Skip to content

Tools

Find land like this: Ai2 opens its satellite AI explorer

3-line summary

  1. Ai2 released OlmoEarth v1.3 and a no-login map explorer on October 22.
  2. Precomputed annual satellite features support similarity searches and change comparisons at 10-meter resolution.
  3. At launch, 2023–2025 cover the global target extent; 2017–2022 cover selected countries.
PracticalOffice workersDevelopers
Two similar regions are outlined on a grid map, with a magnifying glass over one region on a white card against a green background
Illustration: AI Ping

Where else is there land like this field or forest? On October 22, Ai2 released OlmoEarth v1.3, precomputed satellite features and a browser-based Embeddings Explorer. Ai2 says the explorer requires no account or user-side GPU inference for basic analysis.

An embedding condenses observations from a year of satellite imagery into a numerical representation. The technical report describes annual features at 10-meter resolution, available as 128 values or a smaller 64-value representation. Computing these features once lets later analyses reuse them.

Similarity search starts with an area and a reference point inside it. For a classification map, users can label sample points, train a small classifier and export a GeoTIFF. The explorer also supports comparisons between years and checks against the underlying Sentinel-2 imagery.

At launch, 2023–2025 data cover the global target extent, while 2017–2022 cover selected countries. Antarctica is excluded. These are annual summaries rather than live monitoring; a changed feature vector alone does not establish why a landscape changed.

Why it matters

AI Ping sees a useful first step for narrowing a location and research question before running a satellite model. Check output maps against source imagery and separate validation samples before using them for decisions.

Who it's for

Office workers
For environmental or land research, check year coverage first and compare candidate changes with source imagery.
Developers
Start a pilot with precomputed features, and ensure missing pixels are not mistaken for change signals.

Sources

  • How-to guideAbout 18 min reading

    Evaluate AI work beyond first impressions

    Build a small evaluation set, compare drafts against evidence, change one instruction and test on a fresh case. No code required.

    AdvancedPlanningDesignersMarketersSalesProduct researchOffice workersDevelopers
  • How-to guideAbout 17 min reading

    Design an AI workflow with clear review points

    Map intake, drafting, review and approval. Walk through missing information and retries on paper, with explicit stop and handoff conditions.

    AdvancedPlanningDesignersMarketersSalesProduct researchOffice workersDevelopers