Industries / Construction and property / Engineering and architecture / Surveying
Boundaries
Boundary work fixes where one owner's land ends and the next begins: describing a parcel in words and numbers, marking its corners on the ground, and later finding those marks again.
5
Milestones achieved, from 1767 to 1910.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1767
Astronomers end a border feud
The Penn family's Pennsylvania and the Calvert family's Maryland were to meet at the 40th parallel, which lay about five miles north of Philadelphia, and their quarrel over the line dragged on for 82 years, by the account of the Maryland Center for History and Culture. In 1763 the two families hired Charles Mason and Jeremiah Dixon, both of whom had worked at the Royal Observatory at Greenwich. Over 58 months they fixed the lines by observing stars and measuring with chains and leveled wooden rods, allowing for the curve of the Earth, and stones engraved P on the north side and M on the south were set every mile. The American Society of Civil Engineers (ASCE), which lists the line as a historic landmark for 1763 to 1767, says the method became the basis for other boundaries, including the US–Canada border.
LessonA claim never surveyed on the ground can be lost: a 1685 ruling backed Calvert's line, he never had it run, and the border was set 19 miles (31 km) further south.
Credit
Charles Mason and Jeremiah Dixon; the Penn and Calvert families
1785
Land surveyed before sale
On 20 May 1785 the US Congress passed the Land Ordinance, which ordered the western public lands divided into townships 6 miles square, with lines running due north and south and others crossing them at right angles, before any land was sold. Its supporters argued that surveying before sale would prevent overlapping claims and give every owner a common line with a neighbor, by the account of the Bureau of Land Management (BLM). Thomas Hutchins, Geographer of the United States, began at a post on the north bank of the Ohio River on 30 September 1785; by 8 October he, eight other surveyors and about 30 chainmen and axemen had run 4 miles (6.4 km) of line by compass and chain. The BLM says almost 1.5 billion acres (about 6 million km²) have since been surveyed into townships and sections, "the greatest land surveying project ever undertaken."
Credit
the Congress of the Confederation; Thomas Hutchins
1878
Old marks outrank new measurements
Thomas M. Cooley of the Michigan Supreme Court wrote in Diehl v. Zanger (1878), and in his essay "The Judicial Functions of Surveyors", that however wrong the original government survey was, "the monuments that were set must nevertheless govern," even if that made one half-quarter section 90 acres and the next 70, because buyers bought by those marks. When the marks are gone, he wrote, the surveyor must still work out where the original lines were, not set new marks where the old ones should have been. A search of CourtListener, a free database of US court opinions, returned 76 published opinions that mention Diehl v. Zanger.
LessonA line that owners relied on is the boundary, even when the measurement behind it was wrong.
Credit
Thomas M. Cooley
1891
Surveyors need a license
California passed the first law to license land surveyors in 1891, and Wyoming the first to license engineers in 1907, by the account of NCEES, the US national council of the state boards that license engineers and surveyors. A license means a state board has checked a person's education, experience and exam results before that person may practice.
Credit
the California Legislature; NCEES
1910
Contract surveys end in fraud
From 1797 US public land was surveyed by deputy surveyors under contract, paid by the mile. A law of 3 March 1879 let the certificates that settlers received for paying toward a survey be passed to others and used to buy any public land. John A. Benson, a deputy surveyor in California, then ran a "syndicate" that filed survey applications in the names of settlers who were nearly all fictitious, let contracts to deputies of whom some were not surveyors at all, and often made no survey in the field, by the BLM's history. In 1910 the General Land Office ended the contract system and hired salaried "U.S. Surveyors" instead; the BLM's historian writes that this direct system cost less per mile of line and produced better work.
LessonPaying by the mile for lines that nobody checks on the ground buys lines that were never run.
Credit
the General Land Office; Frank M. Johnson, its first Supervisor of Surveys
Watch for
AI agents plot legal descriptions
Orbital, a company that sells AI software for real estate lawyers, says its system splits a US legal description into its parts, the Public Land Survey System references and the metes-and-bounds courses (compass bearings and distances walked from a starting point), has specialized AI agents turn each part into coordinates with geometry code, and plots the parcel on a map, including curves and exceptions. It gives no accuracy figures.
Achieved whenOrbital or an outside party publishes, for a set of parcels whose true outlines are known, the share plotted within a stated distance of the truth, and someone outside the company repeats the test.
Claimed by
Orbital (Andrei Vasilcoi, tech lead), January 2026
Open
01
Corner records against the map
A reader that takes the coordinates surveyors measured for corners of the Public Land Survey System (PLSS) from Utah's monument record sheets, the forms filed when a surveyor finds or resets a corner, and compares each with the position the Geographic Coordinate Data Base (GCDB) gives for the same corner. The GCDB is the GIS record of PLSS corners that maps and parcel layers are drawn on. The output is a county-by-county map of the gaps and a list of corners where the two disagree by more than the GCDB's own reliability figure.
The briefCredit · unclaimed02
What decides a boundary case
A public table of published US court decisions on boundary disputes that records which evidence decided each case (original monuments, long possession or agreement between neighbors, the distances written in the deed, a later survey) and whether the court accepted or rejected a surveyor's line. An AI model reads and codes the opinions, and people code a sample to check it.
The briefCredit · unclaimed
Open 01 · the brief
Corner records against the map
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A reader that takes the coordinates surveyors measured for corners of the Public Land Survey System (PLSS) from Utah's monument record sheets, the forms filed when a surveyor finds or resets a corner, and compares each with the position the Geographic Coordinate Data Base (GCDB) gives for the same corner. The GCDB is the GIS record of PLSS corners that maps and parcel layers are drawn on. The output is a county-by-county map of the gaps and a list of corners where the two disagree by more than the GCDB's own reliability figure.
Why it is still open
Utah's GCDB points come mainly from BLM survey records, with state, county and other coordinates added as they arrive, and each corner carries a summary of its reliability. Utah's Geospatial Resource Center has served monument record sheets through its PLSS web app since 2013 and now lets surveyors upload their own. A search found no published comparison of the sheets' coordinates with the GCDB.
Built with what exists
The Utah PLSS Point GCDB feature service; monument record sheets from the PLSS web app; a vision-language model, an AI model that reads page images, to pull coordinates, datum and units from the sheets; pyproj, a free library, to convert between datums and units; a person checking a sample of readings.
First step
Download the sheets for one township, type the coordinates of 20 corners by hand, and compare them with the GCDB points.
Done looks like
At least 1,000 corners in two or more counties, each with its sheet coordinate, its GCDB coordinate and the distance between them in feet, the share that fall outside the GCDB's stated reliability, and a hand check showing at least 98 of 100 sampled sheet readings correct.
Open 02 · the brief
What decides a boundary case
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table of published US court decisions on boundary disputes that records which evidence decided each case (original monuments, long possession or agreement between neighbors, the distances written in the deed, a later survey) and whether the court accepted or rejected a surveyor's line. An AI model reads and codes the opinions, and people code a sample to check it.
Why it is still open
Surveyors are taught an order of evidence, such as Cooley's rule above, but no coded count of how courts actually rule was found. The one online collection found, which says it holds 245 cases from 52 jurisdictions, shows "landmark" cases such as "Brown v. Smith (2023)" cited as 789 F.3d 123, a citation that CourtListener resolves to an unrelated case, United States v. Parse. The method has been used in a neighboring field: a London School of Economics team hand-coded 600 federal appeals decisions on public-land disputes, trained an AI model on them, and built a set of nearly 5,000 cases.
Built with what exists
CourtListener, the Free Law Project's free database of court opinions; a current AI model to read and code opinions; a coding sheet drawn up with a licensed surveyor and a lawyer; a spreadsheet of hand-coded cases.
First step
Code by hand the published opinions that cite Diehl v. Zanger, as the first answer key.
Done looks like
1,000 decisions from at least 10 states coded and published with citations and links, with two people coding 100 of them independently and the model agreeing with them on at least 90% of fields.