Skip to Fulmerica links
Worth BuildingIndexStarted 4 October 2026

What this technology could be for, one community at a time

A catalog of things worth building with AI and the tools around it, organized by the people they would serve. Each chapter goes deep enough to act on: the facts as of this year, the specific things to build, who is already working on them, and where a reader could start this weekend, this season, or as a career. It exists to show that a hopeful alternative is buildable, and to give anyone who wants to build it a place to begin.

01The idea

Technology should make people more capable together, not just more convenient alone

A more capable system that leaves people more isolated and more dependent is a step backward, whatever its benchmark scores.

Most conversations about AI ask how far it can go. This catalog asks what it is for, and answers with specifics. The commitments behind every chapter are few and plain.

  • Judge by what happens to people. Progress is measured by what people and the relationships among them can do afterward, not by what the technology can do.
  • Keep people in charge. Systems should be legible, put human judgment where the stakes are, and always let people leave with their data.
  • Build capability, not dependence. Success means people understand more and can do more, not only that they do less.
  • Make slow harm visible before it becomes normal: costs that creep, benefits quietly lost, stores closing one at a time.
  • Design for groups. Households, teams, neighborhoods, and towns are users too. The bonds among people are worth measuring.
  • Remove the effort that crowds out human time; keep the effort that builds relationships and judgment.
  • Treat pace, direction, and who benefits as choices, and point to the working alternatives: credit unions, agricultural cooperatives, Linux, Wikipedia.

Three questions for anything in this catalog

More capable, or more dependent?
Can the people who use it explain, catch, and override what it does?
Closer to each other, or closer to the system?
Does time saved show up as time with people, or as more time on screens?
Can people see it, steer it, and leave?
Plain-language decisions, approval at the expensive moments, portable data, no lock-in.
02How to read a chapter

Every chapter has the same shape, so they can be compared

A chapter names a community, shows what the default path and the alternative look like there, maps eight to twelve specific opportunities with current facts and scores, describes an ordinary day when the technology is pointed at the right target, and ends with where to start and who is already doing it.

The score

Each opportunity is scored from 0 to 2 on five questions, for a maximum of 10. Is there a shared layer that nobody owns? Is there a buyer or a deadline within two years? Does using it produce consented records that make a shared dataset more valuable? Does it pass the three questions above? Does it fit a builder who convenes rather than competes?

Nine reusable parts

Chapters mark which of these parts each opportunity reuses. They were first worked out for people with disabilities facing a Medicaid deadline, and they port almost everywhere.

Notice readerA letter or PDF in; who, what, which program, which dates out; explained plainly with a pointer to the source.
Evidence fileWhat proves the claim, who holds it, request letters, templates in the institution's own wording.
Deadline engineThe clocks that preserve benefits and the right to be heard.
Place configurationOpen, versioned rules for each jurisdiction, checked weekly.
Outcome commonsConsented, anonymized records of what institutions did, by place and month.
Care circleRoles and permissions around one person, who can remove anyone and leave.
Records-request pipelineTemplates, tracking, and publication of public-records requests.
Standards cohortFund small tools to adopt shared formats, then win a public buyer.
Member-owned dataA cooperative that holds the shared records and returns value to the people in them.

Why these communities were left behind, and what changes

Five structural causes defeat both venture scale and single-agency scale: fragmentation across tens of thousands of jurisdictions; work that depends on trusted relationships; no payer at the point of need; paperwork and expertise too costly per case; and groups too small to be a market or a program. Models that read and draft at about twenty cents a case, rules kept as open configuration, institutions that lose when people fall through becoming the payers, the group as the unit of design, and a federation that lets many small programs share one core change the economics of each. The chapter "Where Scale Failed" lays this out.

03The chapters

Twenty-three chapters

Every chapter is complete and sourced, and a sample of its figures has been checked by a second, independent pass. All twenty-three are published and checked. Source pages could not be opened during the checks of 5 October 2026, so those checks compared figures against search results; where a figure could not be confirmed, the chapter attributes it to the organization that reports it and says so in its note.

No.ChapterStatusWhat it covers
0Where Scale FailedPublishedThe five causes, the nine parts, and sixteen communities where the same playbook ports. Read this first.
1Disability and AccessPublishedTwenty-four opportunities in seven territories: proof and recourse, portable identity and exit, care circles, place and movement, money and work, public AI, and measurement.
2Food SystemsPublishedFood benefits as notices, small farms and the 2028 traceability deadline, the daily meal visit, the last grocery store, food as medicine, eaters and farmers, restaurants and platforms, school meals, food banks, home kitchens, surplus food, local food governance.
3Care and AgingPublished63 million family caregivers, the direct-care workforce, self-directed services, supported decision-making, the daily visit, fraud and isolation.
4Learning and SchoolsPublishedTeachers' time, the IEP team on a data standard that now exists, small districts, tutoring versus explaining to each other, adult literacy, libraries.
5Home and HousingPublishedTenants facing eviction alone, by-name lists, vouchers, small landlords, accessible units, resident-owned communities, insurance, legible zoning.
6Work and OwnershipPublishedBusiness succession into employee ownership, worker and platform cooperatives, portable benefits, algorithmic management, skills that travel.
7Justice Within ReachPublishedCourts without lawyers, debt and eviction defense, record clearing, veterans' claims, disaster appeals, fair hearings, and who may help.
8Civic Life and Local NewsPublishedNews deserts, paid documenters, small governments, deliberation tools that interoperate, decisions made legible, public comment.
9Health at the Clinic DoorPublishedRural health and the $50 billion transformation fund, community health workers, denials and appeals, referrals between clinic and community.
10Money and the Safety NetPublishedThe notice economy across programs, benefits cliffs, savings accounts for people with disabilities, stolen benefits, the data matches that work.
11Language and BelongingPublished26 million people with limited English as federal language access retreats; a community-owned translation memory; immigrant benefits.
12Place and MovementPublishedParatransit, sidewalks and transition plans, transit data standards, venue accessibility commons, small-town infrastructure.
13Land and FarmingPublished300 million acres changing hands, farm data ownership and repair, traceability, farmworkers, land transfer navigators.
14Friendship and LonelinessPublishedLoneliness, companions, fraud, trusted circles, and a certification for technology used with isolated people.
15Neighbors and DisastersPublishedDisaster recovery as FEMA work shifts to the states: the denial letter and its 60-day appeal clock, the recovery file a family keeps, shared cleanup lists, resilience hubs and microgrids, congregations, 211, community fridges, volunteer fire and CERT, buyouts.
16Building in the OpenPublishedHow small teams build shared tools that outlast their funders: rules as open code, one directory format for services, decision records, reviewing AI-written code, paying maintainers, procurement that buys open, many programs on one core.
17Public AI and DataPublishedWho can say yes to their data, who gets paid, and who can build: data cooperatives, consent signals and licensing, language data governed by its speakers, public models and computing, nonprofit access, deletion rights, incident records.
18Measuring What MattersPublishedMeasures of capability, connection, and exit held by someone other than the seller: a public index of automated-decision audits, AI content labels people notice, a trust mark with an exit line, teen wellbeing, social effects in AI testing.
19Young People in TransitionPublishedAges 16 to 26, when documents, coverage, credit, and skills pass from an agency to the young person: the exit folder from foster care, Medicaid to 26, extended care, credit freezes, first jobs, the care circle at 18.
20Tribal NationsPublished575 nations as 575 governments, with every program led from within: broadband offices and grant clocks, consultation records, data offices under the CARE Principles, language recordings under the nation's license, census counts, ICWA notices.
21Veterans and Their BenefitsPublishedThe free claim-help network and the paid consultants who grew in its gap: PACT Act claims, rating letters and appeal clocks, software in the claim, county veteran service officers, caregivers, benefits lost to rule changes.
22Small OrganizationsPublishedAbout 1.5 million charities, most spending under $500,000 a year: fiscal sponsorship and H.R. 9721, a ledger of terminated federal grants, shared back offices, volunteers and AmeriCorps, one funder profile, mergers and closures.
Why the same parts keep appearing

Every chapter scores its opportunities on the same five questions and marks which of the same nine parts each one reuses. That repetition is deliberate. A notice reader built for a Medicaid letter also reads an eviction summons; an outcome commons built for appeal results also tracks paratransit trips. A reader who learns the parts in one chapter can carry them to a community this catalog does not yet cover.

04Where anyone could start

Three sizes of first step, in every chapter

A weekend

  • Photograph and transcribe a notice you or a neighbor received, with identifiers removed, for a shared test set.
  • Map one thing in your county that no one has mapped: grocery access, a waitlist, a closed unit.
  • Document one public meeting.
  • Export your data from a service and see what travels.

A season

  • Run one chapter's smallest play with one partner organization for ninety days and record outcomes.
  • Write the open configuration for your state or county and get a practitioner to check it.
  • File ten records requests and publish what comes back, including the silences.

A career

  • Lead a program for a community you are from, on the shared parts.
  • Convene a standards cohort where small tools cannot talk to each other.
  • Found a cooperative that holds a community's data and returns the value.
05Contributing a chapter

The guide chapters are written to

Every chapter follows one written guide: the structure, the score, the parts, the rules for facts (every number traces to a source; nothing is invented), and the rules for writing (plain sentences, no slogans). To propose a chapter, email Dylan. Chapters were drafted with AI research assistance and are checked by a person before they are marked published. Drafts say so.

Provenance. Facts are as of the dates stated in each chapter. Scores are judgments made consistently, not measurements. Where a figure comes from advocates or industry, the chapter says so. Chapters name real organizations as examples of work already underway; inclusion is not endorsement, and no organization has reviewed its mention.