dingoandClaude Opus 5 b4c3da56e9 Four of five runs conserve perfectly; three are ecologically worthless
M4.5 closes M4 by measuring what aggregation costs, and adds no library code
except one defect fix the measurement found.

One rule is written once and applied at both resolutions. The control is what
makes the rest mean anything: with one member per class a class average is that
member, so the two runs must agree exactly, and they do, tick by tick, in every
channel. A deliberate one-unit drift in the aggregate rule is caught by it, so the
instrument is verified rather than assumed. Without it a measured gap would be
indistinguishable from a transcription mistake in one of the two runs.

The result is reported, not smoothed into a tolerance. At class widths ten and
fifty the aggregate population survives ten thousand ticks within 375 parts per
thousand on population and 43 on energy. At twenty-five, one hundred and two
hundred it dies out entirely while the individual run sustains seventy-five
members: a false extinction caused by resolution alone, because a whole class
starves the moment its average cannot pay upkeep, where individuals starve a few
at a time and the survivors keep the shared per-capita share up.

The collapse is not monotone in class width. Twenty-five destroys the population
and fifty sustains it. An earlier version of the test asserted that coarser is
never better; the measurement refuted it, so the assertion went rather than the
widths being chosen to make it hold. The non-monotonicity is asserted now, because
it is surprising and because the warning written around it could otherwise become
wrong without failing. A tier policy cannot assume degradation worsens smoothly as
classes widen, and M4.6 inherits that.

Conservation was exact at every tick on both sides of all five runs, including the
three that died out. That is the clearest available demonstration of the thing
ECOLOGY_MODEL keeps saying: conservation is not ecological evidence.

Two flaws in my own experiment were found before any number was believed. The
aggregate model had been producing overlapping classes, so the model measured was
not one a cohort could represent; newborns now extend the youngest class while it
is narrower than the width, and every step is checked to construct a valid Cohort.
The first figures were discarded.

The comparison also found a real M4.3 defect, which is what a validation milestone
is for. plan_advance checked affordability with arrivals added before losses were
taken, but emitted the ledger legs outflows-first, so an interval affordable as a
whole underflowed leg by leg and was refused at commit. A fifty-tick interval
shedding 10,480 against 6,950 held while drawing 13,080 in reproduced it. Arrivals
are the first leg now, matching the order the arithmetic was checked in, with a
regression and a corrected escrow expectation in an existing test.

Release 38/38, analysis, ASan 38/38 and clang-format all ran and passed here.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-09-24 21:57:00 -04:00

Aeternalis Evolution

A custom-engine persistent evolutionary MMO. The world retains its biological history; each player life begins as a primitive cell and ends permanently.

This repository starts with architecture and a headless engineering proof: deterministic synthetic particles, component lifecycle, explicit PRNG streams, canonical snapshots and restored execution. It contains no gameplay, renderer, network service. Dev verification manifests are separate from the simulation.

Start here

Read AGENTS.md before changing code. Architecture and contracts are authoritative over implementation. No external runtime library is needed for the initial proof. C++23 support is checked during configuration.

Build and run

Requires CMake 3.28+, Ninja and a C++23 compiler/standard library supporting std::expected. GCC 14 is the initial Linux CI target; MSVC 19.44 or newer is the Windows target. Detailed verified tool versions are in the validation report.

cmake --preset debug
cmake --build --preset debug
ctest --preset debug
./build/debug/evo-sim --seed 123456 --entities 8 --ticks 1000 --verify-replay

On Windows use a Visual Studio x64 developer shell, or the repository's tools/windows-check.ps1 helper. Executables have .exe suffix on Windows.

./build/debug/evo-sim --seed 123456 --entities 8 --ticks 500 --save state.evo
./build/debug/evo-sim --load state.evo --ticks 500 --output results.json
./build/release/evo-sim --ecosystem --seed 0 --episodes 3000

Ecosystem mode measures seeded bounded episodes of the synthetic feeding community and reports ticks, births, both mortality paths, extinctions and a fingerprint. It shares no state, snapshot or particle metric with the proof run, and its strategy constants are measurement inputs rather than simulation rules. See ADR-0029 for scope.

./build/release/evo-sim --ensemble --seed 0 --seeds-per-cell 4 --strategies 3 --tick-limit 400

Ensemble mode derives strategies from genomes instead of writing them, then runs them on that same driver over a resource and temperature gradient. Each cell's seeds are split in half and both halves are reported separately and never merged, so a difference that appears in one half only is visible as the seed artifact it is. Options belong to exactly one mode: --episodes counts episodes and is rejected here, --seeds-per-cell counts seeds and is rejected in ecosystem mode. A cell's numbers are not fitness, adaptation or balance, and the resource axis is deliberately not monotonic. Lifetimes are censored: an organism still alive when an episode stops is reported as a survivor with the time it was watched, never as a death, so lived_ticks and observed_ticks answer different questions. See ADR-0035 for scope and for why.

The second command continues saved PRNG/component state. --verify-replay compares an uninterrupted run to a midpoint serialization/restore run. JSON reports the seed, ticks, active entities and canonical state hash. Snapshots are bounded local test files; they are not crash-consistent production storage.

Checks and measurements

cmake --build --preset debug --target format-check
cmake --preset analysis
cmake --build --preset analysis
cmake --preset asan
cmake --build --preset asan
ctest --preset asan
cmake --preset release
cmake --build --preset release
ctest --preset release
./build/release/evo-benchmark
./build/release/evo-spatial-benchmark

ASan uses UBSan too on supported Linux compilers. TSan has a separate tsan preset and is never combined with ASan. clang-format/tidy major 19 is the initial analysis baseline. The optional Clang libFuzzer target exercises snapshot restore. See testing strategy and deployment scaffold for scope and commands.

Repository boundaries

include/evo exposes core values and the headless simulation API. src/simulation implements component state, tick ordering and local snapshot codec. src/accounting provides the independent finite-inventory ledger; it is not yet part of particle snapshots. src/processes holds the synthetic producer, feeding, reproduction and episode fixtures built on that ledger; they are bounded experiments, not organism biology. src/biology holds the immutable genome record, the deterministic development that derives a phenotype from it and the metabolic budgets that give capability a cost, and the seeded mutation catalog that proposes successors and the archive that keeps them. No fixture reads a genome yet. apps/evo-sim owns CLI/file I/O and reporting. tests holds contract, fuzz and measurement programs. Future biology, world, networking and persistence modules are documented boundaries; they are added when they have real behavior.

World0 is persistent state. No normal configure, build, test or CI job may create, reset, deploy to or delete it. Gitea builds the proof image for Harbor; a bounded dev verification Job and manual Argo templates are in deploy/. Persistent-world deployment still requires storage, ownership and backup gates.

The project license has not been selected. Do not infer an open-source license from a public dependency's license.

S
Description
No description provided
Readme
7.5 MiB
Languages
C++ 83.4%
PLpgSQL 9.1%
Shell 2.8%
JavaScript 1.5%
CMake 1.3%
Other 1.9%