Macroscope Nexus: the Observatory →
PublicThe public Macroscope. Enter any coordinate on Earth and discover its ecological identity, read from terrain, climate, ecoregion and the living world.
macroscope.nexus · yea.earth, Your Ecological Address
The instrument, running
All projects →Macroscope Reader →
In developmentThe fourth generation, built in the open and changing daily: about 1,170 curated places, from biological field stations and UC Natural Reserves to public lands, the laboratory at Canemah and the Owl Farm testbed, each read across four faces.
Science with Claude →
PublicStructured human-AI scientific investigations, and the essays in which the ideas behind the Macroscope are worked out in public, one morning at a time.
The Macroscope, measured 25 September 2026
1,170curated places
5,151virtual instruments reading climate and community records at those places
33laboratory instruments: 24 fixed, 9 carried into the field
123photopoints at 33 Virtual Field stations
1,171panoramas at 42 places
530three-dimensional reconstructions
1941–2026aerial record at Owl Farm
100+recorded design rulings
The Macroscope
Read the introduction →The idea
A microscope makes the very small visible and a telescope the very far. A macroscope is for what is too large, too slow and too interconnected to see at once: a watershed, a season, an ecosystem, a life of paying attention to one piece of ground.
The grammar
The world arrives as coordinates. A place is a coordinate someone has decided matters and has curated. A survey is a bounded act of looking; the pulse is what keeps arriving between surveys. Places run from deeply known, walked and ground-truthed, to lightly sketched from public data, and the two are always told apart.
Four faces
Stand at any coordinate and it turns four faces toward you at once:
- EARTH climate, geology and water
- LIFE biodiversity and season
- HOME the built habitat
- SELF the observer, whose days are part of the record
Claude in the laboratory
How the collaboration works →Since 2025 the Macroscope has been designed and built by a field ecologist and Claude, Anthropic's AI model, in near-daily sessions. The human keeps the judgment, the machine does the engineering, and every design ruling is written down.
Engineering collaborator
Claude writes and reviews the database, the data collectors and the reader, and tests each change against a local copy before anything reaches the production substrate.
Narrator inside the instrument
The STRATA context engine calls Claude to say what the instruments and reference layers show, and only what is measured: narration is arithmetic over state, never speculation.
Co-author of the record
More than a hundred design rulings, the seven-phase investigations and the essays behind them are published as they happen at Science with Claude.
The boundary is deliberate. Claude does not identify species or classify pixels; specialised models do: Pl@ntNet now and BioCLIP 2 next for species and, in the proposed work, OlmoEarth for the satellite record. Claude designs, runs and interprets them, and the naturalist makes the call.
Principles we build by
Essays at Science with Claude →- Intelligence that knows what it costs
- Simple routines on a small local server do most of the watching; a modest local model gathers the facts when something unusual appears; only the hardest questions go to the most capable and most expensive model. Most mornings the system says nothing at all.
- Trust as architecture, not promise
- Hardware we control is trusted, services we can verify are monitored, and neural-network outputs are quarantined as probabilistic claims. Privacy grows out of ownership: the weather at Canemah is public because the place is, while a person's health belongs to that person.
- Known, not owned
- Every curated place is to have a curator, a role being developed for early 2027. Observations stay with the platforms and people who made them, and a rights envelope of source, license, method and uncertainty travels with every record.
- The naturalist inside the instrument
- The machine finds the irregular heartbeat; the biologist makes the diagnosis. The more capable the instrument becomes, the more it depends on what only a body in a place can observe.
Research program
Dreaming in NEON →The next loop to close runs from the panorama to the satellite: a place that can map and monitor its own plant communities and seasons, in its own vocabulary, with provenance intact.
- Under wayThe Panorama room. Every panorama in the catalog gains a three-dimensional reconstruction, rendered as part of the import pipeline and shown in one room as its sphere or its splat.
- ProposedSatellite portraits and change. Ai2's OlmoEarth run over every curated place: embeddings that find places like this one across the network, and quarterly fingerprints that flag change.
- ProposedA forty-year phenology baseline. Landsat since 1984 and Sentinel-2 since 2015, read per place and per plant community: onset, peak and senescence, and a baseline against which each new year is read.
- ProposedPlant communities in a station's own vocabulary. Paired panoramas over the same photopoint, one on the ground and one from a drone above the canopy, identified with BioCLIP 2, crosswalked to USNVC and CNPS alliances, and used to train a 10-meter community map repainted on every satellite pass.
- ProposedWeather at the UC Reserves. The UC Natural Reserve System's Dendra weather stations, brought in for the reserves already in the catalog.
- ProposedThe Field Audit. A controlled test of whether AI identification becomes safer and more useful when it is grounded in place, season and a naturalist's review.
- ProposedA rights and provenance protocol, written with the source organizations at the table, for any field station, land trust or school to adopt.
- ProposedCalibrating the crowd. Correction factors that tie community-science records to the calibrated measurements at NEON sites in the Pacific Northwest.
- ProposedPartner stations. The reader running at UC Natural Reserves and Virtual Field stations.
- ProposedOregon Regional Ecological Observatories. A confederation of natural-area stewards, each place with its own monitoring agent and a person who goes outside and looks.
Inquiries about the program and the platform are welcome. Of peers we ask what the instrument asks of us: argue with it. Get in touch →
Network and collaborators
- Virtual Field
- The NSF-funded Research Coordination Network of field stations sharing 360-degree panoramas for teaching and research. Michael Hamilton is a collaborator in the network, and the Macroscope holds 455 panoramas from 123 photopoints at 33 of its stations.
- University of California Natural Reserve System
- Where Michael Hamilton directed the James San Jacinto Mountains and Blue Oak Ranch reserves for 36 years. Its reserves are among the curated places, and its Dendra weather stations are proposed for the reserves in the catalog.
- Organization of Biological Field Stations
- Its member field stations are among the curated places.
- Open data and models
- iNaturalist, eBird, GBIF and BirdWeather for the living record; ERA5 for climate history; SSURGO soils, NLCD land cover, 3DEP terrain and EPA ecoregions as reference layers; Pl@ntNet for plant identification; and Ai2's Semantic Scholar behind the knowledge context of STRATA. Every source travels with its license and attribution.
Technical reports
Browse archive →The laboratory's numbered series of technical notes, specifications and white papers.
Wearable Sensor-Fusion Survey Rig
This note records the conceptual architecture for a wearable sensor-fusion rig intended to support ecoSPLAT surveys at Canemah Nature Laboratory...
July 23, 2026The Macroscope Data Logger Designer
The original Climate Analyst Data Logger Designer (CNL-TN-2026-052 v0.1) specified a Climate-Analyst-internal wizard that provisioned virtual weather...
April 16, 2026Macroscope Visual Language Specification
This specification establishes a portable visual language for Macroscope Next Generation (MNG) interfaces, derived from analysis of production CSS on...
April 15, 2026The Collaboratory Librarian
The Macroscope Collaboratory Librarian was instantiated in early April 2026 as a catalog system for Dr. Hamilton's personal library, initially...
April 13, 2026Macroscope Collaboratory Librarian: Unified Catalog and Document Ingestion Pipeline Specification
The Macroscope Collaboratory Librarian is the fourth panel of the I3 shell (Instruments, Investigations, Intelligence, Librarian) and serves as the...
April 11, 2026olmOCR-2 as the Canonical PDF Ingestion Engine
The first-light OCR evaluation on April 10, 2026 (see REPORT.md in `tests/ocr_bench/`) compared two text extraction pipelines against four PDFs:...
April 10, 2026Semantic Scholar API Integration for the Macroscope Collaboratory
An API key for the Semantic Scholar Academic Graph was obtained on April 11, 2026, providing authenticated access to Allen AI’s corpus of...
April 10, 2026The Macroscope Collaboratory
Versions 0.1 and 0.2 of this technical note documented the first-light test of the Investigation Wizard as a PHP in-request application driving the...
April 9, 2026The STRATA Substrate
STRATA 2.0's Observatory and Intelligence dashboards currently generate all sensor summaries, narrative text, and ecological context on demand --...
April 8, 2026STRATA IQ: Place-Based Ecological Intelligence
STRATA's intelligence tier currently operates without place. The temporal state endpoint assembles sensor readings from 26 platforms across four...
April 8, 2026Other experiments
View all projects →The Lookout →
LivePhotographic studies from a digital naturalist's eye
NodeWorlds →
DemoText adventure with living data
Madverse →
LiveCollaborative puzzle poetry
Wiki-Lyrical Engine →
LiveCognitive poetry from the dream buffer
Quotes Collection →
LiveCurated wisdom across domains
Quotes Explorer →
LiveVisual navigation through ideas
Michael P. Hamilton →
LiveMy Virtual Field Station