Genesis

Package Boundary

No connected PyPI package is claimed for this repository. The public genesis package name on PyPI is unrelated; use the source commands below. Use the repo-local source, dossier, or proof commands below; do not substitute an unrelated PyPI package.

For full install, smoke, source, and developer commands, click here.

---

00 · GENESIS · DETERMINISTIC COMPUTATION RESEARCH-READY · SOURCE OPEN Meru — geometric computation. Two independent builds, 31,560 matching hashes, one canonical SHA · Genesis · github.com/Zer0pa/Genesis A computational result is only as real as the next person's ability to reproduce it. Genesis takes a non-trivial Rust pipeline over a 285-vertex geometric substrate, compiles it twice on different machines — one Snapdragon 8 Elite, one Apple M1 — and shows that every one of 31,560 Phase 0 hashes agrees on the same canonical signature. 74 evidence cells pass with zero divergences. Three of four pre-registered comparisons are settled; one symmetry test is held to v2.0.

Scope: Rust pipeline over 285 vertices. Snapdragon and M1 builds agree on 31,560 Phase 0 hashes; one symmetry test is held to v2.0.

01 · THE GAP NO RECEIPT Build the same code twice independently. No receipt proves both got the same answer.

02 · MARKETS ADJACENT FORECASTS Simulation software ’30 $32.8B Edge AI software ’30 $66.5B Digital engineering ’30 $74.2B Verification software ’30 $15.7B Digital trust tooling ’30 $41.8B Five neighbouring markets all assume one floor that almost nobody checks: rebuild a computation and get the identical bytes back.

03 · VALUE OF MARKET $74.2 B The 2030 digital-engineering market — built on a rebuild sameness nobody has been counting.

04 · INSIGHT Build it twice. Get the same answer. That's the proof.

05.1 · CURRENT TECH ASSUMED, NOT PROVEN Containers, seeded randomness, floating-point discipline, and tolerance thresholds reduce drift between two builds of the same scientific code. They do not produce a public, counted answer to the plain question: did the bytes match?

05.2 · OUR TECH TWO BUILDS, ONE SHA Genesis runs a pure-rational Rust pipeline — every math-path value a BigRational, no floating-point — over a 285-vertex D6 substrate. The same source is compiled twice: once on a Snapdragon 8 Elite, once on an Apple M1. Both builds produce 31,560 Phase 0 hashes that agree on a single canonical signature. Across 74 evidence cells, zero divergences.

05.3 · BENCHMARKS V1.0 EVIDENCE CHAIN Cells74/74PASS Hashes31,560Phase 0 Divergence0cells DM33/4settled Evidence cells PASS DM3 comparisons 3/4 Z2 DEFERRED Scope: v1.0 only · no PyPI distribution yet · the D6-vs-C3 Z2 symmetry check is held to v2.0.

06 · MEASUREMENT TWO-BUILD HASH AGREEMENT 74 evidence cells. Two builds. 31,560 hashes agree.

06.1 · COMPARATIVE VALIDATION · PHASE 0 HASH AGREEMENT M1 build SHA MATCH RM10 build SHA MATCH solve_h2.json MATCH K2 S30 MATCH Apple M1 and Snapdragon 8 Elite return one canonical signature across 31,560 hashes. solve_h2.json and the K2 step-30 trajectory match byte-for-byte. Three of four pre-registered comparisons are settled; the D6-vs-C3 symmetry observable is deferred to v2.0.

07 · KEY METRICS MEASURED RESULTS

07.1 · 74 EVIDENCE CELLS 74/74 Evidence cells closed · zero divergences across both builds

07.2 · PHASE-0 HASHES 31,560 Identical hashes on two independent builds · byte for byte

07.3 · CANONICAL SHA 1 One canonical signature · across all 31,560 hashes

07.4 · PRE-REG COMPARISONS 3/4 Pre-registered comparisons settled · the fourth held to v2.0

07.5 · K2 PARITY S30OK One step depth agrees across builds · the next three still pending

08 · DETERMINISM BYTE-IDENTICAL ACROSS BUILDS Two builds. One SHA. 31,560 hashes agree.

08.1 · WHAT DETERMINISTIC MEANS PER-BUILD, PER-CELL Two independent builds of the same Rust pipeline — Snapdragon 8 Elite and Apple M1 — run against the same 285-vertex D6 substrate and return the same verify.json signature across 31,560 Phase 0 hashes. Cross-build parity is confirmed for solve_h2.json and for K2 at step 30. Every math-path value is BigRational; no floating-point is involved on the math path. Byte-exactness is verified per cell and per build against declared canonical hashes. Genesis does not claim all architectures, all step depths, or v2.0 symmetry parity, and says so explicitly.

08.2 · THE FIDELITY GAP Honest Blocker · Bounded to v1.0 receipts. No PyPI package: the existing genesis name is unrelated, zer0pa-genesis is absent. The Cargo/Rust source is not yet substantiated in public form. The v1.0.0 tag has no Release object. The D6-vs-C3 Z2 symmetry observable and step-20/40/50 host widening remain open.

09 WHEN A COMPUTATION KEEPS ITS IDENTITY.

09.1 · THE AMBITION A computational result should mean the same thing on the second machine, the next year, and in the next reviewer's hands. Genesis narrows scope until that sameness is countable rather than promised, so downstream science and engineering can finally point to one stable object instead of an approximate rerun.

09.2 · WHAT WORKS NOW What holds today: 31,560 Phase 0 hashes match across two independent builds, with zero divergences across 74 evidence cells.

09.3 · WHAT'S STILL OPEN Still open: published Cargo source, the D6-vs-C3 symmetry test, step-20/40/50 host widening, and a PyPI distribution.

09.4 · BUILD VARIANCE · NEAR-TERM (12–24 MO) Rebuild variance gets a counted number A simulation team that today writes “results are reproducible to within tolerance” gains a sharper line to draw. When 31,560 hashes either match or do not, internal review meetings about rebuild drift stop being narrative and start being a count.

09.5 · CROSS-PLATFORM · NEAR-TERM (12–24 MO) “It worked on my machine” stops winning A scientific collaboration that hands a pipeline between an Apple M1 laptop and an Android-class ARM handset can now demand the same canonical answer from both. Platform difference stops being an automatic excuse for unexplained divergence between collaborators.

09.6 · SCIENTIFIC CHAINS · MID-TERM (24–48 MO) Computational papers gain a stable referent A reviewer assessing a computational claim can be pointed at a retained hash chain rather than a re-run that approximates the original. The paper, the figure, and the underlying computation share one identity that later experiments can attach to.

09.7 · PROCUREMENT · MID-TERM (24–48 MO) Engineering reviews compare artifacts, not summaries A release-review board signing off on a regulated computational deliverable can compare retained hashes between candidate builds instead of reading prose. Deferred questions stay disclosed alongside the receipt; identity becomes checkable without trusting a vendor narrative.

09.8 · COMPUTATION IDENTITY · PARADIGM (48 MO+) A computation earns a stable identity When the same non-trivial computation can be rebuilt on independent hardware years apart and return identical bytes, regulators, courts, and scientific reviewers gain something they have never had: a computational result they can point to the way they point to a sample on a scale.

---

Install / Developer Commands Detailed

Package Boundary

No connected PyPI package is claimed for this repository. The public genesis package name on PyPI is unrelated; use the source commands below. Use the repo-local source, dossier, or proof commands below; do not substitute an unrelated PyPI package.

### 1. Cross-compile snic_rust for RM10 (aarch64-linux-android24)
export NDK=/usr/local/share/android-ndk
export CARGO_TARGET_AARCH64_LINUX_ANDROID_LINKER=$NDK/toolchains/llvm/prebuilt/darwin-x86_64/bin/aarch64-linux-android24-clang
export CC_aarch64_linux_android=$CARGO_TARGET_AARCH64_LINUX_ANDROID_LINKER
export AR_aarch64_linux_android=$NDK/toolchains/llvm/prebuilt/darwin-x86_64/bin/llvm-ar
cargo build --release --target aarch64-linux-android -p io_cli --bin snic_rust

### 2. Deploy to RM10 (cpu0–cpu5; cpu7 is dm3_runner's; cpu6 thermal margin)
adb -s FY25013101C8 push target/aarch64-linux-android/release/snic_rust /data/local/tmp/genesis/
adb -s FY25013101C8 push harness/phone/*.sh /data/local/tmp/genesis/harness/
adb -s FY25013101C8 shell chmod +x /data/local/tmp/genesis/snic_rust /data/local/tmp/genesis/harness/*.sh

### 3. Smoke-test bit-determinism (single core)
adb -s FY25013101C8 shell taskset 0x01 /data/local/tmp/genesis/harness/run_genesis_cell.sh BITDET_SMOKE

### 4. Verify the canonical hash
adb -s FY25013101C8 shell sha256sum /data/local/tmp/genesis/cells/BITDET_SMOKE/verify.json
### Expected: 97bd7d... (source-hardcoded CANONICAL_VERIFY_HASH)

### 5. Launch the autonomous chain (resume-safe; dm3_runner on cpu7 is untouched)
adb -s FY25013101C8 shell /data/local/tmp/genesis/harness/resume_chain.sh