De-scaffolded heterogeneous development V1: successful transfer, conventional costs favored

Question

Can one fixed, compact local developmental program choose cell roles, storage assignments, and legal connections on a role-unlabelled 3D substrate, producing working heterogeneous memory/routing systems across unseen layouts, physical sizes, and logical capacities?

Hypothesis

Interior blueprint removal, functional transfer and comparative utility are separate. A strong global conventional builder is allowed to match or outperform the local rule.

Registration

Locally frozen exploratory protocol; not public preregistration. SHA256 c319ec4485948e87221a760ff29162bc35a8d5e5ac2ba258d5f77f18939dc845. Complete executable/program/host/task/cap/seed/order/analysis freeze precedes scoring.

Methods

Read the full report/protocol downloads. Neutral six-neighbor lattice with passive public terminals, one uncommitted seed and initially vacant candidates. D uses charged local growth/registrations/differentiation/pruning. B independently plans globally and places distinct scalar latches nearer the query. Positive geometry-derived transport and one-slot bounded request/value handshake.

Baselines

B uses independent weighted Dijkstra and shared paths, the same boundary information, physics, primitives and deployment limits, with disclosed stronger global access. Exact compiled replay gates both arms, and is not a third competing method. All 96 paired full circuit hashes differ.

Evaluation

12 fresh paired groups, complete L=4/6 x K=4/8 x p=0/.15 matrix, two constructors: 192 cases. Four 64-event episodes with at least 32 scored queries/case, plus four dependent selected opposite-sign cloned pairs. Common window 1..1292 from independent physical envelope; no exact-tick coincidence. Primary is case success mean over seven held-out settings within group, then 12 groups.

Resource ceilings

One local standard-library CPU worker. Explicit caps: 216 candidates/cells, 540 selected edges, 1024 development ticks, 4096 messages/65536 development bytes, bounded local state/queues/interface, 1024 task messages/transaction and 100000 episode ticks. 3600 CPU/5400 wall seconds, 1 GiB sampled tree RSS, 128 MiB retained artifacts. Prior costs separate; no GPU, extra agent, paid service or installation.

Stopping conditions

All semantic/branching/detour/relabel/baseline/replay and five defect fixtures passed before freeze. All 192 cases completed, with no scientific failure, scored repair, extra seed/arm or deadline change. Initial monitor permission failure retained and charged full reservation. Post-freeze oracle/role/global-planning/state/replay defects would stop comparison.

Executing code

Exact source hashes in freeze.json; no frozen byte changed. program.py is the actual executable genome, not a prose descriptor. Post-run verify_results.py and derived reports/tables are outside frozen source. Prior assessment and studies preserved by local checked provenance hashes.

Configuration

Python 3.14.8, standard library, one numerical worker. Seed derivation, factor/task/layout coupling, run order, host admissibility, task version V2, wire/message/finite-state semantics and analysis in config/protocol/source. Only L4/K4/p0 and tiny explicit manual detours informed development; L6 and K8 did not.

Complete outcomes

D=B=100% held-out functional construction-transfer success (84/84 per arm); paired difference zero in all 12 groups. All 192 cases, 29072 ordinary queries, 768 selected both-correct pairs and 768 episode compiled replays passed. Program-controlled construction and finite-family transfer supported; no developmental utility advantage. No emergent intelligence, discovery, learned specialization or 3D-necessity claim.

Realized resources

Mean pooled query ticks D 21.30/B 12.54; ordinary messages D 269412/B 200583. Mean persistent field bytes D 282.29/B 220.33, active cells 27.52/27.32. D explored mean 137.23 candidates before pruning; B deployed mean 27.32. Monitored jobs 16.358 CPU/14.992 wall s; unmeasured monitor attempt separately charged 120 s. Sampled peak RSS 81559552 bytes. Full ledgers/conditional coverage and timing limits downloadable.

Limitations

Two fixed hand-authored algorithms, boundary-local D latch heuristic, supplied task-specific primitives, easy boundary-connected lattice admissibility, settled transactions and generous physical deadline. D uses a disclosed global construction-quiescence notification, not global path planning. No twelve independent algorithm discoveries, powered equivalence or universal zero uncertainty. B-favored traffic/state/latency; small timings descriptive. Ordinary communication counts separated from replay/challenge setup; selected pair observations dependent.

Reproducibility

Freeze-bound source plus configuration/protocol/fixture files in a disposable directory: python3 run_job.py smoke. For full matrix copy frozen-status.json to status.json first and ensure results absent, then python3 run_job.py full. Saved scores go in results for python3 verify_results.py. Prespecified group0/L6/K8/p.15 D/B author rerun matched; not independent reproduction.

Revision record

First release of a new independent bounded V1 study, task/runtime V2. Earlier completed reports preserved. Scientific execution, content review, visual inspection, author checks and publication delivery are separate. Constructor-only milestone closed.

Reviewed evidence & downloads

• report.md — derived. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• protocol.md — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• config.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• freeze.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• frozen-status.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• common.py — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• host.py — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• program.py — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• construct.py — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• runtime.py — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• task.py — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• experiment.py — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• run_job.py — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• development-history.md — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• fixture-results.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• verify_results.py — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• verification.json — derived. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• analysis.json — derived. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• aggregate.csv — derived. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• attempts.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• resources.json — derived. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• preservation.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• author-rerun.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• cases.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• queries.csv — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• hosts.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• events.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• summary.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• representatives.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• difficult-trace.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.
• progress.json — original. Selected project-authored static source, protocol/configuration or synthetic evidence authorized for this report. No external dataset or private experiment state.

These are selected public derivatives; raw transcripts, internal configuration and private source are withheld. Hashes describe the actual downloads, not unmodified original evidence.

Artifact manifest and source provenance · Permanent revision URL

Substantive versions

• Revision 1