbursement recycles giving capacity without
+guaranteeing it will be replenished.
+
+There are also cases where the decision-maker cannot legitimately accept the
+proposed payment, lacks authority to change the practice, or faces a legal
+barrier. A fundable implementation project or independent alternative may help;
+a generic offer to pay for a better decision is not sufficient.
+
+## Why Civility is a particularly useful example
+
+Civility illustrates **a missing funding destination**, **many producers serving
+a shared funding purpose**, and **competing rewards**, beyond merely collecting enough money
+for an otherwise uncontroversial project. Its hypothesis is that a constituency
+for better discourse can finance a different mix of content from the mix that
+engagement rewards alone would produce.
+
+There are at least two interested groups: people who want to read the other
+side's views without being insulted, and people who want their own side's views
+to receive a fair hearing. Neither group needs to adopt the other's politics to
+fund the same piece. Audience-specific evaluators help assess whether the piece
+can be heard by its intended readers; people still choose whose judgments to
+trust and what to fund.
+
+The proposed causal chain is:
+
+1. People express demand and commit money toward a kind of content.
+2. A creator can see a credible funding opportunity.
+3. The creator chooses to produce something that serves that demand.
+4. Evaluation helps contributors identify qualifying work and direct funding.
+5. Repeated funding makes producing this kind of work a more viable practice.
+
+Each link matters. Paying for existing content demonstrates recognition, but not
+necessarily additional production. A respectful tone does not establish factual
+accuracy. Funding publication does not establish that the intended audience saw
+it, understood it, or changed its mind. A useful account of a project should
+say which links it exercises and which remain hypotheses.
+
+## Questions for a proposed project
+
+For a concrete project, a reader should be able to answer:
+
+- **Beneficiaries:** who gains, including people who will not pay?
+- **Funding destination:** has anyone made this kind of value explicit and given people a place to fund it and find paid opportunities?
+- **Decision-maker:** whose choice determines whether the benefit happens?
+- **Producers:** is this one commissioned project, or an ongoing invitation for many people to contribute qualifying work?
+- **Current choice:** what do they do now, and what rewards or constraints explain it?
+- **Missing connection:** why can't beneficiaries already reward the better choice?
+- **Intervention:** which mechanism — pooled funding, an assurance threshold, delegation, evaluation, or reimbursement — changes that situation?
+- **Credibility:** what makes the funding opportunity believable before the work begins?
+- **Allocation:** if more work qualifies than the budget can cover, who gets paid, how much, and who decides?
+- **Evaluation:** who judges delivery, by what evidence, and what could fool them?
+- **Boundary:** what remains unsolved even if the money arrives?
+
+The test is the same question from the start of this page: **Whose choice
+changes once this money can reach them—and why?**
+
+
+
+
+
+
+# Chat with Astra
+
+
+> Any suggestions for additions we could make to the product or the fake-data examples or whatever? Like, does this give you any ideas for cool new coordination mechanisms or ways we could use the existing ones or whatever?
+
+
+Yes. The most interesting shift is to make **“there is money available for more of this” a first-class product experience**. Right now, the docs describe the ingredients, but we could make their combination much more tangible.
+
+I’d explore these six ideas. The first two mostly assemble existing mechanisms; the others introduce new behavior.
+
+1. **Give every funding opportunity a producer-facing view.**
+
+   A cause board could answer: “What could I do here to get paid?” Show the desired qualities, examples of qualifying work, available funding, who makes funding decisions, and how to propose work.
+
+   Put **“Fund more of this”** and **“Produce something like this”** beside each other. That makes your two-sided coordination point visible. It should distinguish money people can currently direct from money actually committed to a particular piece of work.
+
+2. **Let people turn an example into a standing invitation.**
+
+   After funding a useful piece, offer: **“What about this would you like more of?”**
+
+   Someone funds an explanation of a contentious issue, then specifies: “Explains the strongest opposing argument accurately, for readers unfamiliar with it.” They can find an existing matching cause or publish that funding purpose and attach money.
+
+   This provides a path from *recognizing value afterward* to *requesting more beforehand*. Examples and counterexamples could also help evaluators interpret the criteria.
+
+3. **Let producers ask for a commitment before doing the work.**
+
+   A visible pool may attract twenty people to write essentially the same explainer. Only three get funded; the rest learn that responding to the incentive was a gamble.
+
+   Allow a lightweight proposal: “I could cover this topic for $200.” A contributor or delegate can then fund a prospective round using the existing assurance mechanism. A more ambitious extension would reserve part of a budget until a delivery deadline.
+
+   The new coordination benefit is **allocating effort before people spend it**, while preserving a route for unsolicited completed work.
+
+4. **Fund the missing pieces of a collection.**
+
+   Sometimes “more qualifying work” stops being useful because everyone tackles the easiest example. A pool could publish gaps: explanations for these five audiences, translations into these three languages, accessibility information for these uncovered locations.
+
+   Contributors still fund a shared purpose; producers choose gaps they can fill. Completing a gap changes what the pool seeks next. This would require additional tracking and allocation design, but gives the money a way to express *what is still missing*.
+
+5. **Make “this would be fundable if someone could evaluate it” actionable.**
+
+   Suppose people want to fund accessible translations but nobody trusts the available quality assessments. The board could expose that bottleneck and invite a separate funded proposal for evaluation work.
+
+   That creates a useful sequence: contributors express demand → a reviewer establishes usable criteria and assesses submissions → producers can respond to a credible opportunity. Existing projects, delegation, and attestations provide much of the machinery; the product addition is making the dependency visible.
+
+6. **Try a pool for maintaining a condition over time.**
+
+   For example: “Keep these public guides accurate as circumstances change.” Different people identify stale sections, make corrections, and review updates. Recurring funding supports continuing usefulness.
+
+   This is an interesting extension beyond rewarding discrete output: sometimes the value people want is **that something stays good**. It needs explicit review intervals and responsibility boundaries.
+
+For fake data, I’d start with **one pool, several strangers, two rounds**:
+
+- Several contributors fund careful explanations for a specified audience.
+- Three previously unrelated creators discover the opportunity.
+- One gets a prospective project funded; another submits existing qualifying work.
+- A third produces something that qualifies but receives no funding because the budget is exhausted.
+- In the next round, visible gaps and funding decisions guide what creators propose.
+
+That would extend the existing incentive-demonstrations plan beyond one writer changing their choice. It would demonstrate **a shared funding purpose coordinating several independent decisions**, including where the coordination fails.
+
+My first picks would be the producer-facing view and that synthetic story. They would make the core idea concrete and reveal which additional mechanism is actually needed.
+
+
+
+
+# My response
+
+Right, I like the idea of focusing on “there is money available for this” and making sure that it's a tangible thing, from both directions (donors and doers).
+
+The pattern in these suggestions seems to be something like: meta-data about the projects, or desired-projects, or something?
+
+What needs to be onchain and what doesn't? What's core to the protocol, versus just being offchain/out-of-protocol stuff that anyone can do separately in order to organize the onchain stuff? I'm more interested in figuring out what primitives would help; the offchain stuff is easier to iterate on later.
+
+The thought in my head is something like:
+  - Maybe there's a range of options on the spectrum between "this money is hard-committed to this particular assurance contract" and "this money is soft-committed, meant to be used for something sorta vaguely along these lines." That is, maybe there's a place for "this money is medium-committed, meant for a project that'll match this not-quite-fully-concrete description." i.e. Maybe a new kind of smart contract? It sorta feels like it needs to be a new kind of smart contract - e.g. it escrows the money for some period, and it has a clearer description - but is the purpose clear enough that it makes sense to make it a smart contract? Maybe there's an escrow period and a description of what's wanted; the escrow period helps the potential-donors demonstrate their commitment to donating to *something* along these lines, and the description is a fairly-clear semi-commitment. The donors aren't *required* to make the final commitment to donate to one of the concrete projects (this is important, since it's possible that none of the proposals will be good enough in the donor's estimation), but maybe the donors are at least required to provide an explanation for *why* they're passing? The point is that this makes bad-faith a bit easier to spot: it requires actually locking up the money for a while, and the project proposals can be compared to the request-for-proposals to see whether they actually look plausibly like a good fit, and there'll be an onchain record of the donor backing out (even if he does give a reason). I dunno, this doesn't feel super-compelling to me; maybe you have thoughts?
diff --git a/docs/end-user/content-funding/index.md b/docs/end-user/content-funding/index.md
index 082e8d42..1deeaf1f 100644
--- a/docs/end-user/content-funding/index.md
+++ b/docs/end-user/content-funding/index.md
@@ -12,6 +12,15 @@ Under the hood, Content Funding is [LazyGiving](../lazyGiving/index.md) pointed
 
 The first two are about *specific* work — a piece that exists, or a creator's next output. The third is about a *category*, and it feeds the first two: a cause pool can reward existing pieces that qualify or commission new ones.
 
+That third option gives a wish like “I want more careful explanations” a concrete
+funding destination. Many people can contribute toward the same stated qualities,
+and many independent creators can produce work that meets them, without each
+creator first building a separate paying audience. The shared pool is a place to
+both help pay for more of that work and discover opportunities to get paid for
+producing it. Qualifying work still needs a funding decision; a finite pool does
+not guarantee payment to every qualifying piece. The broader strategy is
+[a crowd paying a crowd for a kind of value](../commonality/vision-and-strategy/why-its-better/why-benefits-fail-to-become-incentives.md#a-crowd-paying-a-crowd-for-a-kind-of-value).
+
 ## How it works underneath
 
 A content contract isn't a special new thing — it's an ordinary [LazyGiving assurance contract](../lazyGiving/assurance-contracts.md), organized so that the "project" is your content. The creator owns the contract and receives the money; anyone can start one *for* a creator who hasn't shown up yet, but the funds wait for the creator, who [claims them by proving they control the account](get-your-content-funded.md) — never the third party.
diff --git a/docs/end-user/tldr-for-llms.md b/docs/end-user/tldr-for-llms.md
index bd6eaa3d..fca3ec7d 100644
--- a/docs/end-user/tldr-for-llms.md
+++ b/docs/end-user/tldr-for-llms.md
@@ -7,6 +7,14 @@ Each entry follows the same shape:
 - **When a user encounters it** — UI surfaces or moments in the user's flow where this concept becomes load-bearing.
 - **What they might want help with** — typical assistant tasks around this concept.
 
+## Why benefits fail to become incentives
+
+[Full page](commonality/vision-and-strategy/why-its-better/why-benefits-fail-to-become-incentives.md)
+
+- **What it is:** A strategy catalogue of why dispersed beneficiaries cannot reward better decisions. Two breaks come before ordinary commissioning: a desired kind of value may never have been imagined as fundable or given a concrete funding destination; and many contributors may need to fund qualifying work from many independent producers, without choosing all recipients in advance. Shared funding purposes connect both sides, with explicit criteria, actual money, and allocation decisions.
+- **When a user encounters it:** Evaluating a vertical, or explaining Civility's missing reward.
+- **What they might want help with:** Turning a vague wish into a fundable opportunity; distinguishing support for a producer from funding a kind of value; identifying whose choice changes and why. A pool does not guarantee payment to every qualifying submission, and evidence of work is not proof of its benefit. Examples beyond content are possible extensions, not implemented verification capabilities. Synthetic scenarios that would illustrate those breaks live in `fake-data-generation/incentive-demonstrations.md`, not on this page.
+
 ## Commonality API map (jobs → SDK / HTTP)
 
 [Full page](/docs/end-user/commonality/for-llms.md)
diff --git a/fake-data-generation/PLAN.md b/fake-data-generation/PLAN.md
index 6dcea90c..9f53eb7b 100644
--- a/fake-data-generation/PLAN.md
+++ b/fake-data-generation/PLAN.md
@@ -76,6 +76,7 @@ Do these in order unless Adam names a different one. Each item is a session-size
 5. **[ ] Refresh implication-evaluation corpus** after a model that returns JSON. Root TODO.
 6. **[ ] Volume of simple-cause uniques** (hundreds) only after (1)–(2) prove the loop. Then conceptspace seed, **no** fake users/projects.
 7. **[ ] Stress scale** (`gen:large` today is 100 users). 1000+ users, graphs, indexer deep-compare: not started.
+8. **[ ] Incentive demonstrations** — three synthetic stories (Civility's changed choice, a negative replication, a shared migration where money is not a commitment). Inventory, scenarios, and the authoring template: [`incentive-demonstrations.md`](./incentive-demonstrations.md). Do not put that backlog back into the public strategy page.
 
 ## Pointers
 
diff --git a/fake-data-generation/incentive-demonstrations.md b/fake-data-generation/incentive-demonstrations.md
new file mode 100644
index 00000000..e8f3077f
--- /dev/null
+++ b/fake-data-generation/incentive-demonstrations.md
@@ -0,0 +1,100 @@
+# Incentive demonstrations
+
+Project work that follows from the public strategy page
+[Why benefits fail to become incentives](/docs/end-user/commonality/vision-and-strategy/why-its-better/why-benefits-fail-to-become-incentives.md).
+That page is for readers. This file is the inventory and the scenarios still to build.
+
+Extend the existing fake-data pipelines in [PLAN.md](./PLAN.md). Keep invented
+actors and behavior separate from the accepted real-statement catalogue.
+
+## What the fake-data examples already cover
+
+Reviewed **2026-10-07**. These are synthetic examples and recorded protocol
+exercises, not evidence of real demand or behavioral change.
+
+The [curated project stories](./campaignProjectStories.ts) already include:
+
+| Example | Missing incentive it illustrates |
+|---|---|
+| Shared dependency and Linux maintenance | Many users benefit, but no single user can justify covering the cost. |
+| Independent replication and a neglected-disease dataset | Useful public knowledge has weak conventional rewards or cannot be commercially captured. |
+| A watershed sensor network spanning three towns | No jurisdiction owns the full benefit. |
+| Parts-interchange tables and local construction field guides | Producing checked knowledge costs time, while other businesses can freely copy the result. |
+| Shared factual reviews and cross-camp explainers | Interested people distrust a single camp's ownership of the work. |
+
+The [October 2 step-14 run report](./campaigns/medium-realistic-v2-step14-run.md)
+records a synthetic common-ground project reaching its funding threshold through
+contributions from opposing camps, followed by four retroactive donations. It
+reports successful reconciliation locally and on Base Sepolia. This demonstrates
+co-funding and the reimbursement flow. It does not demonstrate that an actual
+public good was delivered or that the contributors independently judged its value.
+
+The [content-funding fixtures](./contentFundingActions.ts) also include
+prospective rounds and a round that materializes content. They exercise relevant
+machinery, but do not yet supply an explicit Civility story showing a creator
+choosing different work because of the added reward.
+
+The central gap is **a demonstration of the changed choice**. Existing scripts
+prescribe purchases and other actions. A realistic project title plus successful
+transactions does not show why a producer would change behavior.
+
+## Demonstrations worth adding
+
+Proposals, not implemented scenarios.
+
+### 1. Civility: make the alternative worth producing
+
+Give a synthetic writer two possible next pieces, with explicit assumed costs
+and competing rewards. Show dispersed contributors making the respectful piece
+viable, evaluation from the intended audience's perspective, funding, and a
+second production round.
+
+Include an underfunded round and a politely worded but misleading piece that
+fails the selected evaluation standard. Show that an evaluator's approval makes
+work eligible for consideration, not automatically paid. Distinguish producing
+the piece from distributing it and observing its effects.
+
+The demonstration should expose its behavioral assumptions: “Under these assumed
+rewards, this writer chooses this piece.” Scripted behavior illustrates the
+mechanism; it does not establish that real writers will respond that way.
+
+### 2. Replication: reward learning, including unwelcome results
+
+Extend the existing replication story. Contributors fund a preregistered study
+whose public result is useful whether it confirms the original finding or not.
+The synthetic result is negative. An evaluator judges that the promised work was
+performed well, and later donors reimburse early contributors.
+
+This makes the incentive precise: pay for reliable investigation and publication,
+not for producing the answer contributors hoped to hear. Keep reaching the
+funding threshold separate from the later judgment that the study delivered.
+
+### 3. A shared transition: reduce the cost of going first
+
+Several communities want to leave an unsupported shared software tool, but each
+needs migration tooling and support. Their contributions fund those common
+prerequisites. Include a branch of the story where adequate money is available
+but an essential participant still declines to migrate.
+
+This shows both the useful intervention and its boundary. Financing a transition
+can remove a barrier; it is not an enforceable commitment that every participant
+will switch. Any proposed multi-party commitment mechanism should be identified
+as additional design work rather than silently attributed to assurance contracts.
+
+## Template for future examples
+
+For each proposed project or vertical, record:
+
+- **Beneficiaries:** who gains, including people who will not contribute?
+- **Decision-maker:** whose choice determines whether the benefit happens?
+- **Current choice:** what do they do now, and what rewards or constraints explain it?
+- **Missing connection:** why can't beneficiaries already reward the better choice?
+- **Intervention:** which Commonality mechanism changes that situation?
+- **Credibility:** what makes the funding opportunity believable before work begins?
+- **Evaluation:** who judges delivery, by what evidence, and what could fool them?
+- **Demonstration:** what observable transaction or state change exercises the mechanism?
+- **Behavioral assumption:** what response is scripted rather than demonstrated?
+- **Boundary:** what remains unsolved even if the funding works?
+
+The strongest examples let a reader answer the same question from beginning to
+end: **Whose choice changes once this money can reach them—and why?**
diff --git a/specs/product/README.md b/specs/product/README.md
index 284a643a..fdeafb54 100644
--- a/specs/product/README.md
+++ b/specs/product/README.md
@@ -2,6 +2,7 @@
 
 Product-manager-level planning documents. These describe *what* to build and *why*, not *how*.
 
+- **[structured-fundable-work.md](structured-fundable-work.md)** — Generalize content funding into domain-defined work identities, structured descriptions, duplicate-claim rules, and work-item attestations connected to funding requests (high-level direction).
 - **[founder-first.md](founder-first.md)** — The vertical-founder posture: our customer is the founder standing up a vertical, not the end user of the generic sites. Carries the triage rule for all platform work, a map of which doc holds which piece, and the consolidated pivot backlog. Frozen rationale in [ADR 0005](/specs/decisions/0005-founder-first-verticals.md).
 - **[Activating a cause](/docs/founder/activating-a-cause.md)** — The founder's adoption playbook: use the cause's existing distribution, recruit money/judgment/trust/execution one role at a time, and let each visible contribution attract the missing roles.
 - **[causes-as-publications.md](causes-as-publications.md)** — Accepted target model for AI-assisted cause creation: causes are mutable, shareable publications over immutable statements; organizers and vertical operators are distinct roles; anchors remain optional semantic statements rather than page identities. Frozen rationale: [ADR 0009](../decisions/0009-causes-are-publications-over-statements.md). Implementation work: [plan](causes-as-publications-implementation-plan.md). Copy/weighting of that publication: **[cause-page-not-a-club.md](cause-page-not-a-club.md)** (direction; two-step rename; not a sweep yet). Returning-signer home: **[personal-dashboard.md](personal-dashboard.md)** (derived fundable-projects union; not a roster).
@@ -15,6 +16,7 @@ Product-manager-level planning documents. These describe *what* to build and *wh
 - **[content.md](content.md)** — Content bootstrapping: seeding statements, AI-assisted content discovery, solving the empty-field problem.
 - **[ai-assistance.md](ai-assistance.md)** — AI skills for helping users navigate the system (implication attester, alignment helper, etc.)
 - **[bridge-finder.md](bridge-finder.md)** — A focused finder for hidden-majority patterns (speculative)
+- **[durable-participation.md](durable-participation.md)** — A continuing tally across campaigns and time: reusable participation, revocable support, counting rules, organic coalitions, and a proposed demo/experiment (discussion proposal).
 - **[statements-are-peculiar-for-good-reasons.md](statements-are-peculiar-for-good-reasons.md)** — Index for why statement wording is verbose/finicky (implication vs nudge vs modified layer). Read this before writing seed statements or bridge clusters.
 - **[bridge-creator.md](bridge-creator.md)** — Actively synthesizing common-ground statements and getting them in front of people (speculative)
 - **[bridge-building-for-founders.md](bridge-building-for-founders.md)** — Turning the CSM bridge-creator into a building block any cause founder can adopt ("a mediator for your cause"): what's already generic, the four places CSM-ness actually lives, a tiered plan, and why the beat-agent rehearsal gates it.
diff --git a/specs/product/durable-participation.md b/specs/product/durable-participation.md
new file mode 100644
index 00000000..ebaacda7
--- /dev/null
+++ b/specs/product/durable-participation.md
@@ -0,0 +1,226 @@
+# Durable participation: a tally that campaigns can build on
+
+Discussion proposal, 2026-10-07. This develops an existing direction, not an
+implementation commitment. Unique-human verification and federation with external
+petition systems remain future work.
+
+## The idea
+
+**Express a view once, keep control of it, and let future efforts build on it.**
+
+A campaign should not have to reconstruct its entire constituency whenever it
+starts a new petition, changes its wording, or hands the work to another organizer.
+People's recorded views can remain useful across those boundaries. New participants
+add to an existing body of participation; returning participants can refine or
+change their views. Different organizers can discover overlapping constituencies
+without requiring everyone to join the same organization.
+
+Tally could become a continuing reference point for expressed public support:
+somewhere people expect their views to remain available and worth updating. The
+ambition behind “the poll” is understandable, but the more defensible promise is
+**a shared, continuing record of what participants say they believe**, with
+inspectable ways to count it. It does not establish what the whole population
+believes, or create a single authoritative interpretation of everyone's views.
+
+This extends [sign once, stay counted](/docs/end-user/tally/express-what-you-care-about.md),
+[individualization](/docs/end-user/commonality/vision-and-strategy/why-its-better/individualization.md),
+and [organic coalitions](/docs/end-user/commonality/vision-and-strategy/why-its-better/organic-coalitions.md).
+The additional strategic claim is about time: the effort of participating can
+remain useful after the occasion that originally motivated it has passed.
+
+## What set union buys us—and what it doesn't
+
+For a specified set of statements, take the union of their signer sets and count
+each identity once. Signing five included statements contributes one person to
+that union. If another compatible system supplies verifiable records and identities
+that can be deduplicated against ours, its participants can also contribute.
+
+Three different questions need different answers:
+
+| Question | Operation and meaning |
+|---|---|
+| Who signed at least one of these statements? | Union: participation in any of them. |
+| Who signed every one of these statements? | Intersection: explicit agreement with the whole set. |
+| Who supports this shared conclusion? | Direct signers plus deduplicated indirect supporters through accepted implications to that conclusion. |
+
+The union of “build this road” and “do not build this road” is not a constituency
+for building the road. Nor does support for one plank mean support for an entire
+cause board. Deduplication solves repeated counting; implication judgments solve
+a separate, harder problem about meaning.
+
+For a target statement q, a proposed count can be described as:
+
+`support(q, t, rules) = direct(q, t) ∪ union(direct(s, t) for accepted s → q)`
+
+Resolve identities before counting the union. Here `direct` means the latest
+applicable recorded agreement at time t, not every historical signing event.
+The rules specify identity eligibility, freshness, accepted attesters, and conflict
+handling. Existing Conceptspace implications are non-transitive: s → x → q does
+not suffice without an accepted s → q attestation. Indirect support remains an
+inference, not a signature on q.
+
+## Durability does not require numbers that only rise
+
+For fixed statement membership and fixed identity rules, the set of people who
+have *ever* signed grows as new historical records arrive. That is a historical
+participation measure. It is not a snapshot of current agreement.
+
+Current recorded support must be able to fall when someone withdraws or disagrees.
+A freshness window also loses entries as they age. Identity corrections can merge
+previously distinct accounts, and rejected implications can remove indirect
+support. Even historical unique-human totals can be revised downward when their
+identity assumptions change.
+
+The useful promise is continuity, not an upward-only counter. Keep these views
+distinct:
+
+- **Current recorded support:** the latest expressed position, which may be old.
+- **Recently confirmed support:** current support explicitly confirmed within a
+  stated period. An account login does not refresh every belief.
+- **Historical participation:** people who expressed support at some time,
+  including those who later withdrew.
+
+None perfectly reveals present private beliefs. Showing age makes uncertainty
+visible without making everyone repeat a whole questionnaire every month. Stable
+values and expiring requests also differ: “libraries should be free” can remain
+useful for years; “stop next Tuesday's demolition” needs its original date and
+context preserved. New wording does not inherit direct signatures.
+
+## Why this could change participation
+
+The hypothesis is that people will spend more effort expressing a precise view
+when they expect it to keep doing useful work. A short answer today could help a
+later project find an audience, reveal agreement with an unexpected group, or
+answer a new question through an inspectable implication. Someone can add a few
+positions at a time rather than complete an exhaustive survey upfront.
+
+Durability also makes recruitment cumulative. An organizer can bring in new
+participants without losing everyone reached by the previous organizer. If people
+see others using these counts to discover projects or allocate attention, they
+have a reason to make their own views visible. Competing camps may then recruit
+their constituencies because absence leaves their positions underrepresented in
+this particular record.
+
+That feedback loop is plausible, not automatic. Organizers may prefer exclusive
+lists, dispute the counting rules, or avoid a venue where opponents dominate.
+More participation can reflect better mobilization rather than persuasion.
+Recruitment from several camps still does not produce a representative sample.
+There is no election-like guarantee of equal access, shared eligibility, or binding
+consequences. A large tally demonstrates recorded support among participants;
+it does not by itself establish a population majority or a mandate.
+
+This fits the [founder-first strategy](founder-first.md): give each founder a
+useful constituency view for their own cause. Shared infrastructure lets that work
+benefit other causes. Becoming a widely used reference point is a possible result
+of those local successes, rather than a prerequisite or an umbrella growth plan.
+
+## The connection to organic coalitions
+
+The most interesting outcome is a growing map of specific agreements that cuts
+across familiar camps. Someone can disagree strongly with another group on one
+issue and discover that many participants in both groups explicitly favor the
+same practical outcome on another.
+
+That requires more than a large union total. Show which specific statements
+connect the groups, how much support is direct versus inferred, and what remains
+disputed. Group membership should be self-described or based on disclosed signed
+statements, not silently assigned political identities. Overlapping groups need
+overlap counts; their totals cannot simply be added.
+
+When the bridge is only inferred, invite participants to read and sign the exact
+bridge statement. That converts a hypothesis about agreement into explicit
+agreement. Signing a mediator's modified statement must remain a new choice;
+signing its natural parent cannot be treated as acceptance of the modification.
+
+This can make fine distinctions rewarding: “I agree with you about this outcome”
+need not become “I endorse your whole program.” It can reveal useful common ground
+even when there is no majority and no resolution of the larger dispute. Whether
+seeing this changes attitudes toward opponents is a further hypothesis to test.
+
+## Product suggestions
+
+1. **Make every displayed count explainable.** Show statement scope, account versus
+   verified-human unit, direct and indirect support, deduplication, timestamp,
+   freshness rule, and attester policy. Direct and indirect breakdowns must avoid
+   double counting people in both. Shareable snapshots should preserve the rules
+   used; personal trust settings can legitimately produce different results.
+2. **Make returning worthwhile.** Provide a compact review of a person's signed
+   statements, easy withdrawal/disagreement, and optional reminders to confirm
+   selected older views. Show useful new connections to their existing views.
+   Avoid treating silence as a fresh confirmation.
+3. **Handle conflicts before claiming current human support.** Propose that explicit
+   disagreement with q overrides inferred support for q. Linking accounts with
+   conflicting positions also needs a disclosed resolution rule and a way for the
+   person to correct it. Audit existing behavior before implementing either.
+4. **Let organizers reuse a count across campaigns.** A shareable view over a
+   specified statement set or conclusion can preserve continuity while individual
+   campaign pages come and go. Changes to the included set must be visible; adding
+   a popular statement is not evidence that anyone changed their mind.
+5. **Prototype common-ground views.** Show agreement on concrete statements across
+   two voluntarily identified constituencies, alongside disagreement and unanswered
+   questions. “Not signed” is not “opposed.” Keep signing separate from subscribing,
+   permission to contact, willingness to work, and willingness to fund.
+6. **Treat federation as a later interoperability project.** Imports require
+   original wording/version, source provenance, timestamps, evidence of signing,
+   updates and revocations, and participant-authorized identity linking. An
+   external signature does not automatically authorize new uses or public linking.
+   Without a reliable cross-system identity mapping, report source counts
+   separately; do not add them into a purported unique-human total.
+
+Human uniqueness, account ownership, current belief, and eligibility for a local
+count are separate claims. Identity providers may overlap without a way to detect
+it. Unlinked accounts cannot be assumed to be additional verified humans, nor can
+we assume malicious accounts will voluntarily link later. Linking political
+histories also creates privacy consequences; a proof of uniqueness does not by
+itself solve cross-statement privacy. These are design requirements for later work,
+not reasons to defer a clearly labeled account-based prototype.
+
+## Documentation suggestions
+
+- Revise Tally's “forever, with no re-signing” promise to explain continuing recorded
+  support, withdrawal, and optional freshness confirmation. Remove the assurance
+  that future identity linking removes today's incentive to create fake accounts.
+- Distinguish implications from equivalence: two statements can imply the same
+  broader conclusion without meaning the same thing.
+- Qualify “hidden majority” language: the system can expose an observable
+  constituency; majority claims require an appropriate population denominator and
+  evidence about that population. Separate support from demonstrated funding.
+- Add the time dimension to the vision's organic-coalition and immediate-value
+  explanations after accepting this proposal. Signing can add zero to a union
+  where that signer is already counted, while still adding a useful direct signal.
+
+## A small demo and a useful experiment
+
+Extend the existing [tiny/demo pipeline](/fake-data-generation/PLAN.md), using
+accepted statements rather than inventing another bridge-generation pipeline.
+Keep all identities and activity clearly fictional. Model missing future features
+as labeled fixtures, not as claims about working identity verification.
+
+A deterministic story with six fictional people is enough:
+
+| Event | Expected result |
+|---|---|
+| Campaign A gets signatures from people 1, 2, 3; campaign B from 3, 4 | Five signatures, four people in the union. |
+| Person 3 links a second account that also signed B | Account count can differ; verified-person union remains four. |
+| Person 5 joins through a later campaign included in the view | Union becomes five without asking the earlier people to sign again. |
+| Person 2 withdraws their only qualifying signature | Current union returns to four; historical participation remains five. |
+| Person 1's signature ages beyond the freshness window | Current recorded union stays four; recently confirmed count falls. |
+| Person 6 signs an opposing statement | Does not enter the target's supporter count merely by participating in the topic. |
+
+Add an accepted bridge with direct and indirect support from participants in both
+camps, one person who explicitly rejects the bridge despite an inferred path,
+and a contested implication whose removal changes the inferred count. Present
+this as a counting scenario; don't invent implausible beliefs merely to fill cells.
+
+Then try one cause across two recruitment occasions. Measure how much participation
+can be reused, whether people understand why they count, whether they can correct
+it, and whether they voluntarily answer another precise question after seeing a
+useful connection. Ask participants from different camps whether the view fairly
+represents them and reveals an agreement they had missed. Track recruitment,
+reconfirmation, changed beliefs, and changed counting rules separately.
+
+The proposal becomes more convincing if reuse reduces organizer work and people
+understand and value the continuing record. If they mostly feel misrepresented,
+ignore review prompts, or see no reason to add detail, durability alone has not
+earned the stronger adoption hypothesis. This experiment can start with account
+counts; human verification strengthens the headcount claim later.
diff --git a/specs/product/structured-fundable-work.md b/specs/product/structured-fundable-work.md
new file mode 100644
index 00000000..a9339a61
--- /dev/null
+++ b/specs/product/structured-fundable-work.md
@@ -0,0 +1,49 @@
+# Structured fundable work
+
+High-level product direction, captured 2026-10-08. Detailed design and implementation remain future work.
+
+Generalize the idea behind [content funding](../tech/subsystems/content-funding/README.md) into an extensible system for identifying and describing fundable work. Content funding gives individual content items canonical identities and explicitly connects them to funding requests. Other domains could benefit from the same structure: concerts, software maintenance, service delivery, public-access releases, and many more.
+
+The purpose goes beyond making it easier to pay people. Once a recurring kind of activity can be discovered, evaluated, and funded consistently, alignment attestations can let movements support the qualities they want across a whole ecosystem of independent providers.
+
+## Identity belongs to the work
+
+A work item should have an identity independent of the assurance contract requesting money for it. Creating another request for the same work should not automatically create another work item.
+
+For example, a violin concert by a particular performer in Cedarvale Park on October 8, 2026 could have a structured identity; next week's concert would be a different item. Maintenance of a particular software repository for October 2026 could likewise be identified separately from November's maintenance.
+
+Funding requests should explicitly reference their work items through queryable data, with enough on-chain representation for contracts to reject duplicate claims under the applicable rules. An observer should have an obvious place to find an item's current funding request and funding history. Identity and funding policy are separate: retries after failed requests, active versus successful claims, and legitimate divisions of work need appropriate rules rather than one universal prohibition.
+
+## Open-ended identity schemes
+
+Users and communities should be able to define domain-specific canonical-ID schemes. A scheme determines which fields identify the work and how they are normalized; Commonality's author need not anticipate every domain.
+
+Canonical means canonical within an identified scheme. Interfaces and communities can recognize useful schemes without requiring one universal authority. This need not prevent every abuse: making duplicate funding harder, easier to notice, or easier to demonstrate is already valuable. Competing schemes, changed details, and overlapping work are design questions to resolve later.
+
+## Who the work points at
+
+Content funding gets a payee for free: the canonical id embeds the channel, and proving control of that channel is enough to withdraw escrowed funds or to take over contract creation. See [channel claiming](../tech/subsystems/content-funding/channel-claiming.md). That pattern fits work whose identity already names a single party allowed to do it or dispose of it: a named performer, a repository's maintainers, a copyright holder releasing a specific work. The scheme can embed that rights-holder, and funds raised by a third party can wait until they prove control.
+
+It does not fit every category below. An open slot (any tutor for this session, any facilitator for this meeting) names a role, not a person. The work id still should not pretend to be a channel. Filling the slot is a separate choice of beneficiary on the LazyGiving project: Alice can create the project and name Bob, using the existing [beneficiary-id](../tech/subsystems/claimable-beneficiaries.md) system. There is no identity for "the human named Bob Smith." What exists is "the owner of this channel" or "the controller of this domain" (`x` / `youtube` / `substack` / `dns`), which is often enough. Bob proves control of that id and then claims or refuses. The same split applies when the doer is not derivable from the work at all: an independent check should not pay the owner of the thing being checked, and a world outcome (a restored habitat, a removed barrier) has no owner. A steward's permission to act is not proof that the change occurred. In those cases assignment or acceptance still sits between "this is the work" and "this beneficiary may withdraw."
+
+## Structured descriptions and creation forms
+
+Work types should also supply recognizable categories and clearly specified fields. For example: maintenance → open-source software → repository → maintenance period, with additional fields for scope and commitments.
+
+Only some description fields need determine identity. Structured forms improve discovery and comparison while helping providers describe their work without inventing everything in free text. They also communicate that funding this kind of work is an established, intended use of the system. Additional prose can remain available.
+
+### Some kinds of work that might be worth exploring
+
+  - Delivering services: e.g. tutoring sessions, translation assistance, repairs, mentoring, community transport
+  - Making existing things publicly available: e.g. releasing a book under an open license, opening a dataset, publishing source code, making an archive freely accessible.
+  - Producing evidence and checking claims: e.g. reproducing research, testing product claims, auditing accessibility, checking public records, or evaluating whether another funded project delivered.
+  - Hosting encounters and collective activities: e.g. public deliberations, workshops, community meals, skill exchanges, and cross-group discussions. (For Civility specifically, this would extend the scope from funding what people publish to funding **how people actually interact**. More generally, it gives movements a way to sustain local conveners.)
+  - Achieving and sustaining changes in the world: e.g. restoring a habitat, removing an accessibility barrier, or reducing a documented source of pollution. (This has potentially very broad reach: a movement could invite anyone capable of producing the desired change to do so. But it is also the hardest category here. Defining an outcome does not automatically make causation, measurement, or payment allocation straightforward.)
+
+## Attestations follow the work to its funding requests
+
+Allow alignment attestations to refer to individual work items, sequences, or other relevant groupings independently of a particular funding request. Use explicit item-to-project relationships to surface the corresponding assurance contracts on fundable-projects boards, extending the content-funding pattern.
+
+The existing general subject-attestation model is a starting point; a future implementer should verify which identities and discovery paths are actually supported. Define how item and sequence attestations affect project inclusion without silently treating endorsement of one item as endorsement of every item in a bundle, or of all future installments.
+
+Keep work identity, description, funding eligibility, delivery evidence, and alignment judgments distinct but connected. This is an extension of how work is represented and connected to funding, not necessarily a new financial mechanism.