Geospatial News Aggregator

Add your feed - @geobabbler, @jamesfee or create an issue on Github.

List of feeds | Raw RSS | Briefings

GeoFeeds Daily Briefing — Tuesday, November 17, 2026

Covering posts from 0800 ET November 16 to 0800 ET November 17. Sources: 168 geospatial feeds.


Three Topics That Stood Out

1. Agents Need Plumbing, and Ground Truth

Esri published two posts on its new ArcGIS agentic developer toolkit, including a Developer's Lounge piece where William Bohrmann explains how the team built it, tagged with agent skills and MCP server. Microsoft opened an AI Workflow preview in Planetary Computer Pro and was frank about why. Running its geospatial models from Microsoft Foundry means defining areas, time ranges and formats, partitioning regions into processing units, calling model endpoints, monitoring jobs and storing outputs. One inference request hides a whole distributed system. The OGC blog supplied the reality check. Dhananjayan KJ built a Stuttgart prototype that joins federal and state building law, development plans and the ALKIS cadastre to answer what can be built on a plot. Questions that take 11 to 21 working days across two city offices came back in under 90 seconds, and the best lessons came where the cadastre and the AI disagreed.

Why this matters: The MCP thread is turning into vendor product, with Esri shipping agent skills and Microsoft wrapping the orchestration. Stuttgart shows what toolkits can't ship. An agent is only as good as the authoritative data it's checked against, and the disagreements are where the real work lives.

2. Commercial Buyers Show Their Hand

TerraWatch led its Essentials issue with CoStar Group buying aerial imagery provider Vexcel for $730M. CoStar gets frequently refreshed imagery over 45 countries and around half a billion properties to expand its property intelligence and AI capabilities. TerraWatch's take is that proprietary geospatial data keeps proving its value outside the traditional EO industry. GeoAI and the Law covered the darker side of that appetite. Thirty-six state attorneys general settled with General Motors for $220 million over OnStar Smart Driver, which tracked drivers' habits and location from 2016 to 2024 and sold the data through brokers to insurers who used it to set premiums. GM now needs express consent. PangeAI signed Sky Blue Impact, which will screen data center sites for grid, climate and hazard factors without consultant-led studies. Spatial Source described Sydney Trains surveying about 850 kilometres of high-voltage wiring by LiDAR twice a year to manage vegetation.

Why this matters: Real estate, insurance, data centers and rail are the verticals these feeds usually miss. They showed up because money and liability forced disclosure, not because buyers started blogging. GM is the warning that location data's commercial value and its legal exposure grow together.

3. Publishing Without a Platform

David G. Smith argues that PMTiles and the Protomaps ecosystem challenge the idea that a map service needs to be a service at all. Running your own tiles used to mean servers, databases, caches, load balancing and monitoring, so teams moved to commercial platforms and then paid for every pan and zoom once a map got popular. OPENGIS.ch released Georama 26.2, which publishes QGIS projects as WMS, OGC API Features or a full WebGIS by wiring QGIS, Django, pygeoapi, QSL and GeoGirafe under an AGPL or commercial license. Spatial Reserves ran an essay on SM Datum, a free browser toolkit with 16 tools and more than 70 operations for converting, reprojecting and repairing files. #geoObserver tried the Devenzi Map Studio WordPress plugin for OSM-based maps. Spatial Source reported that Open Spatial launched a cloud GIS platform in Australia built on open-source tech and open standards, with all data held onshore.

Why this matters: Web mapping gets almost no serious blog coverage, so a day of practical self-hosting options is welcome. It also ties to sovereignty. Open Spatial's onshore pitch and AGI's GeoCom 26 Digital Sovereignty Conference treat control of the stack as policy, not just a hosting bill.


Top Five Posts

1. Connecting Cadastral Data and Building Regulations: Lessons from an AI Prototype in Stuttgart — Open Geospatial Consortium A first-person account of building an assistant that reads German planning law, development plans and ALKIS cadastre records to answer plot-level questions. It shows how a ratio like the GRZ is defined in federal law but set in a small table on a local plan, and it treats the disagreements between cadastre and AI as the main findings. → Connecting Cadastral Data and Building Regulations: Lessons from an AI Prototype in Stuttgart

2. Cloud-Native Map Tiles With No Drama — Thoughts & Insights — David G. Smith Smith lays out the economics that pushed teams onto commercial map platforms and explains how PMTiles changes them by dropping the tile server altogether. It's a practical read for anyone weighing what a successful public map will cost. → Cloud-Native Map Tiles With No Drama

3. GeoAI and the Law Newsletter — GeoAI and the Law Newsletter Kevin Pomfret's issue leads with the $220 million GM settlement and spells out the remedies, from express consent to board-level privacy oversight. It's one of the few places in these feeds where location privacy is handled as a legal fact rather than an abstract worry. → GeoAI and the Law Newsletter

4. Earth Observation Essentials: November 16, 2026 — TerraWatch Space Newsletter The biweekly digest pairs EO market news with short analysis, and this edition opens with CoStar's $730M Vexcel deal. Its take on proprietary imagery as a property intelligence asset is a rare demand-side read on who's actually paying for geospatial data. → Earth Observation Essentials: November 16, 2026

5. Data-Driven Pluvial Flood Modeling For Urban Regions - An Investigation Of Spatial Generalizability — EarthStuff EarthStuff surfaces a paper on machine learning prediction of urban flood depth and velocity, trained on hydrodynamic simulations for Montreal. The framework kept strong performance when moved to Ottawa without retraining, which is the transfer test most GeoAI work skips. → Data-Driven Pluvial Flood Modeling For Urban Regions - An Investigation Of Spatial Generalizability

Powered by Neptune

In memory of Planet Geospatial: 2005-2014

Trans rights are human rights.