On 'Scaling Up' and Being the Right Size (Greshake Tzovaras)

Bastian Greshake Tzovaras works on citizen science and open-knowledge projects, and in an August 2026 post he pushes back on the question he hears most in that work: how do we scale this up? scaling-up-right-size The question takes for granted that scale is wanted, necessary, even good. His answer, borrowed from his background in evolutionary biology, is that nothing scales and stays what it was.

Haldane's gazelle

J.B.S. Haldane's 1926 essay On Being the Right Size shows how physical constraints tie an organism's shape to its size. A gazelle made large would break its legs unless it grew short, thick ones like a rhinoceros or a braced, stretched-out frame like a giraffe, so there is no gazelle above a certain size. Small animals have little to fear from gravity, but surface tension makes getting wet a crushing load for them. Small animals take in enough oxygen through their skin, while large ones need gills or lungs for the surface area. Haldane, a committed socialist, carried the point over to human institutions himself: Greek democracy could not work beyond a small city, and he found it "no easier to picture a completely socialized British Empire or United States than an elephant turning somersaults or a hippopotamus jumping a hedge."

Two planners took the idea further. Jane Jacobs, in her 1979 Massey Lectures, called it Haldane's principle and applied it to institutions: "Big animals are not big because they are complicated. Rather, they are complicated because they are big." Some things can only be done at a large size, she granted, but past some point the complication starts to interfere with the reason the organization exists, as anyone who has booked travel through a large university's administration has seen. Christopher Alexander's A Pattern Language (1977) used Haldane's Greek example to argue for independent regions, because the number of relationships inside a group grows much faster than the group and pushes it toward hierarchy. The post calls that growth exponential; pairwise links grow quadratically, which still makes the point.

Software has its own version. no-silver-bullet records Brooks's observation that a larger program adds different elements, not copies of the same ones, so its complexity grows faster than its size.

The wrong question

For open communities this replaces "can Mastodon be the next Twitter?" or "can Codeberg be the next GitHub?" with a different question: what would getting that big cost? Centralization and economies of scale help some things, like bulk hardware orders. They tend to break social processes. Moderation at scale has to be automated, which removes the human contact that makes a decision legitimate, strains the moderators and produces more wrong calls. He cites a 2026 preprint finding that growing Mastodon instances face moderation pressure much like Reddit's, which suggests that community size constrains moderation whatever the platform architecture. Two of his own projects point the same way: moderation policies co-designed with a small community of autistic people, and a data-collection study that worked because it was small, not despite it.

The same forces apply to politicians wishing for "a European Google". A company with Google's scale and monopoly would not be more ethical for being based in the EU, because many of the harms are built into a business model that needs the scale.

A better question

He suggests asking "how can others make their own?" Jacobs proposed federalism for governments, and he extends it to organizations in general: many reasonably sized alternatives that interoperate, instead of one replacement that becomes the next single point of failure. Even a Codeberg run democratically by its members would, as "the GitHub replacement", inherit the problems of being the one place, the way a second AWS us-east-1 would still take the internet down when it failed. The Fediverse, with all its problems, is his evidence that federation can work, and he thinks the people building federated code forges may manage the same for code hosting. tangled, a forge on atproto with self-hostable repository nodes, is one such attempt in the vault.

He does not claim scaling is never right. He asks that growth be a decision made after weighing its costs, and that a project which chooses to grow expect its current ways of working to stop fitting. Otherwise it ends up as Haldane's oversized gazelle, with broken legs.

Related pages

high-friction-commons reaches a similar trade from another direction, giving up scale on purpose so a commons stays in good faith; it comes from build-a-new-internet. what-even-are-microservices treats service boundaries as a tool for the organizational side of the same size problem inside one company.