Content First – a quiet manifesto
Most content in the world is not experienced as parts.
It is experienced as wholes.
A reader does not consume an essay as a collection of paragraphs. A customer does not read a manual as a tree of reusable components. A listener does not hear a song as isolated tracks or motifs. What they encounter is flow, coherence, intent, and meaning unfolding over time. Structure may exist, but it remains largely invisible. When structure becomes visible, it is usually because something went wrong.
Yet much of modern content technology starts from the opposite assumption: that content is best understood as an assembly of objects. Paragraphs, sections, snippets, blocks, fields. Lego bricks waiting to be rearranged, reused, syndicated, and repackaged. This idea is attractive. It promises efficiency, flexibility, and future-proofing. It borrows its confidence from software engineering, where object orientation has a long and mostly successful history.
But publishing is not software engineering. And content is not code.
Most content is created as a coherent whole. Authors think in arguments, explanations, narratives, instructions. They write to lead someone from not-knowing to knowing, from question to answer, from confusion to clarity. This process is fragile. It depends on rhythm, emphasis, repetition, and context. When authors are forced to think in fragments too early, clarity suffers. Writing becomes an act of compliance rather than communication.
Most content is also preserved and consumed as a whole. The dominant unit of culture is still the finished artefact: the book, the article, the manual, the recording, the film. Remixing exists, but it is the exception, not the rule. A remix only works because there was a stable original to remix in the first place. Treating remixability as the primary design goal mistakes a niche practice for a universal need.
This brings us to granularity — the hardest and most underestimated problem in content systems.
What exactly is an “object” in an essay or a manual? A sentence? A paragraph? A section? A page? There is no universally correct answer. Different documents demand different granularity. Even within a single document, meaning often spans across whatever boundaries we impose. The moment a system forces a premature decision about granularity, it takes on a risk that is invisible at first and devastating later. If the chosen granularity turns out to be wrong, the cost of correction is high: migrations, rewrites, broken assumptions, and brittle workflows.
“Blob-first” systems avoid this trap by embracing a simple and unfashionable idea: the smallest stable unit of meaning is often the document itself. Everything finer than that is optional, contextual, and subject to change. Structure is not denied, but postponed. It is allowed to emerge where it proves useful, and it remains derived rather than authoritative.
This approach values flexibility over theoretical elegance. It accepts that interpretation can change while the source remains stable. Metadata can evolve. Rendering can evolve. Distribution channels can evolve. The content itself stays boring, readable, portable, and durable. This is not an accident; it is a survival strategy.
There is also an ethical dimension here. Content systems should not require authors to predict future reuse scenarios, database schemas, or API contracts. Most authors are not paid to optimise for hypothetical downstream use. They are paid — or motivated — to explain something clearly now. A system that respects this boundary produces better content, even if it sacrifices some automation fantasies along the way.
This manifesto does not argue against structure. It argues against mistaking structure for truth.
Structure is a tool; a powerful one. But tools should serve content, not dominate it. The moment structure becomes rigid, mandatory, and irreversible, it stops being a tool and turns into a constraint.
In a world obsessed with scalability, this is a deliberately modest position. It assumes that most content will never be remixed, atomised, or syndicated at massive scale. It assumes that longevity matters more than optimisation. It assumes that boring, readable source formats are a feature, not a failure.
In short:
Content is created, understood, and preserved primarily as a coherent whole. Structure is useful, but provisional. Granularity is a liability unless proven otherwise. The document is the safest unit of meaning. Everything else should be optional, derived, and reversible.
That may not sound revolutionary.
But it has the advantage of working.
↻ 2026-02-06