HiveMind AI

What to record after every inspection, and what you can safely skip

By HiveMind AI ·

You are standing at an open hive. The smoker is dying, a bee is walking across your veil, and your gloves are sticky to the second knuckle. Nobody writes a paragraph here.

So you close up and tell yourself you will remember. Three weeks later you are back at the same stand asking: was this the one with no eggs, or was that hive four? Did I treat this one, or the split? Was the brood spotty last time, or am I thinking of its neighbour?

Records exist for that second moment, not the first. That makes the job small: hold the handful of facts that will change what you do next time, in a form you can read in ten seconds with a frame in your other hand. Everything else is decoration, and decoration has a cost.

This is written for a paper notebook as much as for a phone. The list is the same either way.

The observations that change a decision

The test for whether something belongs in the record: name the decision it feeds. If you cannot, skip it.

Queen status. Not "saw her" or "didn't", but one status from a short list: queen right, queen seen, not seen, queenless, new queen, virgin queen, drone layer, laying workers. Eggs are the evidence, and better evidence than a sighting. A worker egg hatches after about three days, so eggs in the cells mean a queen was laying within the last three days whether or not you found her (Mortensen, Schmehl and Ellis, UF/IFAS). This field drives requeen, combine or wait, and it is the one you will most want to look back on. Laying workers usually mean a colony has been queenless for one or more weeks (MAAREC), and the record is how you know when that week started.

Worth a category, not a note.

Brood pattern. A four-point scale is enough: none, spotty, half and half, solid. One reading tells you little. When Lee and colleagues swapped queens between good-brood and poor-brood colonies, the "poor" queens' patterns improved after about three weeks in a good colony, which points at the colony's conditions, not the queen (Lee et al. 2019). What the record gives you is the second reading. Spotty twice in a row on one hive while its neighbours are solid is a reason to look closer: disease, stores, a failing queen. Spotty once is a reason to look again.

Worth a scale, not prose.

Frames of bees, brood, honey and empty comb. Four counts, made as you go, per box if there is more than one. Brood tracks the queen's output and the colony's growth. Honey tells you whether to feed and, in autumn, whether the colony can winter. Bees is your population estimate. Empty comb, read with the other three, answers the space question: nine frames covered, brood on six, needs a box; four frames of bees in a two-box stack needs the opposite.

When you do not want to pull frames, a 1 to 5 heft score, lifting the back of the hive, is a coarse stand-in for stores and population together. Crude, but comparable from visit to visit, which is what matters. Cornell's overwintering guide, written for Upstate New York, puts the minimum for a full-size colony at 80 pounds of honey, roughly one full deep box, and adds that the right amount "will depend on the weather and bees". Your local figure belongs on the inside cover of the notebook.

Worth numbers. Numbers are the only observations you can lay side by side.

Temperament. Score it rather than describe it. "A bit cranky" drifts with your mood and how long you have been at it. Four short scales, each 1 to 4: bees flying at you, running on the comb, hanging in clumps, attempting to sting. Add them up. The decision this feeds is requeening a hot colony, and that decision needs more than one bad day of evidence, because defensiveness swings with weather and forage, and a colony that has lost its queen is often agitated for reasons that have nothing to do with its genetics. A number lets you tell a pattern from a Tuesday.

Mite load. Of everything here, this is the number with the most riding on it, and the one hobbyists most often leave as a feeling. Record it as a measurement: the count, the method and the sample size, one line per colony. A half cup of bees from a brood frame is about 300 bees (University of Minnesota Bee Lab), so 9 mites in a wash is 3 per 100. Two colonies on the same stand can sit either side of the treatment line, so never average a yard.

Write the unit every time. A wash, sugar roll or CO2 check gives mites per 100 bees. A sticky board gives mites per day falling from a colony of unknown size, and Penn State's López-Uribe lab says plainly that "there is not a reliable threshold" for a board count. Boards show a trend on one hive. They do not tell you to treat. Keep the two on separate lines so nobody, including you, ever compares them.

Where the treat-now line sits is genuinely disputed. The Minnesota Bee Lab treats above 2 per 100 and aims below 1 year-round; Penn State Extension asks for levels below or around 2, monitored monthly; Mississippi State uses 3 from May to October; Penn State's López-Uribe lab puts it near 5. Note that those last two are one university: two labs, 2 and about 5. Neither is a misprint, and the gap between them is the best evidence available that this figure is a judgement call rather than a settled fact. It is also why the field you write down is the count, the method and the sample size, never somebody's verdict: the number keeps its meaning when you change your mind about the line. Pick your figure, write it beside your winter-stores figure, and count monthly, the cadence Minnesota and both Penn State sources suggest. A count older than a month tells you a recheck is due, not what the load is today.

What you did, and what to do next. Treatment applied, with the product and the date, because dose intervals and withdrawal periods run from that date. Feed given, with the ratio. Diseases seen, or an explicit "none seen", which is not the same as leaving the line blank. And one sentence of next action: "add a super", "check for eggs in a week", "wash this one again before the 15th". That sentence is the most valuable free text in the record, because it is the one you will actually read.

One cheap extra: a single word for the entrance (quiet, normal, very busy, robbing suspected). It matters only when it is abnormal, and then it decides something on the spot: an entrance reducer, an end to open feeding, or a look inside a hive that has gone quiet while its neighbours fly.

What you can safely skip

Every keeper has fields they fill in faithfully and never read again. Some candidates:

  • The time of day. The date is enough.
  • Weather in prose. A one-word sky and a temperature are worth keeping if your phone already knows them, because they are the only context that explains a temperament score. If you would have to guess the temperature, leave it blank.
  • A hunt for the queen. If you saw eggs, "not seen" is a complete answer. Ten minutes of searching to write "seen" is ten minutes of open hive.
  • Frame-by-frame narration. "Frame 1 honey, frame 2 honey, frame 3 brood..." collapses into four counts with nothing lost.
  • Drone counts, pollen colours, honey colour, comb colour. Pleasant to notice. None of them feeds a decision you will make from the notebook.
  • Your smoker fuel, your gear, how many stings. That is a diary of you, not a record of the colony.
  • "Looked good", "strong", "healthy". Conclusions, not observations. Next season you will not know what they rested on.
  • The setup, every visit. Note the box configuration once and record changes.

The case against recording everything is not that it takes too long, though it does. It is that the important line hides. A page of notes gets skimmed, and skimming your own notes is how a queenless hive goes unnoticed for a second visit. Records built from a handful of comparable fields can be laid side by side; records built from paragraphs cannot. And every extra minute per hive is a minute that, on a hot afternoon with twelve hives, gets saved by not recording the last four at all. The record that gets kept beats the record that was thorough.

What a record needs to still make sense in five years

Dates with a year, in one format. 2026-09-01 or 1 Sep 2026. "9/1" reads two ways depending on where you learned to write, and a loose page with no year is a page with no date.

A hive identity that survives a dead-out. The number is painted on a box; the colony lives in it, and colonies die. Decide which one you are tracking. When a colony dies, write the date and your best guess at the cause, and retire that line. The colony that moves into the same box next spring is new, with a new start date, even if the box still says 3. Otherwise you will read a dead colony's mite history as the new one's. Note splits (which hive it came from) and combines (which one survived). Lineage is what makes a two-year trend mean anything.

Units, written every time. Mites per 100 bees or per day, and the sample size. Deep or medium, because six frames of honey is not one quantity. Pounds or kilograms. Fahrenheit or Celsius. Which box is box 1 (usually the bottom). Syrup as 1:1 or 2:1. You know all of this today; the page has to know it in 2031.

Who looked. Initials, if more than one person inspects. Two keepers rate temperament differently, and knowing whose 3 it was is half of reading it.

Photos and voice notes: when each beats typing

A photo beats typing for anything you might want a second opinion on. A frame of brood held to the light settles a "was it really spotty?" argument months later and lets a mentor weigh in from their kitchen. Odd cells, a larva you cannot name, a beetle, a queen cell and where it sits on the frame, the label on the treatment you used. One picture of the stack replaces a diagram. The catch: a photo without a hive number and a date is worthless in five years, and phone metadata supplies the date but never the hive.

A voice note beats typing at the exact moment typing is impossible: gloves on, frame in hand, smoke drifting. Talk through the list above in order: "Hive four. Eggs seen. Brood solid, six frames. Three of honey. Calm. Add a super." Then turn it into fields the same day, because a pile of untranscribed memos is an archive nobody will search. A photo of the paper page preserves the evidence the same way, and still has to become fields.

Pen or keyboard beats both for the categorical fields and the next action, where you want the answer chosen from the same short list every time.

That short list is the shape HiveMind AI's inspection form takes: picker fields and frame counts, a next-action line that becomes a task, up to five photos on each inspection, and a voice note that fills the fields in for you to check before saving. The mite check is its own entry per colony, with the method and sample size stored beside the count, and sticky-board counts kept apart from the per-100-bees line on purpose. A paper notebook ruled the same way gets you most of the way there. Whichever you use, keep the list short enough that you fill it in at every hive, on the bad days too. That is the whole trick.

Filling that form in HiveMind: Logging an inspection in the help centre walks through the fields, the follow-up task and voice fill.

Sources

  • Mortensen, A.N., Schmehl, D.R. and Ellis, J. European Honey Bee, Apis mellifera Linnaeus, and subspecies. UF/IFAS Extension EENY568. https://ask.ifas.ufl.edu/publication/IN1005
  • Mid-Atlantic Apiculture Research and Extension Consortium (MAAREC). The Colony and Its Organization. https://canr.udel.edu/maarec/honey-bee-biology/the-colony-and-its-organization/
  • Lee, K.V., Goblirsch, M., McDermott, E., Tarpy, D.R. and Spivak, M. (2019). Is the Brood Pattern within a Honey Bee Colony a Reliable Indicator of Queen Quality? Insects 10(1), 12. https://doi.org/10.3390/insects10010012
  • University of Minnesota Bee Lab. Varroa Mite Testing & Management. https://beelab.umn.edu/varroa-mite-testing
  • López-Uribe Lab, Penn State University. Varroa Mite Monitoring. https://lopezuribelab.com/varroa-mite-monitoring/
  • Penn State Extension. Methods to Control Varroa Mites: An Integrated Pest Management Approach. https://extension.psu.edu/methods-to-control-varroa-mites-an-integrated-pest-management-approach
  • Mississippi State University Extension. Sampling for Varroa Mites Using an Alcohol Wash. https://extension.msstate.edu/publications/sampling-for-varroa-mites-using-alcohol-wash
  • Cornell CALS Pollinator Network. Overwintering. https://cals.cornell.edu/pollinator-network/beekeeping/overwintering