BRENNEN

Industries / Food, farming and forestry / Agriculture and farming / Nursery and horticulture

Starting new plants

Propagation: making new plants from seed, cuttings, grafts and lab cultures, and growing them into the young plants that growers and gardeners buy.

6

Milestones achieved, from 1920s to 2020.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1920s

    Vegetables get grafted roots

    Grafting joins the shoot of one plant to the roots of another. Masayuki Oda of Japan's National Research Institute of Vegetables, Ornamental Plants and Tea writes that the Chinese grafted trees by at least 1000 BCE, and that growers in Japan and Korea began grafting fruiting vegetables onto disease-resistant roots in the late 1920s. By Oda's count, grafted plants covered 59% of Japan's watermelon, cucumber, eggplant, tomato and melon area in 1990, and 93% of its watermelon area.

    jircas.go.jp

    Credit

    vegetable growers in Japan and Korea; figures from Masayuki Oda (JARQ, 1995)

  2. 1935

    A hormone roots cuttings

    A cutting is a piece of stem taken from a parent plant and made to grow roots of its own. In 1935 Zimmerman and Wilcoxon of the Boyce Thompson Institute reported that two synthetic hormones, indole-3-butyric acid (IBA) and naphthaleneacetic acid (NAA), rooted cuttings better than indole-3-acetic acid, a growth hormone plants make themselves. The institute patented dipping cuttings in IBA (US patent 2,129,599, 1938), and a 2013 review by E.K. Blythe says IBA and NAA are still the most widely used hormones for rooting stem cuttings.

    ishs.orgpatents.google.com

    Credit

    P.W. Zimmerman, Wilcoxon and A.E. Hitchcock, Boyce Thompson Institute

  3. 1962

    Plants cloned in glass

    Micropropagation grows many identical plants from tiny pieces of tissue on a sterile jelly of sugar and nutrients. In 1960 G.M. Morel grew Cymbidium orchids from shoot tips in a paper titled "Producing virus-free cymbidiums", and in 1962 T. Murashige and F. Skoog published the nutrient recipe now called MS medium. A 2022 review calls MS "the most used nutrient supplemented medium in the world for plant tissue culture", and Crossref counts 49,785 works that cite the 1962 paper.

    pmc.ncbi.nlm.nih.govapi.crossref.org

    Credit

    G.M. Morel; T. Murashige and F. Skoog

  4. 1986

    Cloned palms bear no fruit

    Malaysian growers cloned high-yielding oil palms by tissue culture, but Tan Yap Pau of United Plantations noticed that many clones grew "mantled" flowers, which often abort their fruit and give very low oil yields; R.H.V. Corley documented the problem in 1986. A 2015 Nature paper says mantling "largely halted efforts to clone elite hybrids for oil production" until its authors traced it to lost DNA methylation (a chemical tag that keeps a gene switched off) at a site they named Karma, which can now be tested at the plantlet stage.

    LessonA clone is not always a true copy: tissue culture can change how genes behave, and the fault may only show when the plant flowers.

    pmc.ncbi.nlm.nih.gov

    Credit

    Tan Yap Pau, United Plantations; R.H.V. Corley; M. Ong-Abdullah and colleagues (Nature, 2015)

  5. 1992

    Machines graft seedlings

    Grafting by hand takes time, space and skilled workers, and Oda reports that Japan's aging farmers and labor shortage made it hard to keep up. Kobayashi and colleagues began building a grafting instrument for melons, cucumbers and their relatives in 1987, and Onoda and colleagues completed it in 1992; Oda reports that "various grafting methods and instruments were released".

    jircas.go.jp

    Credit

    Kobayashi and colleagues; Onoda and colleagues, Japan

  6. 2020

    Robots stick cuttings

    Sticking is pushing a fresh cutting into a tray of rooting soil. Greenhouse Grower reported in December 2020 that US young-plant growers, among them Spring Meadow Nursery (four machines) and Pleasant View Gardens, were sticking cuttings with robots from ISO Group and Visser, and that, depending on the container, the ISO machines "can output 35% to 50% more per person than manually sticking".

    LessonThe robot changed the work upstream: Spring Meadow had to retrain the people who take cuttings to make them smaller and more uniform.

    greenhousegrower.com

    Credit

    ISO Group; Visser (AutoStix); Spring Meadow Nursery; Pleasant View Gardens

Watch for

  1. Cameras flag contaminated cultures

    Bacteria or fungi in a tissue-culture jar can ruin the plants in it and spread through a lab. S. Hettiarachchi and colleagues built a camera rig that photographs culture jars, and reports that one machine-learning method sorted contaminated from clean rice cultures with 99.6% accuracy on its own images.

    Achieved whenan outside lab tests the method on its routine cultures of horticultural plants, with contamination confirmed by culturing, and publishes how many contaminated jars it missed.

    dergipark.org.tr

    Claimed by

    Hettiarachchi, Sankalpana and Silva, Yüzüncü Yıl University Journal of Agricultural Sciences, April 2026

Open

  1. 01

    Robot takes cuttings

    A robot arm with a small cutter that takes tip cuttings of a set length from potted stock plants (the parent plants kept only to supply cuttings) and lays them in a tray, ready to stick.

    The brief
    Credit · unclaimed
  2. 02

    Predict rooting from a photo

    A photo station that pictures every cutting as it is stuck, a record of which cuttings have rooted three weeks later, and a model that predicts rooting from the first photo, so weak cuttings can be sorted out before they take up bench space.

    The brief
    Credit · unclaimed

Open 01 · the brief

Robot takes cuttings

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A robot arm with a small cutter that takes tip cuttings of a set length from potted stock plants (the parent plants kept only to supply cuttings) and lays them in a tray, ready to stick.

Why it is still open

Robots now stick cuttings, but people still take them. Spring Meadow Nursery says "We take all cuttings on site" and retrained staff to suit its sticking robots. The nearest new project, TTA-ISO's chrysanthemum harvester (EU-supported, April 2025 to March 2027), cuts flower stems for sale, not cuttings. Web, Crossref and GitHub searches found no robot that takes cuttings from stock plants.

greenhousegrower.comfloraldaily.com

Built with what exists

An SO-101 arm with LeRobot (see the Note "No download"); a control model trained from demonstrations with NVIDIA GR00T 2 (see the Note "Five demonstrations per task") or Runway Praxis-1 (see the Note "Open robot model weights"); a small servo-driven snip as the gripper; a wrist camera; potted coleus, pelargonium or chrysanthemum stock plants.

  1. First step

    Buy ten coleus plants, fit a snip to an SO-101 gripper, and record 50 remote-controlled cuts.

  2. Done looks like

    20 filmed stock plants, with a published tally of cuttings taken within 1 cm of the target length and with the right number of leaf nodes, stems crushed or torn, and seconds per cutting against a person. Then robot-taken and hand-taken cuttings are rooted side by side, and the share rooted after 21 days is reported for each.

Open 02 · the brief

Predict rooting from a photo

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A photo station that pictures every cutting as it is stuck, a record of which cuttings have rooted three weeks later, and a model that predicts rooting from the first photo, so weak cuttings can be sorted out before they take up bench space.

Why it is still open

A 2018 trial by the Dutch vision company Beltech at a young-plant grower, reported in Greenhouse Grower, set out to predict cutting quality from multispectral photos but published no results. A 2025 Universidad de Antioquia thesis grades chrysanthemum cuttings by leaf shape to help select "ideal cuttings for rooting", but its abstract reports no rooting outcomes. No public set of cutting photos paired with rooting results was found.

greenhousegrower.comkujane.minciencias.gov.co

Built with what exists

A phone on a fixed stand with a color card; a mist bench or a humidity dome; IBA rooting powder; Label Studio for labels; an open image classifier, such as the classification models in Ultralytics YOLO.

  1. First step

    Stick 200 cuttings of one easy species this month, photographing each against the color card before it goes in.

  2. Done looks like

    2,000 cuttings of one species, each photographed at sticking and scored rooted or not at 21 days, with the model scored on a batch stuck in a different week from the training batches (area under the curve and the share of failed cuttings it catches), and all photos and outcomes released.