Local supplement, 12 November 2026. Not a Zenodo deposit and not Paper X.
Power-form reconstruction
The recorded equation is . A two-byte seed selects and a Python source that prints it. This is not the floor-map packet.
Seed packets and shape to record reconstruct the rational floor map . This page does not. Its seed 399 produces and a 68-byte print source from the two bytes 8f 01. The packet sizes, the 2,520 comparisons, and the 36 program runs on the seed-packet page do not apply here.
Assumptions and the real inverse
Let , , and be fixed reals with and . The general inverse keeps all three symbolic. The reference note uses , , and . An earlier form used . Both meet the nonzero assumption. The symbol in this equation is the constant in the denominator. It is not a new definition of the scalar on the Ω₀ page.
Theorem 1 in the PDF says is a bijection, with
The argument is that is strictly increasing and onto , so both compositions are identities. That is a written proof. The hosted Lean file does not contain it.
The seed family
For a nonnegative integer ,
Theorem 2 says , and the inverse recovers from that power. The inner quotient is exactly , so the cancellation uses only .
Under the reference specialization, Corollary 2.1 puts every seed state in the free rank-4 module
The integer coordinates are , for . The PDF's uniqueness argument uses , the irrationality of , and the transcendence of . Transcendence is cited from Bernard, Bertot, Rideau, and Strub, arXiv:1512.02791. It is not reproved here, and it is not a Lean theorem in the hosted file. The full ring also contains . The rank-4 statement is about , not about that larger ring.
Coordinate maps
On coordinates ,
These are the coordinate forms of and . The PDF says each 2-by-2 block of multiplication by has determinant , so the four-by-four map has determinant 1 and lies in . Applying to the seed coordinates gives , the coordinates of . Applying returns the seed coordinates. Both directions are integer identities on every quadruple. The Lean file states those integer identities. It does not state the real inverse.
Seed 399
, coordinates . The numerator image is . The power is the 60-digit integer
664318888400147317962445992095559074642829054227185303529199
The two-byte payload is 8f 01, because . The shared recipe writes
print(664318888400147317962445992095559074642829054227185303529199)
Those are 6 bytes for print(, 60 digit bytes, one byte for ), and one LF byte: 68 bytes. This archive executed that source. The printed line was the integer above.
Reconstruction
Restrict and encode as two unsigned little-endian bytes. Proposition 4.1 says the decoder is Euclidean division by 256, so it returns , and a fixed recipe then returns . For this family, is the print source. Versioning, recipe selection, framing, and the first distribution of the recipe are extra protocol data. The comparison of two payload bytes with 68 source bytes does not include that setup.
The map is injective on nonnegative integers, and each source contains the full decimal. The 32,768 seeds therefore give 32,768 distinct byte strings. A fixed field that selects all of them needs at least 15 bits. This field uses 16. Across messages, a saving is , with receiver computation added. That is a cost comparison for a shared recipe. It is not universal compression, not a compression advantage caused by , and not a geometric encoder without an explicit map.
| Seed | Payload | Integer bytes | Source bytes |
|---|---|---|---|
| 0 | 00 00 | 1 | 9 |
| 1 | 01 00 | 1 | 9 |
| 2 | 02 00 | 3 | 15 |
| 399 | 8f 01 | 25 | 68 |
| 32,767 | ff 7f | 44 | 112 |
Integer bytes are the shortest nonempty unsigned little-endian encoding. Zero is one byte,00. The power of 2 is 8,388,608. The power of 399 has 60 decimal digits. The power of 32,767 has 104.
The power and the affine iterate
With and , the PDF records , , and . Theorem 3 says these two maps agree exactly at , which is the seed . The numerical pair at , about 0.0137752 and 0.999999999992, is an illustration in the PDF. It was not recomputed here.
What was run here
The three listings were copied out of the PDF and saved as verify_power.py, verify_power.mjs, and PowerProof.lean. The command in the PDF is python3 verify_power.py.
Python 3.10.21 and Python 3.11.2 each ran that command with Node v22.23.2. Both exited 0, stderr was empty, and the stdout files were the same 435 bytes, SHA-256 92b2dc2c2a8582d9e4d3f1be8a562531462b23c07f0c0c3b854b305ddb193ba7. Each run reported 32,768 roots, 32,768 module states, and 32,768 matching Python and Node source records. The printed digests are the ones in the PDF:
Program-record corpus dcce7387a3a2a42122049524ec06ec6ea3b113bd20ef99db259cee0a141ce29f. Length-prefixed integer-power bytes a4b81c948a57d269acf8263d20d090fcc39725ab160ee9dcfe24a37634aa3c88.
The same command under python3 -O also printed those lines. The ten checks are assert statements, and -O removes them, including the comparison with the Node bytes. The hash updates and the print statements remain. The -O transcript is therefore not a second execution of the checks.
Lean, limited to the integer file
PowerProof.lean has no explicit imports. A text scan finds seven theorems and no sorry, admit, or declared axiom. The command was lean PowerProof.lean with Lean 4.19.0, x86_64-unknown-linux-gnu, commit 6caaee842e94, Release. That binary was the upstream release tarball. It is not hosted. The command exited 0 and stderr was empty. The stdout is 401 bytes, SHA-256 ba12d7f41e8ac49dc3b165a246bade2582b934576e39d8cb97b02a10a2aae2e0, saved as lean-stdout.txt.
power_399 depends on no axioms. power_23_injective and two_byte_identity depend on propext. The four coordinate theorems depend on propext and Quot.sound. Those theorems are integer-power injectivity, base-256 reconstruction, the seed-399 power, both directions of the integer coordinate map, and cancellation and recovery for every integer seed. Theorem 1, Theorem 3, and the freeness of are not in the file. This run does not make the real inverse Lean-verified.
Not claimed
- That the real inverse or the freeness argument was checked by Lean. The kernel run covers only the integer file.
- A written-proof review, peer review, or independent reproduction.
- A DOI, a Zenodo deposit, or a tenth source work. Nine deposits remain nine.
- Universal compression, a compression effect caused by Ω₀, or a geometric encoder without the maps above.
- A result about P and NP. The PDF says the integer-root search uses 15 binary-search steps on this domain and does not separate those classes.
- A change to frozen SRSD. SRSD stays the external audit instrument.
The revised draft is OMEGA_NAUGHT_Power_Form_Reconstruction_Proof.pdf, 152,508 bytes, SHA-256 42f2a9f307fc1d04a24933e889ec569075745896644f9605b6caf5795f9fd755. This archive's log is stdout.txt.