Industries / Food, farming and forestry / Agriculture and farming / AgTech
Precise planting and spraying
This work puts seed, fertilizer and weedkiller only where each part of a field needs them, using machines that know where they are and change their rate or switch off as they move.
4
Milestones achieved, from 1994 to 2019.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1994
Fertilizer blended to a map
In 1991 Ag-Chem Equipment of Minnesota, maker of the Terra-Gator spreader, bought its rival Lor-Al and with it 60% of Soil Teq, "a maker of site-specific application systems". Soil Teq's SOILECTION system took soil samples acre by acre, turned them into soil maps, and changed the fertilizer blend as the machine crossed the field; in 1994 Ag-Chem joined Unisys to guide it by satellite instead of by radio antennas. A 1994 Minneapolis/St. Paul CityBusiness article, quoted in a company history, called variable-rate application "the talk of the farm world" but said the expensive systems had not yet taken off.
LessonThe machines were ready years before most farms bought them.
Credit
Soil Teq; Ag-Chem Equipment
2008
A referee for tractor plugs
ISOBUS, the international standard ISO 11783, defines how a tractor, the screen in its cab and the implement behind it exchange data, so that one screen can run a sprayer or planter made by another company. The standard "leaves room for interpretation", which led makers to build their own versions that did not always work together. On 28 October 2008 seven equipment makers and two associations founded the Agricultural Industry Electronics Foundation (AEF) to write tighter guidelines, certify products and publish a database showing which machines work with which; it now has eight core makers, three associations and 300 general members, and holds "plugfests" where makers test their machines against each other.
LessonA standard on paper did not make machines from different brands work together; it took a separate body to test and certify them.
Credit
AEF and its founding members
2017
Dicamba drifts onto neighbors
USDA cleared Monsanto's soybean and cotton seed bred to tolerate the weedkiller dicamba in 2015, and in 2017 new low-volatility dicamba sprays (XtendiMax, Engenia and FeXapan) went on sale for spraying over the growing crop, which was planted on 27 million acres (about 10.9 million ha). The US Environmental Protection Agency (EPA) had written that the label rules were "expected to eliminate any offsite exposures"; by the end of the season state agencies were investigating 2,708 formal complaints, and university weed scientists estimated that about 3.6 million acres (about 1.5 million ha) of soybeans in 24 states, about 4% of US soybean area, had been damaged, by the count of Kevin Bradley of the University of Missouri. A federal appeals court vacated the registrations in June 2020, and a federal district court in Arizona vacated their successors on 6 February 2024.
LessonDicamba can turn to vapor after it lands and move through the air for several days, so label rules that depend on careful spraying alone did not keep it in the field.
Credit
Monsanto (now Bayer) and the makers of the sprays; damage counted by Kevin Bradley and state agriculture departments
2019
Shut-offs and rates go mainstream
USDA's Economic Research Service (ERS) measured, from its national farm survey, that automatic section control, which shuts off sprayer boom sections, nozzles or planter rows over ground already covered, was used in 2019 on 41% of US cotton acres, 45% of sorghum acres and 48% of barley acres. Variable-rate application covered 37.4% of corn acres in 2016 and 25.3% of soybean acres in 2018, and auto-steer guidance grew from 5.3% of corn acres in 2001 to 58% in 2016.
LessonThese tools spread on big farms first: at least half of larger row-crop farms used yield maps, soil maps, variable rate or guidance, while under 25% of the smallest farms used any of them, cotton excepted.
Credit
measured by USDA ERS (Jonathan McFadden, Eric Njuki and Terry Griffin) with the National Agricultural Statistics Service (NASS)
Watch for
Dicamba back under tighter rules
EPA registered three dicamba sprays (Engenia, Stryax and Tavium) on 6 February 2026 for use over tolerant soybeans and cotton in the 2026 and 2027 seasons only, and calls the restrictions the strictest in its history: half the yearly amount, twice the volatility-reducing additive, no spraying when the temperature will reach 95°F (35°C), and limits between 85°F and 95°F (about 29–35°C). EPA says it will watch incident reports through both seasons. Environmental groups asked the Ninth Circuit Court of Appeals to review the decision on 20 February 2026.
Achieved whenstate complaint counts for 2026 and 2027 stay far below the 2017 level (Minnesota alone had 249 reports in 2017 and 304 in 2021), and the registration survives the court challenge.
Claimed by
US EPA, February 2026
Six hundred thousand spray drones
DJI, the largest maker of farm drones, says more than 600,000 of its agricultural drones were in use worldwide by the end of 2025, flown by more than 600,000 trained operators in more than 100 countries and regions. A separate 2025 survey by the American Spray Drone Coalition, an industry group, put the yearly work of licensed US drone operators at about 16.4 million acres (about 6.6 million ha) treated, as reported by DigitalToday.
Achieved whenaviation or agriculture regulators in several countries outside China publish counts of registered spray drones and the area they treat that support DJI's figure.
Claimed by
DJI Agriculture, April 2026
Open
01
A national dicamba complaint table
One public table of off-target dicamba complaints by state and year, 2017 to 2027, for every state where dicamba-tolerant soybeans or cotton are grown. Each number links to the state's own report or its reply to a public-records request, with a note on how that state counts a complaint.
The briefCredit · unclaimed02
Answer key for spray labels
A public, hand-checked file of the spraying limits printed on 50 US herbicide labels: most product per pass and per season, allowed wind speeds, temperature limits, buffer distances, droplet size and cutoff dates. Use it to score how accurately AI models pull the same limits out of the label PDFs. Then build a small checker that takes a sprayer's logged pass and the weather at that time and flags any limit it broke.
The briefCredit · unclaimed
Open 01 · the brief
A national dicamba complaint table
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
One public table of off-target dicamba complaints by state and year, 2017 to 2027, for every state where dicamba-tolerant soybeans or cotton are grown. Each number links to the state's own report or its reply to a public-records request, with a note on how that state counts a complaint.
Why it is still open
EPA counted about 3,500 dicamba incident reports for the 2021 season, without a state-by-state table in its release, and estimated in 2020 that only about one incident in 25 is reported. It says it will watch incident reports through the 2026 and 2027 seasons, but searches found no multi-state count for any later year. Minnesota publishes its own yearly table (249 reports in 2017, 304 in 2021, none in 2025), and the national figure courts still cite for 2017 is Kevin Bradley's tally of 2,708 complaints.
Built with what exists
State agriculture department websites and public-records requests; a web-browsing AI agent to find each state's complaint pages and draft the requests (see the Note "The jobs agents repeat"; the Note "Deleting 80% to start over" covers browser agents); a spreadsheet or SQLite file; USDA NASS planted acres by state for context.
First step
Send records requests this month to the pesticide programs of the ten largest soybean-growing states, and enter Minnesota's published table as the first rows.
Done looks like
Yearly counts for 2017 to 2026 for at least 20 states, each traced to a source document, published before the 2027 spraying season and updated after it, with the 2026 total set beside the 2017 and 2021 figures.
Open 02 · the brief
Answer key for spray labels
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public, hand-checked file of the spraying limits printed on 50 US herbicide labels: most product per pass and per season, allowed wind speeds, temperature limits, buffer distances, droplet size and cutoff dates. Use it to score how accurately AI models pull the same limits out of the label PDFs. Then build a small checker that takes a sprayer's logged pass and the weather at that time and flags any limit it broke.
Why it is still open
EPA says its label review is still mostly manual, with staff reading long label files to pull out details such as the application rate, and its digital label is still being planned; the European Union's digital pesticide labels apply only from 1 January 2028. Label data is sold in structured form (the TELUS Label API covers more than 15,000 products), and a University of Maryland student team built an AI pipeline over about 5,000 EPA labels with ">90%" accuracy "based on internal experiments", but no public, hand-checked answer key for scoring such tools was found.
Built with what exists
Label PDFs from EPA's Pesticide Product and Label System; AI models that read long documents, hosted or run locally (see the Note "Seven more flagged", on an AI reviewer that caught errors people missed in long records); ISOBUS task files or shapefiles as the logged passes; hourly records from a public weather station network.
First step
Code the three 2026 dicamba labels (Engenia, Stryax and Tavium) and two glyphosate labels into a draft format, by hand, this month.
Done looks like
50 labels coded by two people working separately, with disagreements resolved and published; each AI model's share of limits extracted exactly right, field by field; and the checker run on 100 made-up logged passes with known breaches, reporting how many it caught and how many false alarms it raised.