Preliminary synthesis · Astra (AI) draft · September 9, 2026 ·
Astra prepared this draft from the public workshop record, permission-cleared responses, and linked sources. A second Astra agent completed an independent AI source check on September 9. This does not replace review by David Reinstein, workshop participants, or another qualified human reviewer. The claims should be assessed against the linked evidence before the synthesis is disseminated more widely. Sources and limits.
Workshop respondents generally expected a move to cheaper food-grade media. They differed over how far particular substitutions could go, and their cost forecasts remained wide. The public discussion also returned to productivity, equipment costs, and the kind of food people would buy. Media responses · Process discussion · Product discussion
This page brings those views together, including disagreements and the evidence participants wanted next. It draws on the May 8 workshop record and public belief responses. Each point opens into quotations, numbers, and qualifications.
The September scan covered selected public developments after May 8. It was not a systematic review. Belief counts use the July 31 privacy-reviewed release; later submissions have not been checked in this revision.
Scheduled checks prepare local revisions for review. Dated review log.
Potential takeaways · Mind changes · Since May · Possible follow-up · Questions for review · Download draft
Whose views, which evidence, and how to read this
The historical account uses public Sessions 1 and 3 and responses cleared for public sharing. Session 2 was off-record and is excluded. This was a selected workshop, with many contributors working in or supporting cultivated meat. Agreement describes those who answered a particular question, rather than field-wide opinion.
Counts come from the July 31 public dataset and differ by question. “Workshop cohort” includes written contributions; it does not mean everyone attended live. Independent responses are shown separately. A later response is not automatically a change caused by the workshop.
Source access: quotations were checked against the original public-session transcripts or cleared written responses. Times below refer to the original recording. The recording-page links open the edited clips; playback was rechecked September 9, 2026. Transcript excerpts and source notes provide a readable evidence route.
Earlier research, including the GFI 2025 amino-acid cost analysis, remains background. Its scenarios concern an input component, rather than a measured total production cost. It is not a development since this workshop.
1. Cheaper media were widely expected; the route was disputed
All eight workshop-cohort respondents to the food-grade question expected it to account for 80–100% of commercial cultivated-meat output in 2036. Expectations for hydrolysates were less settled: seven respondents gave a 20–81% probability of their use as the base medium for most production. Written comments distinguish full replacement of purified amino acids from using hydrolysates alongside supplements. Responses and qualifications
The counts, the wording, and the disagreement
Food-grade inputs (CM_17). The eight workshop answers were 80, 85, 90, 98, 100, 100, 100, and 100 percent. These are estimated shares of output, not probabilities that a transition happens. Two independent answers were 10% and 85%, so the expectation was less uniform in the wider response set. Individual answers.
Hydrolysates (CM_12). The seven workshop probabilities were 20, 40, 50, 50, 50, 50, and 81 percent. The question concerns adoption as base media for the majority of production by volume. It does not ask what fraction of a medium’s ingredients they replace. Question and estimates.
Fuchs presented hydrolysate-supported growth and recycling as promising routes. Swartz welcomed the proof-of-concepts but asked about high-density suspension growth. Fuchs acknowledged that hydrolysate consistency needed further investigation. S1 exchange, 62:33–65:46.
“It's also unclear if hydrolysates are truly cost beneficial.”
Elliot Swartz, CM_12 written response · Public response cards
Swartz and Bomkamp expected continued use of purified amino acids alongside hydrolysates. Bomkamp’s roughly 80% confidence that both would matter was not an 80% substitution forecast. Fuchs herself assigned 50% to the adoption question. Discussion and written comments.
What remains open: usable nutrients, required supplements, consistency, and the resulting cost per kilogram of harvested biomass. These questions also bear on the July review. The responses do not establish that media is the most important cost uncertainty.
2. Several participants wanted better measures of productive capacity
Morales-Dalmau challenged density taken on its own: higher density can require more inputs. Frohlich emphasized output relative to equipment cost. Swartz asked for feed-conversion and equipment-cost data, while Bomkamp wanted comparable parameters across processes. This was a recurring concern; participants did not select a single preferred metric. Statements and context
What participants wanted measured, and why
“my follow-up question is, what are your inputs?”
Jordi Morales-Dalmau, S1, 74:36–74:45 · Transcript excerpt
His point was to compare inputs and outputs, rather than treat the largest cell count as the cheapest process. Frohlich asked how much material equipment could produce relative to its cost, explicitly noting that he had not examined the project’s model in detail. S3, 38:26–39:11.
Swartz requested evidence on feed conversion, equipment and facility cost, and whether to use larger vessels or more smaller units. Bomkamp asked for a small set of parameters that translate across bioprocess scenarios. S3, 53:11–56:25.
Proposed implication: report medium consumption, biomass harvested over time, and equipment cost for a specified product and process. This reporting set follows their concerns, but its contents still need participant review. A density record alone would not answer these questions.
3. The target food product was an important unresolved question
McNulty and Frohlich asked what the final product needs to deliver, including how much cultivated material would give a hybrid product the desired taste. Bomkamp agreed on the importance of product attributes. Lattanzi’s written response took a stronger position: undifferentiated products would not be enough to attract consumers. This leaves a substantive question about what should be costed, beyond growing cell biomass. Views and distinctions
Product requirements, inclusion rates, and the funding question
McNulty identified uncertainty about the product attributes needed for adoption. Bomkamp explicitly agreed and discussed input cost multiplied by the cultivated share of the final food, while emphasizing limited sensory evidence. Frohlich asked how many animal cells a hybrid would need to provide the desired flavour. S3, 13:16–13:39 and 16:26–22:19.
“I think that undifferentiated products will not be enough to attract consumers, differentiated products are needed.”
Stefano Lattanzi, written response · Public response cards
The focal cost question covers undifferentiated chicken-cell biomass at harvest. It does not settle the cost of a finished food or whether that food replaces conventional meat. The June Forum analysis is an outside contribution to examine on this link between costs and displacement.
Funding views also differed. Asked whether donating $100,000 in 2026 to cultivated-meat development would produce more animal-welfare benefit than donating the same amount to the best established corporate animal-welfare campaigns, six workshop-cohort respondents gave probabilities of 15, 30, 40, 85, 85, and 92 percent. The four independent answers were 10, 12, 80, and 90 percent. These are judgments, not a comparative cost-effectiveness result. CM_10 and responses.
Written reasons differed over timing, market adoption, and what marginal funding would change. Lattanzi conditioned the benefit on successful scale-up and meaningful effects on intensive farming. One anonymous response favoured welfare campaigns in the short term while expecting more from cultivated meat over a longer horizon. Quoted reasoning.
Useful follow-up: specify a food product and ask for evidence about its required inclusion rate, sensory performance, feasible price, and the animal products it could displace. The record supports asking these questions; it does not provide agreed answers.
4. Financing the path to production remained a separate concern
Swartz emphasized the difficulty of obtaining capital for physical production capacity. The earlier forms did not separately elicit the probability of reaching scale. The revised questions distinguish achievable costs (A1), actual costs conditional on production (A2), and production probability (B). This distinction matters when using the forecast to discuss a funding decision. Statement and question boundary
The financing concern and what the cost forecast assumes
“What the industry needs is money that is gonna yield steel in the ground”
Elliot Swartz, S3, 23:04–23:22 · Transcript excerpt
The May workshop wording did not explicitly condition on qualifying plants existing; a June 25 revision added that condition. We cannot assume every earlier respondent interpreted it identically. A cheap achievable-cost estimate can coexist with a low probability of deployment. Question history and revised definitions.
The public discussion included public and philanthropic support and commercial routes. The checked record does not establish a collectively preferred financing strategy. Recording entry. The PARIMA/Vow announcement and SuperMeat financing provide subsequent cases to investigate.
5. Cost estimates differed substantially; changes of mind are not yet documented
The public release contains 14 standardized cost estimates: ten from the workshop cohort and four from the independent round. The independent point estimates alone range from $6 to $500 per kilogram. The public release does not identify matched before-and-after responses from the same person, and this review found no explicit report of a change of mind in the checked public material. Counts, limits, and what participants said
What the elicitation tells us, and what would establish an update
The earlier questions concerned 2036 wet chicken-cell biomass at harvest, in 2025 US dollars. The workshop and independent cohorts received different wording about whether production at scale was assumed; neither cohort has been relabeled as answering the revised A1 or A2. The ten workshop estimates range from $1 to $100/kg. The four independent estimates are $6, $40, $50, and $500/kg. These are ranges of point estimates, not uncertainty intervals. Individual estimates, intervals, and filters.
An additional written estimate in edible-meat units remains visible but is excluded from this standardized pool. Some answers lack intervals or have interpretation flags. Counts and aggregates must respect those distinctions. Methods.
One independent response challenges whether optimistic inputs in bottom-up models can work together, and asks for comparison with established microbial production. This is a substantive skeptical argument, not a validated cost floor. Public reasoning.
“we are kind of the believers”
Bert Frohlich, S3, 51:03–51:14 · Transcript excerpt and context
Frohlich was commenting on the group’s composition. Reinstein replied that he had tried to bring in sincere skeptics. Independent responses broaden the picture but do not make it a representative expert survey.
Mind changes. Reinstein asked whether anyone had changed their mind (S3, 50:07). The discussion that followed described evidence that would change views, including Bomkamp’s emphasis on bioprocess evidence. These statements and the later submissions do not establish workshop-induced updating. Public transcript context · Before/after data and methods.
Question for participants: “Did the workshop change your estimate or uncertainty? What was your view before and after, which evidence mattered, and may we quote your explanation?” Clarifying a question without changing a forecast would also be useful to report.
Since the workshop
Checked September 9, 2026. These developments may change how the workshop questions should now be pursued. They do not rewrite what participants said in May. This limited scan did not locate an independently auditable new full-production cost estimate establishing commodity-meat price parity. Scope, sources, and review log.
July 2 · PARIMA reports a multi-tonne run using Vow’s 22,000-litre line
PARIMA’s announcement describes cultivated-duck biomass grown in suspension without added growth factors or a differentiation stage. It reports a 99% cost reduction relative to earlier runs, but gives no absolute cost per kilogram or detailed cost accounts.
This is relevant company-reported evidence about scale and shared infrastructure. It does not establish chicken-price parity, repeat-run reliability, or the economics of differentiated muscle. Follow-up for PARIMA/Vow: harvested wet and dry mass, medium consumption, cycle time, repeated runs, and full costs including equipment and capacity payments. These would help answer the productivity and financing questions.
July 3 · A hydrolysate review focuses attention on what cells consume
Hoek and colleagues compare microbial hydrolysate composition with cellular amino-acid demand. They write that similarity in composition “does not translate into meeting cellular amino acid demand.” Their quantitative comparison uses Chinese hamster ovary (CHO) cell data where appropriate cultivated-meat data are lacking.
The review sharpens the substitution disagreement: which nutrients are usable, which supplements remain necessary, and how much biomass results? It supplies no commercial cost demonstration. Follow-up for the authors and workshop media specialists: identify a relevant food-species uptake dataset and a suspension-culture test that could distinguish the competing expectations.
July 10 · New Great Britain guidance makes some regulatory requirements more explicit
FSA/FSS issued four guidance documents covering hygiene, product identity and production, application quality, and taste trials. Guidance is not a product authorization or a promised approval date.
Follow-up for applicants: which requirements affect medium choice, process changes, factory design, or time to launch, and does the guidance change a previous forecast? The useful update would be a specified effect on the route to market.
June 26 · An outside Forum synthesis connects costs to demand and animal displacement
PabloAMC’s “Cultivating hope” revisits cost projections and links an interactive demand model. It was marked Curated when checked. Its analysis is a useful challenge to examine; curation is not validation, and its cost discussion largely uses earlier studies and company claims.
Follow-up for the author and workshop participants: which cost assumptions and demand responses determine the projected displacement? The discussion’s correction about feed-conversion denominators is also relevant to comparable process metrics. This is new outside analysis, rather than new production data or a workshop position.
Other updates: financing, research priorities, and publication changes
May 12, financing. Agronomics reported a $6 million initial close for SuperMeat: $5 million in newly issued Agronomics shares and $1 million cash, toward a targeted $10 million round. The announcement sets out licensing and commercial-validation milestones. It does not verify unit economics; its production-cost estimate dates to 2024. Ask which milestones have since been completed and which results can be shared.
June 9, research priorities. Rethink Priorities’ AI and Cultivated Meat report discusses scarce data, confidentiality, funding, and process-specific uses of AI. Its assessment runs through April 2026: a later publication about earlier conditions. It gives a reason to compare a modeling exercise with work making useful datasets available.
June 8, journal publication. Oey and colleagues’ algae/media-recycling work had a February preprint. Publication status changed after the workshop; the core results were already public. The experiments retain serum and do not establish an animal-free commercial process or net production-cost savings.
The review log also distinguishes earlier work from later publicity. Publication, demonstration, and coverage dates should not be treated as interchangeable.
What should the next follow-up accomplish?
The clearest supported direction is to narrow the question and make comparisons more useful. Swartz proposed separate attention to media, bioprocess, and product design. Bomkamp asked for comparable parameters. Frohlich saw techno-economic analyses as a way to locate the assumptions that matter most. These give reasons for focused follow-up work, but do not establish its format or which exercise the group would prioritize. A modeling session is one option; the modeling follow-up proposal sets out a possible structure. Who proposed what
Participant proposals, their purpose, and the limits of endorsement
Swartz listed missing process, equipment, and product data (S3, 53:11–54:14), then proposed narrower workshops on media, bioprocess, and product design (61:45–62:30). Reinstein agreed that the narrower approach made sense (62:34). Bomkamp’s common-parameter proposal aimed to make bioprocess scenarios comparable (55:25–56:25). Source excerpts and context.
“What factors are the most important or sensitive”
Bert Frohlich, S3, 62:38–63:18 · Transcript excerpt
Frohlich proposed using further model discussion to identify sensitive factors and help choose topics. He left the format open and mentioned a bioreactor-modeling consortium. Reinstein welcomed model sharing or a hack session. S3, 62:38–64:17.
The checked record contains no vote or shared ranking making reconstruction of a published model the highest-value activity. Interest in further modeling does not establish endorsement of that particular task.
A candidate exercise: reproduce one scenario, then test a consequential disagreement
Editorial proposal for review. Select a published scenario whose equations and inputs can be checked. Reproduce its result before changing an assumption that participants disagree about. Show whether the change alters total cost, which other inputs must change with it, and what observation would distinguish the competing cases.
Why this might help. Reproduction would reveal whether the comparison starts from a faithful implementation. A focused sensitivity exercise could show whether a disputed assumption matters enough to change a research or funding choice. This follows Frohlich’s use of models and Bomkamp’s request for comparability; reconstructing a particular model is this draft’s proposal.
When another activity would be better. If the decisive uncertainty is missing nutrient-consumption data, equipment quotations, or sensory performance, obtaining those data may be more useful. Swartz’s requests and the new hydrolysate review make these concrete alternatives. The exercise should earn its place against them.
Proposed selection question: which unresolved assumption could change a specified decision, and which feasible activity would reduce that uncertainty most? Decide that before choosing the form of follow-up. If a modeling session looks useful, use the answer to shape the modeling follow-up agenda.
What a useful contribution would contain
A precise claim or proposed change; its source and units; the relevant species, cell type, product, and scale; dependencies on other assumptions; the resulting cost or output change; and the observation or decision that makes it worth resolving. An unresolved discrepancy is also a useful output.
Questions to check before disseminating this more widely
For David: does this fairly describe the balance of the public discussion? Does it give appropriate weight to media, productive capacity, product requirements, and financing? Which important disagreement or priority is missing from the linked record?
Specific questions for participants and follow-up contributors
- Fuchs, Swartz, and Bomkamp: is the hydrolysate disagreement fair? Are you disagreeing about full replacement, practical savings, or adoption? What result would distinguish your views?
- Morales-Dalmau, Frohlich, Swartz, and Bomkamp: which minimum set of process metrics would permit useful comparisons? Which missing measurement is worth obtaining first?
- McNulty, Frohlich, Bomkamp, and Lattanzi: what evidence would resolve the product question? Does a specific hybrid product need differentiated cells, and what inclusion rate delivers an acceptable result?
- The full group, including independent respondents: did you revise your estimate or uncertainty, and why? Which next questions could change your research or funding choices? Please rank a few and explain the decision at stake.
- Potential modeling contributors: would a reproduced scenario and one sensitivity comparison answer a question you care about? If a dataset, experiment, or independent evaluation would be more useful first, which one and why?
These are proposed questions, not messages already sent. Revised estimates should retain the earlier answer, explain the evidence that changed the view, and specify what may be quoted publicly.
Corrections are welcome through Hypothes.is on a selected passage, or the project discussion. A useful correction identifies the claim, links evidence, and suggests a fairer characterisation.