1. Probability-weighted process mix removed
15 Nov 2026 · v0.3
There's no point imagining a frontier plant which uses a worse process, and anyway process doesn't seem to be the major issue here.
Alex Bates's rebuild, with first-pass edits. A step-structured rebuild of the Unjournal cost model by Alex Bates (file v0.9.7, 16 November 2026), with host notes and minor changes made in response to comments: typeface, labels and wording, explanations, and display of the "What has to go right" rows. Alex's model, inputs and code logic are unchanged. The main Unjournal model remains the master copy (Alex suggested this; neither model is fully validated).
Every input can be changed: open Assumptions under each step (central estimate and 95% optimistic value, per process where it matters). Alex removed some structural choices of the main model, such as the probability-weighted process mix and contract manufacturing; see his Notes at the bottom. Open points from the first round of comments: Questions for Alex. Comment with Hypothes.is: the ‹ tab on the right edge.
Unjournal pivotal questions · cultured meat · rebuild v0.9.7 · first-pass edits
For usability, a very simple comparison between dry cell mass and commercially available chicken is assumed.
Changes to discuss with the team.
15 Nov 2026 · v0.3
There's no point imagining a frontier plant which uses a worse process, and anyway process doesn't seem to be the major issue here.
15 Nov 2026 · v0.3
There's no point imagining a frontier plant which contracts manufacturer - if there was, we should just model the costs of the contract manufacturer.
15 Nov 2026 · v0.4
Relevant to the scope.
15 Nov 2026 · v0.4
Not relevant to the scope.
15 Nov 2026 · no model change
Finding. The Learn page puts the seed train at 5–10% of production cost, and a few lines later at 5–15%, "mostly labour". It cites Humbird's $1/kg labour figure, but that covers the whole plant's staff (95 people for fed-batch, 132 for perfusion) against a total cost of $37/kg, so all plant labour is about 3% of cost. The page also says seed-train medium volumes are orders of magnitude smaller than in production. In the model the seed train is equipment only, about 0.5% of cost, and that capital ratio is consistent with Humbird (about $2M of seed-train reactors against $34M of production reactors).
15 Nov 2026 · no model change
Finding. The cost breakdown summary says capital is 15–35% of total cost and fixed running costs 10–25%. In the model they are about 7% and 3.5% across all worlds, but about 13% and 22% in the cheapest 10% of worlds. Published cost studies model plants near parity with conventional meat, so their shares match the model's cheap worlds, not its typical one.
16 Nov 2026 · v0.6
The original bought a blended share of reactor capacity (55% ± 15) at food-grade prices and the rest at pharma prices. This doesn't make sense at a frontier factory, and is more relevant as an industry average.
16 Nov 2026 · v0.8
There's no point imagining a frontier plant which uses a worse process.
16 Nov 2026 · corrected in v0.8
Repeated issue with the model where technology which is cheaper on average is used even when simulated result makes it more expensive in some particular world.
16 Nov 2026 · no model change
Finding. The share of output lost to failed runs grosses up every cost line, including harvest. But a failed run is thrown away, not harvested, so harvest shouldn't scale with failures. Recovery losses in the harvest step itself do belong there.
16 Nov 2026 · v0.8.6
Change. (1) Reactor prices are entered as two prices, pharma-grade and food-grade (default 0.35 × pharma-grade), drawn together so they move as one; this replaces the original's separate 0.15–0.55 ratio. (2) The whole-plant multiplier also has a food-grade column, since a food-grade plant may need far less cleanroom space, sterile pipework and validation (default: no difference). If food-grade reactors work, the plant uses whichever installed cost is lower. (3) Fixed running costs ($1–6/kg wet at a 20 kt reference plant, scaled by plant size and utilisation) and utilities and consumables ($0.30–3/kg wet) are merged into one plant operating cost per kg: central $23.50, 95% optimistic $9.00 per kg dry ($4.70 and $1.80 per kg wet), matching the median and 2.5th percentile of their sum. The fixed-cost scale exponent goes with them.
16 Nov 2026 · corrected in v0.8.7
The original model has no limit to the scaling factor of capital, which means the best plant is just one which grows massive. This is not plausible, and leads to spurious effectiveness.
16 Nov 2026 · v0.9.4
Errors in the original model, corrected here.
David Reinstein read through this page on 17 November 2026 and left comments with Hypothes.is (open the sidebar on the right to see them). Claude, working with David, collected the open points below. Each one says what the issue is, why it matters, what we would suggest, and what we would like to know from you. Several of them apply to the main Unjournal model as well, which shares the same structure in these places.
We will not change how your model calculates anything without your agreement. You can reply with Hypothes.is on this page, on Slack, or simply by changing your own version.
A few terms used below. A world is one simulated draw of every input. The competitive zone is your range of cell costs between parity with chicken ($12/kg dry) and $36/kg dry. The share of worlds below $36/kg dry is our main summary of how often cultured meat comes out price-competitive. A minimal set is one of the smallest combinations of favourable inputs that gets a world into the competitive zone, as computed in your "What has to go right" section.
Plant operating cost: one input or three?
Your page merges two of the main model's inputs, fixed running costs (labour, maintenance, overhead) and utilities and consumables, into one plant operating cost. Its central value, $23.50 per kg of dry cells, is the median of the sum of the main model's two ranges. Those ranges have no published source, so the merged number does not either.
This input matters a great deal: it appears in 98% of the minimal sets, more than any other input. But its parts are driven by different things. Labour depends mostly on the size and complexity of the plant. Maintenance and insurance are usually estimated as a share of the plant's capital cost (Humbird 2021 uses 4% and 5% of capital a year). Utilities and consumables scale with output and differ by process; for example, perfusion uses filter membranes that fed-batch does not need. Drawing one combined number independently of everything else means a simulated world can pair very cheap running costs with a very expensive plant, which is not realistic.
We suggest splitting it into three parts: labour per plant; maintenance and insurance as a share of capital; and utilities and consumables per kg, with values that differ by process. This follows point 3 of your own literature review. Does this fit how you would want the frontier plant modelled? (There is also a sidebar note on this.)
Perfusion membranes are not costed.
Perfusion and continuous culture keep the cells in the reactor while medium flows through, which needs cell-retention equipment such as filter membranes. Neither your model nor the main model charges for these. In Humbird's perfusion design the membranes alone cost about $4.53 per kg of wet cells; Pasitka et al. report that single-use filters cost $8.78 per kg of wet cells more than reusable ones.
This matters because almost all of the worlds that reach the competitive zone use perfusion or continuous culture. Leaving out a cost that only those processes carry makes them look better than they are. Would you add a membrane cost for those two processes, with a range that allows for reusable filters?
The limit on economies of scale.
You fixed a real problem: in the main model, reactor cost per litre keeps falling as the plant gets bigger, with no limit. Your fix prices every world's reactors as if the plant held 6,500 m³ of reactor volume. But the worlds that reach the competitive zone are high-productivity worlds that only need about 530 m³ (median) to make the same output. Pricing them as a 6,500 m³ plant gives them a much bigger volume discount than before, so the share of worlds under $36/kg dry rises from 0.049% to 0.080% when this change is made on its own. The 6,500 m³ figure also has no source yet.
Could the limit be tied to engineering instead? Most economies of scale come from making each vessel bigger, and vessel size is limited by oxygen transfer, mixing and sterility; beyond that, plants add identical vessels. So we would suggest a scale discount on vessel size up to a maximum vessel size, a smaller discount for shared plant infrastructure, and a plant no larger than the 2036 market and lenders would plausibly support. More detail is on the notes page.
Inputs that can be set separately for each process.
Many inputs have a tick box, "Same for all three processes", which a user can untick to enter different values for fed-batch, perfusion and continuous culture. This leaves a lot for the user to invent. Some inputs should clearly be the same for every process (input prices, the cost of capital); others differ in systematic ways that a user would have to work out for themselves.
We suggest: lock the inputs that should not differ by process; give the inputs that do differ separate default values for each process, with sources; and where the difference follows a pattern, write it as a rule (for example, perfusion density as a multiple of fed-batch density, or membrane costs charged only for perfusion and continuous culture) rather than as free columns.
Other choices besides the process.
The page treats the process (fed-batch, perfusion, continuous) as the main fork in how a plant could work. Other forks may matter as much: the cell line (for example genetically engineered or not, immortalised, adapted to grow in suspension without serum), the route to cheap medium, and whether a scaffold is used. Should some of these be modelled the same way as the processes, as a few named routes, each with a probability and its own inputs?
How growth factors are represented.
"Cheap growth factors exist by 2036" is a yes-or-no switch inherited from the main model, and it is fairly arbitrary: "cheap" is really a matter of degree, and several different advances could get there, with different consequences. Cheaper production of the proteins (for example by precision fermentation) lowers the price per gram. Longer-lasting or more potent factors lower the dose. Cell lines that make their own growth factors, or need fewer, lower the requirement altogether.
A cleaner approach might be to elicit the growth-factor cost per kg of cells directly, as a range, and to treat those advances as distinct routes. What do you think?
The "industry maturity" factor, and inputs that should move together.
Both models draw one "industry maturity" value per world and use it to nudge four technology probabilities and the cost of capital up or down together. Which inputs it links was a simplifying choice, not based on evidence, and in practice it has almost no effect: giving each of those inputs its own independent draw moves the share of worlds under $36/kg dry only from 0.049% to 0.048%. Meanwhile, things that plausibly share a cause are not linked at all. Better cell lines, for instance, would likely lower growth-factor needs, raise cell density and nutrient yield, and make long continuous runs more reliable, all at once.
On the notes page we propose replacing the single maturity factor with a few named common causes (cell-line quality, food-grade supply chains at scale, and the investment and policy climate), each linked to the inputs it plausibly affects, while keeping every input's own range exactly as entered so that nothing is counted twice. A related question: the whole analysis is arguably conditional on a large R&D effort. Should that be stated as an explicit scenario rather than represented by a random maturity draw?
Sources for every input.
Each input should show where its default comes from, even if only in a tooltip or a link. At the moment some have no source at all, including the two cell-bank inputs, the plant operating cost and the 6,500 m³ scale limit. Most of your other defaults come from the main model, whose sources are in its technical reference.
Medium recycling.
The "medium-use multiplier" is the number of reactor volumes of fresh medium used per harvest. You do not allow it to go below 1, because using less than one reactor volume of medium per harvest is only possible if spent medium is cleaned and reused, and the model does not charge anything for that. We agree that recycling should not be free, but it is physically possible, so ruling it out entirely may be too strict.
We checked how much it matters, by running the main model's settings with and without the limit and comparing the same simulated worlds. Without the limit, free recycling lowers the cost in only 3.7% of worlds, because in most worlds the amount of nutrients the cells need already sets how much medium is used. In those worlds, adding the limit raises the cost by 7.6% at the median (by 28% or more in a tenth of them). Among the worlds under $36/kg dry, 29% are affected, by 8.9% at the median.
Because the typical effect is under 10%, we would suggest keeping values below 1 but charging a cost for each litre of medium saved by recycling (covering the equipment, energy and replacement nutrients), rather than adding another yes-or-no choice for users. Does that seem reasonable to you, and do you know of a source for the cost of recycling medium?
Whether breakthroughs arrive together: a new optional setting in the main model.
This follows from point 7. We worked out that the maturity factor links the technologies with a correlation of only about 0.01, which is effectively no link (a correlation of 0 means two technologies succeed or fail independently; 1 means they always go together). On 17 November we added an optional setting to the main model, under Technology Adoption on the Advanced page, that links hydrolysate medium, cheap growth factors and food-grade proteins more strongly while keeping each entered probability exactly as it is. With the moderate setting (0.3), cheap growth factors work in about 55% of the worlds where hydrolysates work, instead of 50%. It is off by default.
The 0.3 value is a convention borrowed from other fields (cost-risk analysis and a drug-development model), not a measurement; we found no measurement for cultured meat. We are inclined to make a moderate link the default, and would like your view first: is 0.3 a reasonable starting point, and should it also cover whether perfusion and continuous culture work at scale, as your review suggests? Background is in a research note (written by an AI research agent, not yet reviewed).
A small correction to check in your literature review: the Pasitka plant figures.
The review gives the cost of cell mass in Pasitka and colleagues' 50,000-litre plant as about $24 per kg of wet cells, alongside their headline $6.2 per lb for the finished product. In the paper's supplementary Table S10 these two numbers belong to different plant designs. The design using tangential-flow filtration to retain cells costs $24.06 per kg of wet cells and $6.80 per lb of product; the large-scale perfusion design costs $21.49 per kg of wet cells and $6.22 per lb of product. So "$21 to $24 per kg wet, depending on the design" would be accurate.
We could read the paper's abstract and supplement but not its main text, which is behind a paywall. So we could not confirm three figures the review takes from the main text: the test run of 1.73 kg of wet cells from 110 litres of medium (about 64 litres per kg), the cell densities of 130 million and 28 million cells per mL in the two media, and that the paper states no dry-matter content for the cells. If you can share your copy, we will check them; one of our pages quotes the 64-litre figure.
Where the competitive zone ends.
You set the upper edge of the competitive zone at $36 per kg of dry cells, about three times your parity line and roughly the price level of organic chicken. Is that the right line for "price-competitive"? For reference, $36 per kg of dry cells works out to $9.90 per kg as meat made entirely of cells (at the water content of chicken meat), or $4.73 per kg for a product that is 25% cells and 75% plant-based filler at $3/kg. US wholesale boneless chicken has averaged about $3.50 per kg.
The wording, label and display changes already made in this version of your page are listed in the hosting script (scripts/host-bates-rebuild.mjs, in the list called EDITS). We are happy to copy any of them into your own version.