ACLs are broken. Zero-trust is a PDF.
The problem: We copy data, then bolt on a permissions table, then pray no one leaks the UUID.
Cryptographic Set-Chemistry flips it:
Only those monads holding mutual .me refs + keys can derive the route.
For everyone else, the path doesn’t 403. It doesn’t exist.
Each added monad forges a fresh compound with its own resolution surface — private by construction, visible only to those holding the mutual keys.
The Math: I = (path, ciphertext, T, A, C)
Every island of state is:
T = Topology set: Where it exists → {office, iphone, backup}
A = Audience set: Who can interpret it → {me, wife}
C = Capability set: What you can do → {read, delete}
The axiom: Encryption doesn’t define where the data is. It defines who can collapse its meaning.
// Replicated everywhere
T = {office, iphone, backup}
// Decryptable by two identities
A = {me, wife}
// Result: widely distributed, narrowly perceived
Access isn’t checked. You either intersect with A or you get nothing.
Sharing = A ∪ {bob}
Revocation = A \ {bob}
Delegation = expand A
All enforced by cryptography, not policy.
Routes Are Chemical Compounds
You + Wife = New Compound
Route = A ∩ B ∩ Keys`
If you’re not in the intersection: no route, no data, no attack surface.
Pure emergent, sovereign topology.
No PDP. No PEP. No policy logs. The ciphertext is the boundary.
Why This Matters for the Agent Era
Agents need subspaces, not firehoses. friends[age > 18] should be resolvable only by callers in A.
In .me, the boundary is the data structure:
javascript
me.wallet["_"].secret = "seed"
// A = {self} by construction
// Not “403”. The path doesn’t exist for others
Bonus: O(K) Reactive
1M node graph. Change one sensor. 6 nodes recompute in 0.256ms.
Demo: Same medical canister = "biohazard" for surgeon context, "box" for loader context. Context is path. Meaning computes live.
You’re not modeling files or permissions.
You’re modeling: who can perceive what, where it exists, and what can be done.
MIT. TS/Python/Rust.
Code: https://github.com/neurons-me/monad
Docs: https://neurons-me.github.io/.me