New here. Looking for interest in Autonomous AI enabled grower

[harvesterai]

Hello all. I am a cannabis enthusiast but I've never grown my own. I am a software engineer by trade. I was looking to get info and a sense for interest for a product I am engineering. The product is an autonomous grower. The system would support (1 - 6), 5 gallon deep water cultures (DWC). The basic system comes with just one 5 gallon DWC but you can add up to 6 DWC per unit. The system would use a central reservoir where a Nutrient A, Nutrient B, PH up and PH down containers would be mixed into the central reservoir to provide nutrients during different stages of growth. The light would also be controlled based on growth cycle. The system involves some tinkering by the user unlike other systems like Hey Abby that are for brown thumbs. The base model logic for dosing, lighting and environmental factor control will be done locally and customized by the user. However a subscription base model will also connect to the cloud for monitoring and dosing recommendations. Eventually there will be enough data where Artificial Intel will monitor and recommend dosing based on metrics such as strain,environment, local factors, etc. Please any advice, comments, and recommendations are welcomed.

 

[Galgrows]

I'm not at all interested in robot growing

 

[TheGoldenRoad]

I second that.

 

[ArtfulCodger]

I'll suggest that you learn to grow before you try to automate growing.

 

[voltage]

I feel like an autonomous robot atm.

growing it is half the fun

 

[amneziaHaze]

ac infinity allready has that.
but for your sistem to dose small amout of nutes you need to hit that perfect mart you need one of those tiny pumps and they break after 2000 uses....
there is a couple of diy sistem allready made there is allsoo a youtuber that made it all for free you just have to assemble it but the code if open source

 

[ATLien415]

peristaltic pumps and stuff can be automated already, I don't see how AI is going to know anything about the cultivars much less dosing. You need to know the cultivar like the back of your hand, top to bottom, front to back, before you try to automate your runs for a cut. Also, like others have said ACI is way ahead of you on that front. The problem with any AI in the grow room is bro science. AI is nothing but the bastardization of linguistics, and is only as good as the training data. What training data could you even use? Not any of the major forums because bro-science and misdiagnoses.

You really want to help the community out with your skills, make something open source for a DIY cannatrol that you actually share with the community. Last two EEs who did this went dark the minute it was done.

 

[Bobcat.Branch]

But who will calibrate the measuring instruments? Will it drive itself to the store to buy nutrients? Will it detect, diagnose, and remove pests and disease?

I'd suggest you find a way to better mankind and make the world cleaner, healthier, and less violent. Mmmm, k?

AI can eat a dick, too. Fox your life, sonny.

 

[amneziaHaze]

peristaltic pumps and stuff can be automated already, I don't see how AI is going to know anything about the cultivars much less dosing. You need to know the cultivar like the back of your hand, top to bottom, front to back, before you try to automate your runs for a cut. Also, like others have said ACI is way ahead of you on that front. The problem with any AI in the grow room is bro science. AI is nothing but the bastardization of linguistics, and is only as good as the training data. What training data could you even use? Not any of the major forums because bro-science and misdiagnoses.

You really want to help the community out with your skills, make something open source for a DIY cannatrol that you actually share with the community. Last two EEs who did this went dark the minute it was done.

super easy with with ppm and ph knowledge.

cannatrol is extreamly simple its just a dehumidifier inside a fridge.....

 

[ATLien415]

super easy with with ppm and ph knowledge.

cannatrol is extreamly simple its just a dehumidifier inside a fridge.....

Based on growing experience, agree to disagree. Until you ran a girl every which way multiple times, you don't know how she is going to like to be fed...and chances are you aren't pulling tip-top-shelf from her until then. Sure you can do some elementary level checks on the EC (not PPM if you want it to be done right) and the pH and automate the feeding to an extent. I would love to see your methodology of using only 'ppm and ph knowledge' applied to a 100 cultivar grow of common commercial strains and we can judge how appropriate that one-size-fits-all approach is to the average plant in the population. My point is that you're going to need to be able to adjust things on the fly, things that in order to automate would require a dozen or more reservoirs and an equal amount of pumps.

Completely lose me here, the cannatrol is (to me at least) extremely simple but is very much not 'just a dehumidifier in a fridge.' That is really trivializing the science of how it works and why it works and requires a plethora of assumptions to even be a true statement. I think the patent offices agree, as a 'dehumidifer in a fridge' would very much not be patentable. You can't say, here is an iphone in a microwave and get a patent.....

 

[Jimmie]

Hey, whip me up something to get beer out of the fridge for me, that’s a start.

 

[amneziaHaze]

Based on growing experience, agree to disagree. Until you ran a girl every which way multiple times, you don't know how she is going to like to be fed...and chances are you aren't pulling tip-top-shelf from her until then. Sure you can do some elementary level checks on the EC (not PPM if you want it to be done right) and the pH and automate the feeding to an extent. I would love to see your methodology of using only 'ppm and ph knowledge' applied to a 100 cultivar grow of common commercial strains and we can judge how appropriate that one-size-fits-all approach is to the average plant in the population. My point is that you're going to need to be able to adjust things on the fly, things that in order to automate would require a dozen or more reservoirs and an equal amount of pumps.

Completely lose me here, the cannatrol is (to me at least) extremely simple but is very much not 'just a dehumidifier in a fridge.' That is really trivializing the science of how it works and why it works and requires a plethora of assumptions to even be a true statement. I think the patent offices agree, as a 'dehumidifer in a fridge' would very much not be patentable. You can't say, here is an iphone in a microwave and get a patent.....

well i have been growing like that for the last couple of years and each year i grow 2-4 different strains and never have i burned one some get underfeed for some time but eventually it balances.

as for the canatrol can you explain whats soo special i have a home made version and buds come out perfect. you keep the room cool and you keep the air humidity at 55-60 what else does the canatrol do spits out a graph.the machine is soo simple they need you to put a wet spounge inside to keep the dehumidifier working

edit
isnt ppm the same shit as ec?

 

[ATLien415]

my point is that a shotgun spray approach to nutrition would get very complicated or be done poorly, like doing the automation right would require single element reservoirs for all macro and micro, pumps for each, testing capabilities, and feedback loops (or just guessing based off macro level metrics like pH and EC). Then once you have the capacity to automate it, then you have the additional complexity of discerning plant-specific behavior which is also distinct within the same genome across phenotypes. Having an AI actually pull top-shelf flower would certainly require some imaging and processing, to actually dial in the plants (because EC and pH don't tell you much about WHAT the plant is eating). Once you have the ability to do all that, then you need to do what the internet has not done yet - get a properly labelled dataset for cannabis dich. key steps...

A cannatrol, or any properly made mockup, is going to have two distinct thermoelectric coolers. A larger one running a feedback loop based solely on temperature. The smaller one running a feedback loop for RH only as the T is held constant. I get the impression that you have not looked at the patent or any of the whitepapers surrounding the product. The key is the peltier, not a fridge or a dehumidifier, peltiers specifically. A device which can target humidity and temperature without leveraging a compressor cycle, which is mayhem on the local scale of trichome heads. This is science from meat curing and cheese making taking to the umpteenth degree. We do not care about the RH, only when the T is constant because then we aren't targeting RH are we? No, we are actually targeting dew point. The DP is where ALL the magic happens. There is a setting where in 80% RH you never grow mold...this is science...this is the power of dew points. Like you know that the sponge is also key. Why though? The sponge is providing unbound water which is a further mitigation step against vapor pressure gradients and trichome head contractions. The product is a first principles approach to preserving and maximizing potency in cannabis.

Like here is an example setting:
Dry Stage
Dry bulb = 22.2°C /72°F
Dew point = 12.8°C /55°F → 55% RH Transition = Slope 3 days
Cure Stage
Dry bulb = 23.9°C /75°F
Dew point = 13.5°C /56.3°F → 52% RH Transition = Step 2 days
Hold Stage:
Dry bulb = 20°C /68°F
Dew point = 12.6°C /54.6°F → 62% RH

The RH is only a matter of circumstances, a derived value of the dew point and temperature.

In a way, yes they are. EC is superior though as it relates to dimensional analysis (the units, namely)

 

[harvesterai]

this is what I was thinking.

New Use Case: Phenotype Hunting for Breeders and Cultivators​

Overview:
Harvester AI provides precision-controlled, per-plant automation and environmental logging, making it a powerful tool for phenotype selection in breeding and research environments. Cannabis breeders and craft cultivators can streamline their pheno hunts by collecting detailed sensor data, automating unique nutrient profiles, and correlating plant performance with environmental variables.


Value Proposition:

• Automate nutrient and lighting variations across buckets to test how different phenos respond.
• Collect real-time environmental and plant health data per bucket, including pH, EC, temperature, water level, and PPFD.
• Tag buckets with strain ID, seed number, generation, and log detailed phenotype observations (e.g., structure, resin content, terpene notes).
• Integrate with lab test data for THC/CBD/terpene profiling.
• Use Azure ML to identify optimal traits and growing conditions for elite phenotypes.

Target Users:

• Commercial cannabis breeders
• Genetic R&D labs
• Cultivators selecting mother plants
• Licensed producers aiming for consistent premium output

Competitive Edge:

• Most competitors focus on generic environment control; Harvester AI uniquely supports per-plant experimentation and data-driven phenotype selection with edge + cloud integration.

 

[LoveGrowingIt]

I am a software engineer by trade. I was looking to get info and a sense for interest for a product I am engineering. The product is an autonomous grower.

What so-called 'autonomous' systems cannot do, AFAIK, is to look at a plant and diagnose its condition and prescribe what it needs. Until that's achieved, there cannot be autonomous growing.

 

[harvesterai]

What so-called 'autonomous' systems cannot do, AFAIK, is to look at a plant and diagnose its condition and prescribe what it needs. Until that's achieved, there cannot be autonomous growing.

You’d be surprise at the capabilities machine learning and visual recognition.

 

[Bobcat.Branch]

this is what I was thinking.

Target Users:​

• Commercial cannabis breeders
• Genetic R&D labs
• Cultivators selecting mother plants
• Licensed producers aiming for consistent premium output

Rule #1: Know your audience

 

[Newty]

AC Infinity has the controller AI+, can automate everything in your grow with the proper adapters and sensors.
Not only automates but AI learns from the environment to run the equipment the best it can.

They even have a soil moisture meter, definitely my favorite sensor!!!

For the Hydro people, they released a dosing pump to deliver precise doses of nutrients.

Even have a liquid presence sensor, so it can automatically refill reservoirs, humidifiers, etc

 

[ATLien415]

Nice addition for keeping your pH in line, wonder what the sensor refresh rate is... 21 mL intervals means likely diluting yourself prior and would be interesting with buffering solutions....unless you were running like a 40 or 60 gallon setup off this little guy.

Regardless, they have cool stuff in the pipeline. Who knows when we will get it. Says shipping by EOM April. Can't wait for their dehumidifier to drop, they have a page up but isnt out yet.

 

[ATLien415]

You’d be surprise at the capabilities machine learning and visual recognition.

You'd be surprised at the pitfalls of machine learning and visual recognition. The core problem is training data. Get me a large and accurately labelled dataset for a cannabis dichotomous key to do this right and.....you and I are rich enough to buy this site and send every member a car.