The bubble framing describes a magnitude error: money over-rotates, returns disappoint, the field corrects, and what got funded was pointed the right way even if too much of it got funded. What concerns me about monitoring technologies is a sign error, not overpaying for something good, but funding something that works exactly as designed while the design points the wrong way. A magnitude error corrects itself, because disappointing returns are the signal that something was wrong. A sign error doesn't, because in that case success is what does the damage.
Annex V of Council Directive 2007/43/EC is headed "Criteria for the use of increased stocking density". The criteria are two monitoring conditions and a mortality threshold, and satisfying them moves a producer from 39 to 42 kg/m². Sweden is more explicit. Enrolment in the Foot Health Programme is "mandatory to be allowed the maximum stocking density of 36 kg/m²", and the programme factsheet spells out the mechanism: flocks are scored at slaughter, poor scores trigger "a reduction of the maximum stocking density (range between 20-36 kg/m2)", and "when the farmer makes corrections the maximum stocking density (up to 36 kg/m2) is increased again". Density is what good measurement gets paid in. The Commission's 2018 report on the Directive calls this "the first EU piece of legislation which introduces animal-based indicators as a way to regulate animal welfare", and finds that "slightly more than a quarter of EU production is at the highest stocking density", out of roughly 6.5 billion birds a year.
Set that beside the validation literature the post is already worried about. Stygar et al. 2021, reviewing 129 commercial dairy technologies: "In total, only 18 currently retailed sensors have been externally validated (14%)." Gómez et al. 2021, on 83 commercially available pig technologies: "only 5% had been externally validated". Garrido et al. 2023, on respiratory disease detection, found three of 23 studies met their reliability criteria under field conditions, two of those had high risk of bias, and "only one swine technology fully fit our criteria". So the credential that unlocks the density band is largely unestablished as a welfare instrument, which means the evidence gap isn't only wasting money, it's making the credential cheap to issue.
The USDA sequence shows the other half. 83 FR 49048 set criteria for waivers letting young chicken establishments run at up to 175 birds a minute, with FSIS saying the data collected would be used "to assess the ability of NPIS establishments to maintain process control at higher line speeds and to inform future rulemaking, if supported". The February 2026 proposed rule writes the higher rates into the regulations, so they no longer depend on a waiver, and eliminates the annual worker safety attestation that came with them. Monitoring bought the increase, then once the increase was secure, conditions attached to it came off.
None of this is a novel observation, and the clearest version I know comes from inside this funding stream rather than outside it. Lewis Bollard, in Make Cruelty Unprofitable Again, on the last time productivity and welfare moved together: "But this humane progress was quickly undone by further technological progress. New antibiotics, vaccines, and synthesized vitamins let farmers keep vast numbers of animals alive in wretched conditions, severing the historic link between good treatment and productivity."
The bubble frame also assumes the correction is neutral, and I don't think it is. Bubbles deflate but the assets don't evaporate, they get bought, and here the buyer has been named in advance. Spring's Eitan Fischer, talking to AgFunderNews at launch, points to the sizable addressable market and "strategic acquirers such as Zoetis, Merck Animal Health, and Elanco", says "the bottleneck is not the science but crossing the valley of death to commercialization, so we're looking at those early bets where a purely commercial VC may not necessarily want to take on that risk", and expects some portfolio companies to "raise a more traditional venture capital round, and potentially deliver the kind of exits that make sense for venture capital".
That is a fair description of what philanthropic capital is being asked to buy. Donor money carries a welfare technology across the valley of death, the stage where it might fail and where no commercial investor will fund it, and the exit is onto the balance sheet of an company whose revenue depends on selling throughput and profitability to factory farms.
So the failure mode doesn't need the bubble to pop. It is what a success looks like. After the sale the roadmap belongs to the acquirer, nothing in an acquisition obliges them to keep optimising a product for the welfare metric it was funded to serve, and where welfare and throughput diverge it is the owner's objective that decides what the next release is built around. The set of possible buyers is also very small, so "graduate to non-philanthropic funding" resolves to a handful of firms with near-identical incentives.
So the ask is a screening question rather than a moratorium. If this works, does its output become a credential that unlocks density, throughput, or head-per-worker, and once the philanthropic money is out and an animal health company owns it, what stops the gain being re-optimised for the owner's objective? Where there is no good answer to that, is there a counterfactual use of the same philanthropic capital that raises welfare in the short-term without making factory farming more efficient, more durable and more scalable in the long-term?
The bubble framing describes a magnitude error: money over-rotates, returns disappoint, the field corrects, and what got funded was pointed the right way even if too much of it got funded. What concerns me about monitoring technologies is a sign error, not overpaying for something good, but funding something that works exactly as designed while the design points the wrong way. A magnitude error corrects itself, because disappointing returns are the signal that something was wrong. A sign error doesn't, because in that case success is what does the damage.
Annex V of Council Directive 2007/43/EC is headed "Criteria for the use of increased stocking density". The criteria are two monitoring conditions and a mortality threshold, and satisfying them moves a producer from 39 to 42 kg/m². Sweden is more explicit. Enrolment in the Foot Health Programme is "mandatory to be allowed the maximum stocking density of 36 kg/m²", and the programme factsheet spells out the mechanism: flocks are scored at slaughter, poor scores trigger "a reduction of the maximum stocking density (range between 20-36 kg/m2)", and "when the farmer makes corrections the maximum stocking density (up to 36 kg/m2) is increased again". Density is what good measurement gets paid in. The Commission's 2018 report on the Directive calls this "the first EU piece of legislation which introduces animal-based indicators as a way to regulate animal welfare", and finds that "slightly more than a quarter of EU production is at the highest stocking density", out of roughly 6.5 billion birds a year.
Set that beside the validation literature the post is already worried about. Stygar et al. 2021, reviewing 129 commercial dairy technologies: "In total, only 18 currently retailed sensors have been externally validated (14%)." Gómez et al. 2021, on 83 commercially available pig technologies: "only 5% had been externally validated". Garrido et al. 2023, on respiratory disease detection, found three of 23 studies met their reliability criteria under field conditions, two of those had high risk of bias, and "only one swine technology fully fit our criteria". So the credential that unlocks the density band is largely unestablished as a welfare instrument, which means the evidence gap isn't only wasting money, it's making the credential cheap to issue.
The USDA sequence shows the other half. 83 FR 49048 set criteria for waivers letting young chicken establishments run at up to 175 birds a minute, with FSIS saying the data collected would be used "to assess the ability of NPIS establishments to maintain process control at higher line speeds and to inform future rulemaking, if supported". The February 2026 proposed rule writes the higher rates into the regulations, so they no longer depend on a waiver, and eliminates the annual worker safety attestation that came with them. Monitoring bought the increase, then once the increase was secure, conditions attached to it came off.
None of this is a novel observation, and the clearest version I know comes from inside this funding stream rather than outside it. Lewis Bollard, in Make Cruelty Unprofitable Again, on the last time productivity and welfare moved together: "But this humane progress was quickly undone by further technological progress. New antibiotics, vaccines, and synthesized vitamins let farmers keep vast numbers of animals alive in wretched conditions, severing the historic link between good treatment and productivity."
The bubble frame also assumes the correction is neutral, and I don't think it is. Bubbles deflate but the assets don't evaporate, they get bought, and here the buyer has been named in advance. Spring's Eitan Fischer, talking to AgFunderNews at launch, points to the sizable addressable market and "strategic acquirers such as Zoetis, Merck Animal Health, and Elanco", says "the bottleneck is not the science but crossing the valley of death to commercialization, so we're looking at those early bets where a purely commercial VC may not necessarily want to take on that risk", and expects some portfolio companies to "raise a more traditional venture capital round, and potentially deliver the kind of exits that make sense for venture capital".
That is a fair description of what philanthropic capital is being asked to buy. Donor money carries a welfare technology across the valley of death, the stage where it might fail and where no commercial investor will fund it, and the exit is onto the balance sheet of an company whose revenue depends on selling throughput and profitability to factory farms.
So the failure mode doesn't need the bubble to pop. It is what a success looks like. After the sale the roadmap belongs to the acquirer, nothing in an acquisition obliges them to keep optimising a product for the welfare metric it was funded to serve, and where welfare and throughput diverge it is the owner's objective that decides what the next release is built around. The set of possible buyers is also very small, so "graduate to non-philanthropic funding" resolves to a handful of firms with near-identical incentives.
So the ask is a screening question rather than a moratorium. If this works, does its output become a credential that unlocks density, throughput, or head-per-worker, and once the philanthropic money is out and an animal health company owns it, what stops the gain being re-optimised for the owner's objective? Where there is no good answer to that, is there a counterfactual use of the same philanthropic capital that raises welfare in the short-term without making factory farming more efficient, more durable and more scalable in the long-term?