In August and September the network grew from 740 to 1,310 members. This second report looks at whether what the first one showed held up with so many new people: which day length gets picked, how people automate and how close each grow gets to the environment it set for itself.
It covers two months because the September report did not come out. Putting them together leaves the series with no gaps.
| Data | #1 · as of Jul 31 | #2 · as of Sep 30 |
|---|---|---|
| Members | 740 | 1,310 |
| Active zones | 918 | 1,555 |
| Photoperiods configured | 275 | 440 |
| Of those, longer than 24 hours | 55% | 52% |
| Sensors | 133 | 298 |
| Zones with climate evaluated | 53 | 156 |
| Time with humidity in the critical zone | 25.0% | 19.6% |
| Harvests with the yield filled in | 1 | 3 |
Each section is calculated over the zones that filled in that particular field, and they are different subsets: 440 have a photoperiod configured, 156 have their climate evaluated, 131 described their lighting, 120 described their substrate. Against 1,555 active zones, it is always a part.
That part is not a random sample. Whoever connects sensors, describes their room and keeps a journal is, almost by definition, whoever grows more methodically. It is reasonable to expect the network as a whole to look less like these numbers than the size of each cut would suggest. Every chart states how many zones it covers, and it is worth looking.
August brought 302 sign-ups and September 268: a steady pace, above July and far from the June peak. In September 466 members logged in at least once, 36%; in August, with a smaller network, it had also been 466, 45%.
The share that measures is growing faster than the network. Of the 1,555 active zones, 460 have a device connected, 370 keep a journal and 211 have a sensor assigned: 14%, against 11% in July. The rest of this report is about that instrumented subset, which is the one that leaves data behind.
The map is recalculated every day, so it does not match the report window: it shows where the network is today. We group by city and a dot only appears once there are at least five members; the dot sits at the center of the city, never on an installation. The panel beside it is the opposite — frozen data from the report: what light cycle each region configures, which is what the map cannot tell you.
The map could not load. Per-country totals are still available in the list.
Regions that do not reach the minimum number of cases we publish on their own are grouped together. Base: 433 of the 435 active zones with a photoperiod; for two the region could not be determined.
Data is shown grouped and may update with some delay. Cities or regions with fewer than 5 members are grouped or hidden. We do not publish anyone's location or activity.
Supercycler lets you define a light cycle of any length, not just 24 hours, and the question that organizes this report is what people do with that freedom. With 165 more photoperiods than in July, the answer barely moved: 228 of the 440 configure a day longer than 24 hours.
440 photoperiods configured
308 zones
127 zones
The proportion dropped a little, from 55% to 52%, and the reason lies with the newcomers: of the 165 additional photoperiods since July, 76 are long-day, 46%. That is still almost half, but less than among the first to arrive.
By stage, the long day is 50% in vegetative — down from 57% — and 55% in flowering, where it did not change. By region it evened out: the Southern Cone is at 51% and Southern Europe at 52%, and the rest of the world, which was at 71% in July, dropped to 64% with more zones. #1 said that difference was fragile, and it was.
This whole cut is about the cycle each zone has configured. Whether the light actually follows it is another matter. Of the 225 active zones with a day longer than 24 hours, in 116 the platform switches the light on and off; in 66 the light is in manual mode — run by the grower or by a timer of their own — and 43 do not have a main light registered yet.
In those 109 the cycle is declared, but we cannot confirm it is being followed. On 24-hour days the share switched by the platform is even lower: 80 of 205. That makes sense: an ordinary timer knows how to do 24 hours and does not know how to do 26.
A photoperiod is two numbers: hours of light and hours of dark. There are 46 distinct combinations in use; these seven cover 378 of the 440 zones:
13 hours of light and 13 of dark, on a 26-hour day, is still the most used setup, and by a wider margin than in July: 155 zones, almost double the 12/12 (87) and more than double the 18/6 (70). Among the risers is vegetative's 20/4, with 14.
308 zones · the classic is 18/6
127 zones · the classic is 12/12
The change in this period is in the right-hand column. In July 12/12 led flowering, 22 against 19. Today 13/13 comes first, 42 against 40. A two-zone difference is not a trend and could flip next month; what can be seen is that the zones that moved into flowering in these two months did not go back to 12/12.
In vegetative 13/13 still beats 18/6, 110 against 64. The pattern from #1 holds: whoever adopts the supercycle does not lengthen the light period, they split it half and half and stretch the whole cycle.
After 24 and 26 hours the count drops off fast. Lengths with fewer than five zones are grouped at the end, from a 2-hour cycle up to a 56-hour one.
A protocol is the grow's task calendar. There are 376 journals open — 204 of them new in these two months — and 345 have a protocol assigned, 92%. These are the most picked:
In July the most used was the vegetative one. Now the two flowering protocols have overtaken it, together adding up to 184 journals against 81. It fits the calendar: June was the month of peak sign-ups, and those grows are the ones that should start bringing in harvests.
Activity keeps pace: in two months 3,145 protocol tasks were completed and 1,351 journal entries were written, more than double the total accumulated up to July. Technique protocols and modifiers are still barely used: four journals in total.
131 zones described their lighting in My Lab, more than double July's figure, and declared 624 fixtures.
624 fixtures declared across 131 zones
LED is still at 90%: 563 of 624 fixtures. The rest is a mix of HPS, metal halide, LEC and models that cannot be classified by name, none with enough cases to be published separately.
The bulk is still between 300 and 600 W, but the top group grew: 39 zones go past 1000 W, 30% of those that filled the field in. Declared power adds up to 166.5 kW, and 130.5 kW of it is in zones with a photoperiod configured.
Light is still the first device people connect, with 381 of 677. But climate is no longer marginal: between ventilation, extraction, dehumidifiers, humidifiers, heaters and cooling there are 188 devices.
Across the 677 active devices, this is what decides when each one turns on:
Sensor rules tripled: from 28 in July to 92. The 310 devices with no rule on the platform are not off, nor are they necessarily switched by hand: 177 are lights — many with their own timer, as seen in section 2 — and 64 are pumps, which are run from the irrigation programs, which work separately.
The 188 climate devices:
Seven out of ten climate devices now respond to something: a sensor, a schedule or the light. 53 are left with no rule. It is what changed most since July, and it connects with section 8, where humidity — the hardest thing to hold by hand — is precisely what improved.
298 sensors, more than double July's figure. A single unit usually measures more than one thing:
Temperature and air humidity cover almost everything. What was barely measured in July has tripled: 42 substrate moisture sensors and 28 CO₂ sensors. They are still few, but they are the base that will let us look at irrigation in the next reports.
120 zones described their substrate in detail, against 49 in July. It is still the smallest cut in the report and the most specific.
112 of the 120 zones filled this in
104 of the 120 zones filled this in
In July there was no dominant medium. With more zones, one appears: coco, with 34, ahead of compost and peat. The other half of the picture barely moved: 43% describe their substrate as living or biologically active (47% in July) and 16% work in inert media.
The first four are still biological, and trichoderma rose to second place. The corrective minerals sit well behind, as in July.
In the two months there were 828 health scans by 193 growers: almost as many as in the previous three months combined.
What follows are suspicions, not diagnoses. The analysis is done by an AI model on a single photo, with no soil or tissue sample: accurate enough to point you somewhere, not to confirm anything.
828 scans by 193 growers
The split is almost the same as in July. Three out of ten scans find nothing, and the system suspected a nutrient deficiency in 45% (49% in July). Pests, together, are 8%.
Across the 373 scans with a deficiency, counting every time a nutrient is named:
Magnesium and nitrogen are still in front. As in July, most of the findings — 240 of 373 — name more than one nutrient, because in a photo they look alike; 112 point at a single one and 21 do not specify which.
63 of the 828 scans ended in a pest or a fungus:
Spider mites are still the dominant pest, with 49. Fungi and leaf spots dropped from 19 to 9. Thrips, whitefly and aphids still do not reach the minimum on their own.
164 traps across 88 zones, three times as many as in July. Each number is an insect counted, not a photo:
Whitefly still dominates the traps, with 800 individuals, and thrips appear with 205. As in July, the trap and the scan measure different things: the trap counts what flies through the tent, the scan what is already on the leaf. That is why whitefly fills traps and barely shows up in scans, and spider mites, which do not fly, do the opposite.
Most trap photos — 136 of 164 — found neither whitefly nor thrips.
In two months, 269 growers opened 700 conversations with the assistant and wrote 5,057 messages, double what had been written in the previous three months. As in #1, we do not read what they wrote: we look at which tool the assistant executed to answer. That came to 1,575 executions.
1,575 tools executed · 269 growers
Zone status still comes first, but it dropped from almost half to a third. What grew is something else: 384 executions, one in four, ended in a proposed change — changing the photoperiod, starting a grow, assigning a protocol, entering zone data. In #1, with the same rule, it was 81 of 1,538: one in twenty.
The AI does not carry out those changes on its own: it prepares them and the grower decides whether to confirm. Those requests came from 82 distinct growers.
Searching the documentation moved into first place and is, once again, the query that reaches the most distinct growers: 128. Behind the status queries come the action ones: 73 requests to change the photoperiod, 66 grows started from the chat and 49 support requests opened that way.
We neither read nor classify the content of the conversations. This whole cut comes from the name of the tool the assistant executed, which is a system record and not anyone's text.
Every zone has two ranges set by its grower: an ideal one and a critical one. Across 29,411 complete days from 156 zones — three times as many zones as in #1 — this is how the time split.
#1 set critical humidity as the baseline: 25% of the time. In this period it dropped to 19.6%, and all three metrics spent more time in their ideal range. But the whole-network average also moves because new zones came in, so the comparison that matters is a different one:
| The same 42 zones | In ideal range | In critical zone |
|---|---|---|
| Humidity · May-Jul | 38.8% | 26.0% |
| Humidity · Aug-Sep | 46.4% | 22.0% |
| Temperature · May-Jul | 45.8% | 7.8% |
| Temperature · Aug-Sep | 48.4% | 8.1% |
In the zones that were in both reports humidity also improved: eight points more in the ideal range and four fewer in the critical zone. Temperature barely moved. The zones that joined in August and September started out better than the old ones: 17.8% critical humidity.
86% of the zones with a photoperiod are in the Southern Cone. July is the middle of winter there and September is already spring: the outside air changes, and with it what is hard to hold inside a tent. A four-point improvement in critical humidity is consistent with the network managing its climate better — section 4 shows it automates it much more — and also with the season simply having changed.
With two periods there is no way to separate the two. It takes a full year, or a comparison against zones in the northern hemisphere, which today are very few. In the meantime we publish it for what it is: a move in the right direction, with no proven cause.
The 0.6% of critical VPD still describes the configuration and not the grow: almost nobody configures a critical VPD range.
Everything above is a snapshot closed on September 30 and it does not change. What follows is the opposite: it is recalculated every day, so it does not carry the same date as the rest of the report and it is not usable for citing the period.
Every month we gather the activity logged across the network: cycles started, new members and grow records. The series are kept apart so each number answers a single question and can be read cleanly.
Once a zone is connected, the system can read its environment, decide on a response and send it to the right device. These figures show how much gets measured, how many actions run and how many decisions are verified across the network. They are not promises: they are activity records.
Average of the last 30 days. The page updates automatically from the real system logs.
Connecting a device turns a routine into a process that can be measured and repeated. From then on the system can log the environment, run actions and leave evidence of what happened.
The underlying question — whether a 26-hour day yields differently from a 24-hour one — needs harvests closed and with the yield filled in. The harvests have started to arrive; the grams have not.
| Data | #1 | #2 | What would be needed |
|---|---|---|---|
| Grows with a harvest date | 9 | 25 | 14 with consistent dates |
| Harvests with the yield filled in | 1 | 3 | Around 30 per group to compare |
| Complete grow closures | 3 | 17 | It is the format that enables the comparison |
| Potency measurements from the network | 0 | 0 | The integration exists and nobody in the network has used it |
Complete grow closures went from 3 to 17, from 14 distinct growers: the format has started being used. But only 3 harvests have the grams recorded, and nothing can be compared with that.
We are not publishing an average cycle length yet either. Of the 25 grows with a harvest date, 8 have the harvest on the same day as the start or earlier, and 9 are still marked as active. And among the 14 consistent ones there are 16-day cycles, which in all likelihood are plants that started being logged when already grown. It is the same problem #1 pointed out and it is still not solved on our side.
The observational base, on the other hand, grew: 228 zones on a long day and 207 on a 24-hour day growing in parallel. It is not a controlled trial, but the year-end harvests will land on two large groups.
The data is taken from the network's grows and aggregated before being published. The rules are the same as in #1, and that is what makes it possible to put one report next to the other.
No equipment manufacturer, fertilizer, breeder or commercial protocol is named. What gear the network uses is not public information.
Everything published is totals and averages, and no group with fewer than five cases is broken out. No number corresponds to one member or one zone in particular.
The period data runs from August 1 to September 30, 2026 and is not modified afterwards. Corrections are published as a dated erratum.
Every number comes from a query saved alongside the report. The definitions and denominators are in the methodology annex.
Sign-ups, scans, queries and environment cover August 1 to September 30. The configurations — photoperiods, devices, sensors, light and substrate — are a snapshot from the day the report was run, October 4, just as #1 was run a few days after its close.
One measurement changed: member activity now counts who logged in during the month, not the date of the last login. Recalculated that way, July gives 303 instead of 306. The annex explains why.
The findings in section 6 are automated readings of a photo, not a lab analysis. The automated accounts the app stores create are excluded.
The methodology annex has the definition of every number, how many cases it is calculated over, and the log of corrections. Report #1 remains published exactly as it came out.
The next report comes out at the beginning of November and covers October. If you would like us to let you know, write to us using the form below.
With #2 every number has something to be compared against. Every zone that starts measuring and every harvest logged with its grams makes the next one say a little more.