Homecastr, a new real estate tool that surfaced on Hacker News this week, lets users sweep across an interactive map to see AI-generated home price forecasts by neighborhood — a different take on the predictive pricing that Zillow popularized with its Zestimate, but aimed squarely at what's coming rather than what things are worth right now.
The interface is the main pitch. Instead of typing in an address and getting back a single number, you get a map. Forecasts are layered geographically, so a buyer can scan across a metro area and pick out where values are expected to move without running individual lookups. The forecasting engine draws on standard AVM (automated valuation model) methods — machine learning trained on historical sales data, local market signals, and broader economic indicators. That's the same basic toolkit the big players use; the novelty here is presenting the output as a spatial layer rather than a search result.
The project landed as a Show HN submission, Hacker News's format for indie developers and small teams showing early work. The minimal page content — a loading dashboard backed by a heavy JavaScript client — puts it firmly in MVP territory. Data coverage is the obvious question for any small team going up against institutional players with deep data agreements and years of model tuning.
The map-first approach is genuinely distinct from how most real estate tools serve forecast data, and it's easy to see the appeal for someone trying to get a read on an unfamiliar market quickly. Whether the underlying model is accurate enough to drive real decisions is the harder test — and the one that will separate a useful product from an interesting demo.