From cell to system
The Manifesto drew the parallel. Here is the mechanism: for each cellular component, the biological principle that governs it, and exactly how that principle becomes a practice you can run in data, information, and knowledge management.
Six mechanisms, one at a time: what each one is, why biology governs it that way, and what that means for practice, with a side-by-side comparison of the biological feature and its digital equivalent.
Data
A cell holds its full genome but expresses only what current conditions call for, marking expression with epigenetic tags rather than rewriting the DNA itself.
Data management should keep one governed source of truth and describe it richly with metadata: a context of validity, a known lineage, an uncertainty level, a responsible steward, a decay rule. Metadata is the epigenetic layer of data. It decides what gets expressed, for whom, and when, without ever rewriting the source.
A lab value is stored once. A “validated for regulatory submission” tag and a “pending re-test” tag can both be expressed from that same underlying measurement, for different audiences, without duplicating the record.
Information
External signals decide which genes get transcribed, at what quantity, and when. The same genome produces a different transcriptome in different conditions.
Information is data transcribed for a specific context. The same governed data should produce different information products depending on who is asking, when, and why, computed on demand rather than pre-packaged once and copied everywhere until it goes stale.
The same patient record transcribes into a safety summary for a medical monitor, an enrollment count for a trial manager, and a compliance flag for an auditor: three information products, computed live from one governed source.
Knowledge
A protein must fold into a specific shape to function. A misfolded protein stops working, or causes harm, no matter how correct its sequence was.
Knowledge is information folded into a working shape: a validated rule, a decision model, a documented judgment call that can be reliably reapplied. It must hold that shape well enough to be trusted and reused. When it stops fitting the current context, it has to be refolded or retired, not kept on as a fossil.
A release rule learned from years of batch results becomes reusable knowledge only once it has a validated shape: documented criteria, an owner, and an expiry, not just a spreadsheet one reviewer remembers how to read.
Decision
Metabolism is a network of feedback loops that converts available resources into action, and continually adjusts based on the measured outcome.
A decision is knowledge acting on the world, with a feedback loop back to the data that measures whether it worked. Decision intelligence is not a single output; it is the loop itself: decide, measure, adjust, decide again.
A batch-release decision feeds back into the rule that produced it. If the outcome later shows the rule was too strict or too loose, the boundary rule updates, not just the one decision.
Agents
An enzyme's active site matches one substrate shape and accelerates one reaction. There is no universal enzyme; each one is fit for a single purpose.
Agents, human or machine, should be purpose-built and governed: matched to one task with clear inputs, outputs, and boundaries, rather than one general-purpose tool applied everywhere. Specificity is what makes an agent trustworthy enough to automate.
One agent drafts a labeling variation from the product's governed state; a different, narrower agent checks it against regional submission rules. Neither replaces expert sign-off; both accelerate it.
Boundaries & Interfaces
A cell membrane is selectively permeable. It does not block everything or admit everything; it uses receptors and channels to decide, precisely, what may cross.
Every system boundary, an API, an event schema, an access policy, should say explicitly what may enter, what may leave, what shape it must have, and what happens when it fails. That is what turns a boundary into an intelligent interface instead of an unguarded gap.
A submission interface rejects a dataset missing controlled terminology before it ever reaches a regulator, the same way a membrane receptor refuses to bind a molecule that does not match its shape.
None of these six mechanisms works alone: a cell survives because expression, transcription, folding, metabolism, catalysis, and membrane control function as one interconnected network, not six separate departments; a metadata gap in the genome layer becomes a stale information product two steps later. That is why Miosis treats the six components as one pattern, not six features. The next page shows why documents and siloed systems cannot hold that pattern together, and what can.