GIS.Coffee

A practical GIS field guide

GIS into the future

A modern GIS connects trusted data, processing tools, services, and apps so people can make decisions from geography.

Think coffee bar: keep the recipe open, even if ArcGIS Pro is your favorite mug.

A connected GIS, not a single product

Keep data contracts and services stable while tools plug in around them.

Inputs

Organization data

Enterprise systems, field updates, files, and public data.

Workbenches

Choose the right tool

ArcGIS Pro for hands-on GIS; GDAL/OGR or headless QGIS for repeatable processing.

Shared boundary

Data and services

ArcGIS services, PostGIS, files, and APIs give other tools a stable way in.

Experience

Apps and teams

Esri apps for standard tasks; custom web apps for workflows that need a tailored fit.

Automation and agent glueGDAL/OGR · qgis_process · ArcPy · Python · LLM-generated adapters, reviewed and tested
ArcGIS Pro remains a first-class manual workbench; repeatable pipelines can use whichever stable tools and interfaces fit.

Tools by layer

Use the parts that solve a real workflow; this is not a shopping list.

  1. 01

    Keep authoritative data clear

    Define ownership, schema, and CRS.

    Leave governed data in the databases and ArcGIS services your team trusts; use GeoPackage or GeoParquet for portable handoffs and analytics.

  2. 02

    Automate repeatable work

    Make a recipe you can rerun.

    GDAL/OGR handles format conversion and common processing; qgis_process runs QGIS models and algorithms headlessly; ArcPy fits ArcGIS workflows.

  3. 03

    Share through stable interfaces

    Separate processing from clients.

    Use ArcGIS feature services, PostGIS, OGC APIs, or tiles as clear boundaries between data pipelines and apps.

  4. 04

    Choose an app for the user

    Manual, configured, or custom.

    ArcGIS Pro stays useful for manual GIS; Experience Builder and Instant Apps cover standard needs; the ArcGIS Maps SDK, MapLibre, or Leaflet support custom web apps.

  5. 05

    Orchestrate and verify

    Focus effort on architecture.

    Agents can help write glue code and tool calls; version the recipes, test outputs, protect credentials, and review changes before release.

Agentic development for GIS

Let the agent learn the commands; you design the system.

An agent can draft GDAL commands, qgis_process models, ArcPy scripts, API clients, and web UI. That can reduce how many tool-specific details you need to memorize; your leverage is defining the architecture, data contracts, outputs, and checks that keep every tool honest.

Good at connecting tools

Drafting a GDAL conversion, wrapping qgis_process, wiring an ArcGIS REST call, or building a focused app.

Architecture stays human-led

Decide where truth lives, how services connect, who can write data, and what makes a result valid.

Test the actual output

Check a known CRS, geometry validity, feature counts, permissions, and the map a user will see.

A practical agentic loop

An agentic tool can inspect a project, edit files, and run commands; you stay responsible for the goal and the review.

  1. Frame itState the goal, boundaries, sample data, and what “done” means.
  2. Build a sliceAsk for one focused pipeline or app change, not a whole GIS rewrite.
  3. Run the checksTest real geometry, a known CRS, permissions, and the rendered map.
  4. Review and keepRead the diff, verify behavior and permissions, then decide what ships.

DHH’s writing and Omarchy reflect an opinionated, keyboard-first approach: keep the workflow direct and avoid complexity that does not earn its keep. Agents can learn the tool syntax; you keep the architecture deliberate.

Back to GIS Grounds