Skip to main content

Scope & roadmap

Pilot today. Outlook tomorrow. Discipline always.

Two surfaces on this page. The North Star below says where we are going and why. The four bands beneath it show the discipline we are building it with — what is verifiable in the repo today, what's in active development, what is planned if signal supports it, and the things we are deliberately choosing not to build.

Where this is going

Build the layer the country needs. Earn it in Ohio first.

Every parcel decision in America starts with the same question: do I actually trust this data? Today, for most decisions outside the residential mainstream, the honest answer is — not really. You either pay a consultant to spend two weeks assembling the right steward sources by hand, or you accept an AVM's answer and silently hope the assumptions hold. There is no middle road that survives a county attorney's review.

Strata88 builds that missing middle. Every claim in every brief carries its source — the steward, the ingest run, the artefact hash — rendered, not described. The bet is that this pattern, done right in one state, becomes the way every land-touching application in the country eventually checks its work. We chose Ohio because 88 counties is the right unit of difficulty: dense enough that the discipline shows, varied enough that the pattern proves out, and home to a land-bank movement already looking for tools that can survive a county attorney's review.

Four rungs

From a county brief to the layer the country leans on.

  1. TODAY
    Show your work, on every parcel.

    A broker, an attorney, or a land-bank director opens a brief and sees not just the answer but where the answer came from. Every flood line, every soil class, every owner trail carries its steward, its ingest run, and its artefact hash. In production today, across the Hocking pilot — Franklin, Vinton, and Athens next.

  2. NEXT
    Stop checking. Get told.

    The same brief, but it watches itself. When FEMA republishes the flood layer or ODNR updates a well record, the people who actually care about that parcel get a signed, idempotent alert — not a re-marketing email. The notification spine shipped this quarter; the watch-list UX is on the workbench.

  3. THEN
    One source of truth across the deal.

    A broker, a buyer, a closing attorney, a lender, and a land bank all read from the same provenance-clean brief. The brief stops being a deliverable handed around and starts being the handoff itself. Title, escrow, and county recorders integrate with it instead of around it.

  4. HORIZON
    The layer they reach for, not the one they rebuild.

    The next decade of land-touching software — title, lending, conservation tech, climate-risk underwriting — won't roll its own parcel-data layer any more than today's apps roll their own payment processor. They will call Strata88. The briefs become a side-effect; the steward graph is the product.

Why now

Three reasons this works now and didn't five years ago.

  • Verifiability just got rare and valuable.

    LLMs sorted every B2B surface into two camps: the ones that paraphrase, and the ones that prove. The proving lane is wide open — and parcel data is one of the cleanest places to occupy it.

  • Counties are publishing faster than anyone is assembling.

    State and county GIS programs are putting out more structured parcel data every year. The integration tax is everyone's problem and nobody's job. That gap is a moat for whoever decides to do it well.

  • The fastest-growing land segments want provenance, not automation.

    Land banks, probate, conservation easements, climate-driven dispositions. These are customers AVMs were never built for. They are exactly who we built this for.

Strata88 is not a feature; it's a layer. We're looking for two kinds of people right now. One: design partners — land-bank directors, probate attorneys, and rural-county brokers in Ohio with a real transaction in flight, willing to see what a counsel-clean brief looks like inside their workflow. Two: investors who back verifiable-data infrastructure instead of outcome-promising AVMs. If either is you, the buttons below land in a human's inbox.

Become a design partner Investor conversation

Direct human reply. No deck-spray, no marketing list.

Discipline against the ambition

Four bands. Receipts on the first, × marks on the last.

Each band below is sourced from src/data/scope.ts. Every item in Shipped cites a real path in Projects/BULU/. Every item in Not in scope states the principle being preserved, not the "for now" we'd say to a customer in conversation.

Shipped today

Verifiable in the repo right now.

Each item below maps to a file, route, or test in Projects/BULU/. No screenshots, no "trust us" — open the source path and read the code.

  • Steward-attributed parcel briefs (Land Card)

    Every field in every brief carries its steward, ingest run, and source-artefact sha256 — rendered, not hand-waved.

    source: Projects/BULU/app/viz/land_card.py

  • Counsel-gated field rendering

    Commercial-maturity tier + counsel-hold flags gate every overlay value at render time. There is no `--override` flag anywhere in the brief renderer.

    source: Projects/BULU/app/services/contracts/rollup_field_dictionary.py

  • Catalog-driven overlay viewer (MapLibre POC)

    Single manifest drives the layer list and the steward metadata. Adding a steward source in BULU surfaces here with no UI code change.

    source: Projects/BULU/app/api/viz_routes.py (GET /viz/poc)

  • Cesium 3D pilot scene

    Bounded pilot-parcels GeoJSON over OSM imagery with optional ION terrain. Caps response size so a public demo can't be DOS'd.

    source: Projects/BULU/app/viz/cesium_poc.html

  • Diligence-diff notification spine

    Email + Slack transports with idempotent delivery, modular recipient policy, and a no-op default. Wired through Celery.

    source: Projects/BULU/app/services/diligence_diff/notifier.py

  • Brief share-link UX with audit logging

    Tokenized share links to read-only briefs; every fetch is audit-logged for counsel review.

    source: Projects/BULU/app/api/viz_share_routes.py

  • Deterministic Land Bot (LB0)

    Tool calls to deterministic Python templates — no LLM narrator, no paraphrasing. Every response cites the tool that produced it.

    source: Projects/BULU/app/mcp/tools_impl.py (parcel_provenance_chain_tool)

  • Stewardship CI gates

    Catalog snapshot drift detection + commercial-gate validation run on every PR. Blocks merges that strip provenance metadata.

    source: Projects/BULU/scripts/validate_catalog_commercial_gate.py

In active development

On the workbench this quarter.

Honest progress, no ship dates we can't keep. Each item points at the doc or file driving the work.

  • Franklin, Vinton + Athens snapshot expansion

    Catalog and parcel snapshots beyond the founding Hocking pilot — Franklin and Vinton in partial lab build, Athens onboarding as the newest. Same pipeline, no per-county code paths.

    source: Projects/BULU/docs/ops/county_manifests/

  • Parcel-watch overlay rollups

    Background materialization of per-parcel overlay snapshots so diligence-diff alerts can compare runs without re-querying overlays at notification time.

    source: Projects/BULU/app/services/rollups/materialize_parcel_overlay_rollups.py

  • Heir liability shield product spec

    Probate-adjacent product surface — capability scoping in progress; counsel review precedes any code.

    source: Projects/BULU/docs/ops/HEIR_LIABILITY_SHIELD_PRODUCT_SPEC.md

  • Tier-2 land-bank GTM materials

    Outreach narrative and demo collateral for county land-bank directors beyond the founding pilot.

    source: Projects/BULU/docs/ops/LAND_BANK_TIER2_GTM_PLAN.md

  • Probate outreach productization

    Operationalizing the manual probate-attorney workflow into a repeatable checklist with stewardship audit trail.

    source: Projects/BULU/docs/ops/PROBATE_OUTREACH_EXECUTION_CHECKLIST.md

  • Public Hocking showcase plan

    The execution path that turns demo.strata88.com from a tunnel-only showcase into a publicly-shareable evaluator URL.

    source: Projects/BULU/docs/ops/PUBLIC_HOCKING_SHOWCASE_PLAN.md

Planned (counsel-cleared, conditional)

Where we go if signal supports it.

Each item below has a written plan or design doc and a clear gating condition. None of these ship before the conditions are met.

  • Live MCP Land Bot interactability

    Browser-facing Land Bot panel hitting the real MCP server. Gating signal: a customer asking for it. Plan is written; held until then.

    source: Projects/BULU/docs/ops/STRATA88_LIVE_LANDBOT_INTEGRATION_PLAN.md

  • Managed 24/7 cloud host for demos (Option B)

    Migrate the showcase from a Cloudflare-tunneled operator machine to Fly.io ($7–30/mo). Trigger: a paid customer pilot or sustained cold-outreach traffic.

    source: Projects/BULU/docs/ops/STRATA88_PUBLIC_DEMO_DEPLOYMENT_PLAN.md §3

  • Cesium ION terrain in the public scene

    Real elevation data in the 3D scene instead of the ellipsoid fallback. Trigger: licensed Cesium ION terrain layer cleared by counsel.

  • Branded MapLibre base style

    A custom map style replacing OSM imagery. Phase-2 polish — does not gate the product story.

  • LLM-narrator Land Bot (LB4)

    An optional narrative layer above the deterministic LB0 templates. Gating signal: counsel sign-off on a paraphrasing-disclosure model AND a customer use case the deterministic layer can't serve.

  • Multi-state expansion beyond Ohio

    The same pipeline applied to a second state. Trigger: every Ohio pilot county is in the catalog AND a paying customer requires it.

Explicitly NOT in scope

What we are choosing not to do.

These are deliberate exclusions, not omissions. Saying "no" out loud is the cheapest way to keep the product honest and counsel-clean. We will not be talked into these by demand alone.

  • Title insurance or underwriting of any kind

    We attribute. We don't insure. Bundling underwriting into the brief would create liability that the stewardship chain does not support.

  • LLM-generated narrative briefs without per-field provenance

    Every value in every brief must resolve to a steward, an ingest run, and an artefact sha256. Free-form paraphrasing is incompatible with that invariant.

  • Predictive valuation or automated AVM

    Valuation is an outcome promise. Strata88's job is to give a broker, attorney, or land-bank director the receipts they need — the valuation is theirs to make.

  • "Find me hot deals" outcome promises

    The marketing site and the product are deliberately free of outcome-style rhetoric. We surface capability, not lead generation.

  • Any bypass path for counsel-hold flags

    No `--override`, no `--skip-counsel`, no admin button. If a field is held, it does not render. There is no escape hatch.

  • Multi-state launch before Ohio's 88 counties are covered

    The brand name is Strata88 for a reason. Scaling east-west before Ohio is complete dilutes the catalog discipline that makes the product defensible.

Adjacent

Where to go next.

Live demos →

The two POCs (Viz overlay viewer + Cesium 3D pilot scene), with deployment status and a one-edit path to flip them live.

How it works →

The four architectural pillars (counsel-clean by design, steward-attributed every claim, etc.) on the home page.

Engineering →

Test counts, CI gates, observability. The numbers behind the "verifiable in the repo" claim above.

Counsel rationale and the rules every entry above must follow live in Projects/BULU/docs/ops/STRATA88_MARKETING_COUNSEL_CHECKLIST.md §4d.