What Persists?

TL;DR

Invariants survive component replacement — find what stays true across swaps.
Constraints outlast strategies — physics, law, and logic bound all plans.
Agreements persist when enforcement and benefit align — otherwise they erode.

How to use this lens

List components that could be replaced — people, servers, cells, tools.
Ask what must remain true for the system to still be “the same thing.”
Separate hard constraints from soft conventions — which break the system if violated?
Find identity keys — IDs, schemas, DNA, constitution, brand promise.
Check persistence mechanisms — replication, backups, rituals, audits.
Note what you want to persist vs what accidentally does — they differ.

Core concepts

Takeaway: Persistence enables coordination at scale.

If every fact had to be renegotiated daily, nothing complex could run. Durable state — whether in CRDTs, databases, or constitutions — lets strangers cooperate with low friction.

Takeaway: Two kinds of persistence.

Physical persistence: atoms in a wall. Logical persistence: account balance after server migration. Distributed systems teach that logical persistence is engineered, not guaranteed by hardware.

Takeaway: Invariants simplify reasoning.

When you know “total money conserved” or “schema version monotonic,” you can reject huge classes of bugs. Invariants are compression of experience into rules.

Persistence vs change

Persists	Changes freely
API contract major version	Implementation internals
Company mission (ideally)	Tactics
Blood type	Heart rate
CRDT merge semantics	Replica location

Mini-examples

Sleep: circadian rhythm persists across nights; sleep need resets but baseline need persists across life stages.
Office: legal entity persists through employee turnover; shared vocabulary persists in culture.
Blood: blood type invariant; vessel topology persists while individual cells replace constantly.

Curriculum bridge

Formal persistence tools:

CRDTs + Probability — merge without coordination
Consensus Systems — agree on a single ordered history

Try it now

Name one thing about your team or body that ** stayed the same** through five replacements of parts (people, cells, laptops). That is your invariant candidate.

Going deeper

Takeaway: Persistence trades flexibility.

Strong invariants speed reasoning but resist change. Constitution-level rules need amendment processes — meta-persistence for the rules themselves.

Takeaway: Entropy erodes soft persistence.

Without maintenance, agreements fade. Documentation rots; habits decay without sleep. Persistence requires energy input — the second law wearing a business suit.

Takeaway: Identity under replacement is a design choice.

The Ship of Theseus is not a paradox — it is an engineering question. What ID do you assign the ship after each plank swap? Systems answer with primary keys.

Knowing what persists tells you what you can safely refactor — and what you must never break silently.