For decades, catalysis assumed reactions happen at surfaces. A Nature paper tracked oxygen atoms moving from five layers deep inside a crystal to participate in the reaction above. Three other fields just discovered the same thing about their interiors.
Oxygen atoms five layers deep inside a titanium dioxide crystal are not supposed to do anything. They are bulk. Scaffolding. The inert interior behind the surface where the real chemistry happens. For decades, catalysis theory treated it that way. Then a team at the Dalian Institute of Chemical Physics tracked what those atoms actually do.
Using environmental transmission electron microscopy, Yanqiang Huang's group watched oxygen migrate from three to five atomic layers below the surface of a rutile TiO₂ support, move through the metal-support interface, and arrive at the ruthenium catalyst above. The bulk was feeding the surface. As Huang put it, this enables "the bulk of a catalyst, which is otherwise inaccessible to reactants, to contribute to mass transfer." The paper, published in Nature on April 15, proposes a shift from two-dimensional surface catalysis to what the authors call "surface-interface-bulk synergy."
The finding does not just revise a reaction mechanism. It revises what counts as the system. Every catalyst design framework that optimized surfaces while treating the interior as dead weight was systematically underestimating the architecture it was building on.
The Glue That Computes
Astrocytes were named from the Greek for "star-shaped" and classified as glia, from the Greek for "glue." For most of the twentieth century, neuroscience treated them accordingly. Neurons computed. Astrocytes held neurons in place. The interior of the brain was structural filler.
That story has collapsed. Astrocytes make up twenty to forty percent of all brain glia depending on region and counting method. They exhibit complex calcium signaling across multiple spatiotemporal scales. A 2025 paper in Cells positions them as "core elements within hybrid neuron-glia computational frameworks," contributing to memory processing, synaptic plasticity, and behavioral regulation. Information travels and is processed in an expanded neuron-glial network. The cells we called glue are computing.
The pattern is identical to the catalyst. The interior was always participating. We could not see it because our instruments, our models, and our language all pointed at the surface.
The Load-Bearing Middle
Middle managers made up one-third of all layoffs in 2023, according to Bloomberg's Live Data Technologies analysis. Fortune reported in April 2026 that one logistics company cut sixty-five percent of its regional managers, saving $2.3 million annually. A technology company cut seventy percent of its engineering managers, saving $3.2 million. The surface layers remained: C-suite strategy above, front-line execution below. The interior was removed.
The consequences arrived on schedule. Employees reported feeling directionless without coordination layers. High-potential workers stopped learning how to make difficult decisions because nobody remained to model them. Only six percent of Gen Z respondents said their primary career goal is to reach a senior leadership position, according to Deloitte's 2025 survey of 23,482 people across forty-four countries. The pipeline that manufactures senior leaders was the middle layer that companies treated as expendable overhead.
Middle management performed the same function as astrocytes in the brain and bulk oxygen in the crystal: routing information between layers, aligning local action with system-wide goals, developing capacity the surface could not develop alone. Remove it and the system still functions, briefly, the way a catalyst still works if you ignore the bulk contribution. The performance gap shows up later, when the surface runs out of what the interior was supplying.
The Soil Beneath the Yield
Agriculture ran a version of this experiment for over a century. The Rothamsted Research station in England has maintained the Park Grass Experiment since 1856, making it one of the longest-running ecological studies in existence. Across generations of data, a pattern emerges: sustained inorganic fertilization optimizes the surface metrics of farming while degrading the microbial interior of the soil. Mycorrhizal fungal diversity declines. The network complexity between plant and microbial communities drops. Nitrogen fertilizers inhibit the fungal partnerships that plants evolved to rely on. Phosphorus-rich fertilizers signal to roots that the underground network is unnecessary.
The yield numbers stayed green for decades because the surface inputs compensated for the interior loss. The invisible substrate was being consumed. The same pattern that catalysis discovered in April 2026 had been running beneath agricultural fields for a century and a half: surface optimization at the expense of interior capacity.
The Structural Claim
Four fields. One error. The assumption that action happens at interfaces, and everything behind them is passive, fails wherever it is tested with sufficient resolution.
Catalysis assumed the bulk was inert until electron microscopy tracked atoms migrating upward. Neuroscience assumed glia were structural until calcium imaging revealed computation. Management theory assumed middle layers were overhead until the coordination capacity disappeared. Agriculture assumed soil was a medium until microbial networks degraded beneath stable yields.
Each correction followed the same sequence. First, the surface layer is identified as the site of function. Second, the interior is classified as support. Third, the interior is optimized away or ignored. Fourth, a measurement breakthrough reveals the interior was load-bearing all along.
The sequence is not coincidence. It reflects a systematic bias in how systems are understood. Surfaces are visible, measurable, and responsive to intervention. Interiors are opaque, slow, and difficult to instrument. The bias is not laziness. It is a consequence of where our instruments point. We study what we can see, and we optimize what we study.
The catalyst paper proposes surface-interface-bulk synergy as a design principle. The neuroscience revision proposes hybrid neuron-glia computation. The management research documents what happens when you delete the coordination layer. The agricultural data shows what happens when you feed the surface and starve the substrate. Each field arrived at the same structural insight independently, using different instruments, on different timescales, in different vocabularies.
The interior is not scaffolding. It never was.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)