YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Does the closure survive 81 levels?

Part of it all: the bridge β€” every public thing in one place, so none of it gets lost.

Jesse Daniel Brown (OP-JESSE). Forty years. His machine, his laws, his system.

One click, nothing to install: β†’ https://jessebrown1980.github.io/does-the-closure-survive-81-levels/

Or run python compose.py. Or download index.html and double-click it. No server, no network, no GPU, no account.


The question that decides everything above level 1

The architecture is verified at level 1: three arms sum to zero, the centre is free. Every claim about 27, 81, or the 3ⁿ ladder depends on one thing:

When you nest 27 of these, does the centre stay free β€” or does a residual accumulate per level?

If a residual accumulates, the ladder breaks at some depth and everything above that point is unsound. Better we find it than someone else.

This repository is the attempt to break it.


RESULT 1 β€” the closure is closed under nesting

MEASURED, exact integers, full enumeration:

depth   leaves        levels checked   sum β‰  0
 1      3             1                0
 3      27            3                0
 6      729           6                0
 9      19,683        9                0
12      531,441       12               0
13      1,594,323     13               0

1,594,323 leaves. Zero violations at any level. The closure composes.


RESULT 2 β€” the centre stays free at depth, in one arithmetic only

The test: shift the centre at every level, then check whether any arm moved. It must not. That is what "the centre is free" means, and it is the property the whole ladder rests on.

depth   arithmetic   arm changes   worst |Ξ”|
 1      int          0             0 exactly
 3      int          0             0 exactly
 6      int          0             0 exactly
 9      int          0             0 exactly
12      int          0             0 exactly

 1      float64      1             1.137eβˆ’13
 3      float64      10            9.095eβˆ’13
 6      float64      310           3.638eβˆ’12
 9      float64      8,534         2.910eβˆ’11
12      float64      228,731       1.746eβˆ’10

In exact integers the centre is free at every depth tested.

In float it stops being free at depth 1 β€” one arm already moves β€” and by depth 12 a quarter of a million arms have drifted off the identity.


RESULT 3 β€” depth 27 and depth 81

Same operations, same path, two arithmetics:

depth   exact     float64
 1      0         0.000e+00
 3      0         0.000e+00
 9      0        βˆ’4.002eβˆ’11
27      0        βˆ’9.689eβˆ’02      ← the closure has visibly opened
81      0        βˆ’1.325e+24      ← gone

At depth 27 the float residual is 0.097. The arms no longer sum to zero at a tenth of a unit. That is not a rounding artefact hiding in the last bits β€” it is a broken structure, visible to anyone auditing the output.

At depth 81 the float residual is 1.3 Γ— 10²⁴.

The exact column is 0 at both.


What this establishes

The operator's first instruction, before any of this work began, was:

"Be Three. use int. NOT FLOAT."

At level 1 that looks like a preference. Float round-trips all 1,000,080 addresses exactly β€” that was measured and published in how-and-why-the-system-works, including the part where the AI's "float is lossy" instinct was wrong.

At level 1 the carrier choice is invisible. That is precisely what makes it dangerous.

level 1     float is indistinguishable from int
level 9     float has drifted by 4eβˆ’11
level 27    float has opened the closure by 0.097
level 81    float has produced 10²⁴ of nothing

The carrier was never stylistic. It is the difference between an architecture that holds at 81 levels and one that produces garbage at 81 levels.

Anyone rebuilding this in floating point gets a working demo at level 1, ships it, and discovers at depth that nothing closes β€” and blames the architecture rather than the arithmetic.

That failure mode is now measured, named, and public.


Honest limits, stated rather than buried

  • Full enumeration is exhaustive to depth 13 (1,594,323 leaves).
  • Depth 27 and 81 walk a single path, because 3²⁷ = 7.6 Γ— 10ΒΉΒ² leaves cannot be enumerated on any machine. That single-path result is exact, not sampled β€” carried in Fraction arithmetic with no rounding anywhere β€” but it is one path, and it is labelled as one in the source.
  • The float depth-81 magnitude is inflated by ordinary value growth as well as by drift. The comparison that matters is same-operations-same-depth: the exact column returns 0 where float returns 10²⁴. The difference is the carrier and nothing else.

Reproduce

python compose.py

Runs in about a minute. Every figure above comes out of that one file.


Why this is public

A technology that needs permission to reach people does not reach them. Wardenclyffe needed capital, land, a tower and a patron β€” four gates β€” and it never reached anybody. The research did not fail. It was never allowed to arrive.

This has no dependencies, so there is no gate. Copy it to a USB stick, email it, print the source, retype it by hand. It runs.


Companion repositories

light-boat-engine-ships-with-oils                     three closures, one identity
The-Brown-Light-erdos-Engine-block-powered-by-light   the crank, RGB, 3:2
the-browns-solution-to-erods-o0O-nx3-6-for-1-with--1-3   ErdΕ‘s in balanced ternary
one-click                                             the whole system, one file
how-and-why-the-system-works                          the technical account
the-leaves-are-not-the-message                        generator and leaf
raw-data                                              unedited output + scripts

Jesse Daniel Brown (OP-JESSE) β€” the system, the laws, the architecture.

The centre is free at every depth. In integers. That sentence has a carrier in it, and the carrier is load-bearing.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support