A foundation lab for physical development
Building the benchmark, training data, and model that predict how places change.
Homecastr pairs dated institutional forecasts with realized outcomes to create a time-valid benchmark, training corpus, and built-environment model. Parcel and neighborhood forecasting is the live execution wedge, not the company boundary.
Why Homecastr exists
Public institutions have spent decades publishing forecasts about population, housing, land use, transportation, infrastructure, utilities, hazards, and development. Observed history supplied the outcomes, but the original expectations remain scattered across reports, spreadsheets, archives, and incompatible geographic definitions.
The missing asset is a normalized pairing of what was expected with what happened next. Homecastr is building that point-in-time training and evaluation surface so planners, utilities, developers, lenders, and investors can test long-range claims against the physical world.
What exists now
Current proof demonstrates national-scale data, model serving, and forecast evaluation.
50M+
Parcel observations
14M
Production requests each week
~7%
One-year-ahead median error
Live
Forecast product and API
Production requests measure system activity, not people, customers, demand, or revenue. Homecastr remains pre-revenue. The national forecast-outcome benchmark is still being built.
How the current product supports the broader model
Live forecast layer
Parcel, tract, and neighborhood forecasts make geography, time horizon, and uncertainty explicit through lower, median, and upper modeled outcomes.
Time-valid benchmark
Historical public forecasts can be preserved with their original publication date, target date, geography, assumptions, and realized outcome.
Additive institutional validation
Agencies can retain their consultants and planning models while adding an independent forecast, calibration history, uncertainty range, and measured performance.
Arizona provides a bounded validation wedge across four public forecast vintages. It is not the completed national corpus, and national normalization remains financed work.
Founder
Daniel Hardesty Lewis
Founder, Homecastr
Daniel previously built climate, land, water, and disaster-planning models under DARPA's World Modelers effort, including 1M-node-scale jobs on Frontera, a top-5 most powerful supercomputer in the world.
Previous affiliations
Support and credits
We publicly acknowledge documented communities and infrastructure-credit providers whose support can be described accurately. People and organizations we have spoken with are not presented as customers, partners, or endorsers unless that status is explicit.
View credits and acknowledgmentsWork with us
Institutional validation
Explore independent forecast validation and calibration for planning, utilities, infrastructure, and institutional analysis.
daniel@homecastr.comBuild with forecast data
Access parcel and neighborhood forecasts through REST and MCP interfaces.
View API documentationJoin the team
See the current roles supporting product, platform, institutional validation, and applied modeling.
View careersExplore the live forecast layer.