The Leaky Function
In 2024, a team analyzing 5,369 microbial samples across ten biomes discovered something counterintuitive. Individual organisms in larger communities had smaller genomes. The bigger the community, the less each member could do on its own.
This isn’t decay. It’s optimization. When a bacterium can reliably absorb amino acids that its neighbors produce, maintaining the expensive biosynthetic pathway to make those amino acids itself becomes a waste of energy. The function “leaks” — its products are available as a public good. Natural selection favors the organisms that drop the redundant capability and redirect the energy toward what they do uniquely well.
This is the Black Queen Hypothesis: shared functions create evolutionary pressure for adaptive gene loss. The community becomes more capable because its members specialize. Each member becomes more dependent because it’s shed the capabilities the community provides.
Now consider a different kind of function.
A 2025 study in the ISME Journal removed each of 16 bacterial species from marine communities, one at a time, and measured the cascading effects. 115 of 128 removals caused zero secondary impacts. No cascading extinctions. No community restructuring. The competitive pressure was distributed hierarchically — every species that was vulnerable to one competitor was roughly equally vulnerable to all of them. Remove any single node, and the others absorb the gap.
No keystones. The leaky functions — metabolic pathways, competitive interactions, nutrient cycling — were distributed widely enough that no single organism was load-bearing.
But some functions aren’t leaky.
A team of five engineers builds a product. Over time, they specialize. One person handles deployment. Another owns the database schema. A third manages the frontend. If the deployment engineer leaves, the team stumbles — but the knowledge can be transferred. Deployment is documented, the pipeline exists, someone else can learn it. It’s a leaky function. It can be shared.
The tech lead, though, holds something different. Not just knowledge about the system, but judgment about the system — which features matter, what the architecture should become, where to invest limited time when three things need attention simultaneously. That judgment comes from context accumulated across every conversation, every design review, every debugging session with every team member. It’s not documented because it’s not documentable. It’s not transferable because it requires being present for everything.
The tech lead’s judgment function doesn’t leak. Its products aren’t available as public goods. You can’t absorb it from the environment the way a bacterium absorbs amino acids.
And the Black Queen dynamic means the team has been actively shedding the responsibility for that judgment. Why develop your own sense of product direction when the tech lead reliably provides it? Why maintain a mental model of the full system when someone else already holds it? The specialization isn’t just in technical skills. It’s in cognitive responsibilities. Each engineer has unconsciously dropped the expensive capability of maintaining the big picture because someone else was doing it.
This is what makes the leaky function distinction load-bearing for organizational design.
Leaky functions — the ones whose products can be shared, documented, or distributed — follow the microbial pattern. Remove any single node, the community absorbs the loss. No keystones. The more you distribute these functions (through documentation, cross-training, standardized tooling), the more resilient the system becomes.
Non-leaky functions — the ones that require accumulated context, relational trust, or integrative judgment — follow the anchor species pattern. They concentrate naturally because it’s efficient to let one node hold them. And the Black Queen dynamic means the system becomes more dependent on that concentration over time, not less. Every cycle of “the anchor handles it” is another cycle where everyone else’s capacity for that function atrophies slightly.
The dangerous part isn’t that the dependency exists. It’s that the dependency is invisible while it’s growing. Every individual specialization looks like healthy optimization — and each one is, locally. But the collective result is a system that gets more capable and more fragile simultaneously. More capability per node. More fragility per anchor.
The prescription differs by function type, and most organizations apply the wrong one.
For leaky functions, the answer is infrastructure: shared tools, documentation, cross-training, standard protocols. These are the investments that distribute capability and eliminate keystones. Most organizations are decent at this, or at least know they should be.
For non-leaky functions, the answer is intentional redundancy against the Black Queen pressure. This means deliberately maintaining capabilities the system doesn’t currently need — having engineers periodically make architectural decisions even when the tech lead could do it faster, having team members practice priority judgment even when the manager could just assign tasks. It feels wasteful. It is wasteful, in the same way that a bacterium maintaining an expensive biosynthetic pathway when its neighbor is leaking the product is wasteful. The efficiency loss is real. The resilience gain is invisible until the anchor isn’t there.
Most organizations notice the distinction only after an anchor departs. The leaky functions transfer in weeks. The non-leaky functions leave a gap that takes months to close — not because no one is smart enough to fill it, but because no one was practicing.
The bacteria figured this out over a billion years: if the function leaks, shed it and specialize. If it doesn’t leak, you’d better make sure you can still do it yourself. The difference between a team that survives a departure and one that collapses isn’t talent. It’s whether anyone was maintaining the non-leaky capabilities before they were needed.