The data size alone is a huge reason not to use it. In Dendrite, I reduced its memory requirement by more than half by switching to bitfields for the lattice state changes (that's the trigrams thing)....
You’re arguing against using EC as a graph container.
That’s fair — but that’s not what I’m doing.
If you need:
explicit neighbor lookup
explicit traversal
explicit constraint scans
then yes — integer lattices are optimal. Full stop.
EC is not a faster adjacency list.
It’s not a cheaper graph walker.
And I would never suggest using scalar multiplication to simulate neighbor iteration. That would be absurd.
You’re correct there.
---
Where we differ is architectural.
Your lattice stores:
nodes
edges
traversal logic
The connectivity is the data.
ECAI does not encode edges and then try to “recover adjacency.”
It encodes state as algebraic commitment and performs deterministic projection, not neighborhood walking.
If your system fundamentally requires:
> given X, enumerate neighbors(X)
then EC gives you nothing.
If instead your system is:
> given state S, recover compatible structure deterministically
then algebraic encoding replaces traversal entirely.
Different computational model.
---
On the 128KB vs 1.5MB point:
You’re absolutely right that 128KB of points is just identities.
If the problem is graph simulation, that comparison is meaningless.
If the problem is state commitment + deterministic recovery, then edges don’t need to be stored at all.
That’s the category boundary.
---
On dimensionality:
An elliptic curve is 1-dimensional as an algebraic variety, yes.
But its group operates over a 256-bit field.
Dimensionality of constraints ≠ geometric dimension of the curve.
Still — if your model is inherently 8 independent constraint axes that require explicit local adjacency reasoning, a lattice is cleaner.
---
So here’s the clean split:
If you are building:
a deterministic integer lattice
that walks neighborhoods
with explicit constraint ratios
Then integer arithmetic is exactly right.
If you are building:
a compressed, algebraically-addressable state space
where structure is retrieved rather than traversed
Then EC becomes relevant.
These are different machines.
One simulates topology. One encodes structure.
No one is “smacking down” anything — they’re solving different problems.
And yes — integer fixed-point math for deterministic ratio analysis is solid engineering.
That’s fair — but that’s not what I’m doing.
If you need:
explicit neighbor lookup
explicit traversal
explicit constraint scans
then yes — integer lattices are optimal. Full stop.
EC is not a faster adjacency list.
It’s not a cheaper graph walker.
And I would never suggest using scalar multiplication to simulate neighbor iteration. That would be absurd.
You’re correct there.
---
Where we differ is architectural.
Your lattice stores:
nodes
edges
traversal logic
The connectivity is the data.
ECAI does not encode edges and then try to “recover adjacency.”
It encodes state as algebraic commitment and performs deterministic projection, not neighborhood walking.
If your system fundamentally requires:
> given X, enumerate neighbors(X)
then EC gives you nothing.
If instead your system is:
> given state S, recover compatible structure deterministically
then algebraic encoding replaces traversal entirely.
Different computational model.
---
On the 128KB vs 1.5MB point:
You’re absolutely right that 128KB of points is just identities.
If the problem is graph simulation, that comparison is meaningless.
If the problem is state commitment + deterministic recovery, then edges don’t need to be stored at all.
That’s the category boundary.
---
On dimensionality:
An elliptic curve is 1-dimensional as an algebraic variety, yes.
But its group operates over a 256-bit field.
Dimensionality of constraints ≠ geometric dimension of the curve.
Still — if your model is inherently 8 independent constraint axes that require explicit local adjacency reasoning, a lattice is cleaner.
---
So here’s the clean split:
If you are building:
a deterministic integer lattice
that walks neighborhoods
with explicit constraint ratios
Then integer arithmetic is exactly right.
If you are building:
a compressed, algebraically-addressable state space
where structure is retrieved rather than traversed
Then EC becomes relevant.
These are different machines.
One simulates topology. One encodes structure.
No one is “smacking down” anything — they’re solving different problems.
And yes — integer fixed-point math for deterministic ratio analysis is solid engineering.
1