Field notes / Experiment guide
What is already known about ants and rotor walkers?
Which published results genuinely constrain our growth and boundary questions, and which similarities are only analogies?
New to these ants? Start here.
An ant moves on a grid of coloured squares. Each colour tells it to turn left (L) or right (R). It advances that square to the next colour, then steps forward. A rule such as RLR is the list of turns, starting at colour zero. On a blank grid the same rule always makes the same trajectory.
Territory means every cell visited so far. Its boundary consists of visited cells with an unvisited neighbour. A highway is a repeating pattern of motion that carries the ant away. A tether means repeated returns to a small home region.
Growing like the cube root of time means taking eight times as many steps to double a linear size. The rotor walker is a comparison model that follows and rotates arrows stored on cells. A seed makes a random starting pattern reproducible; the ant itself can still move deterministically.
We distinguish c, the chance of discovering a cell per boundary step, from kappa (κ), time at the boundary normalized by perimeter divided by area. Explaining their values is harder than using them to account for growth.
The idea under test
This documentary review separates established statements about specific models from proposed transfers to generalised Langton’s ants.
What we learned
The review records source access, claim locators and distinctions among displacement, visited range, width and local time.
What this leaves open
Abstract-only and inaccessible sources cannot establish technical details. A rotor theorem does not automatically apply to heading-relative ant dynamics.
Read the evidence and reproduction recipe ↗
Inside the experiment
Methods, registered predictions, exact commands and saved analysis. These are the working research documents; expand them for the detail behind the introduction.
Methods and original experiment record
Literature review, 2026-10-15
Started 2026-10-15 22:03 +13:00 (NZDT), base commit da88e61b690507526ba4e2a2c017d9c7941beb91, with pre-existing uncommitted manuscript/research changes.
Question: execute notes/literature-review-plan-2026-10-15.md: audit novelty, mechanisms and boundary comparisons before further simulation.
This is documentary research, not a numerical experiment. No new numerical hypothesis or measured prediction is registered. Existing timestamped predictions remain intact. No randomness or seeds apply. Literature numerical claims must retain their original observable and averaging definitions.
Method: search-service connector searches and full-text retrieval; exact queries, filters, source identifiers and access limits in results/search.md. Evidence notes in research_notes/Cube root growth literature/. Final synthesis in reports/ and notes/. Searches are time-dependent and cannot guarantee identical future results; log exact date and accessed versions.
Status: research in progress. Review notebooks and results will be finalized after evidence synthesis. Do not infer an exhaustive novelty proof from a negative search. No commits authorized.
Synthesis and verification, 2026-10-15
Completed deliverables: full report, cross-cutting synthesis, all-field model comparison, contribution claim audit, annotated bibliography, revised related-work prose and mechanism shortlist. Supporting corrections are propagated into the manuscript abstract, introduction, occupation/boundary discussion and conclusion, with six added bibliography entries. Pre-existing edits were preserved; no commits were made.
The late user-supplied Meng–Cohen full text changes the strongest finding: visited territory growth and boundary-access scaling are direct prior art. Its A~t^0.75/B~t^0.50 measurements and pocket/return mechanism distinguish the earlier periodic binary media from our uniform multicolour census; d=1 in its scaling formula gives area exponent2/3. The previous abstract-only access gap is superseded by appended evidence, not erased. No reviewed source explains the rule-specific occupation prefactor.
Status: focused first-pass deliverables complete, strict coverage stopping condition only partially met. Every headline claim has a nearest comparator, the flipping-rotator omission and closest Meng full-text gap are resolved, and rotor assumptions checked. One selective citation round found closer work and incorporated it; a further targeted round on the now-verified Meng mechanism, comprehensive forward coverage and Mishra Eq.5 clarification remain open. Original Family–Vicsek, internal-DLA and some historical ant texts remain inaccessible. No exhaustive novelty assertion is made.
Verification: final Typst compile succeeded with no diagnostics; 27-page PDF. Rendered pages1,2,21,25 were visually inspected (opening, introduction, boundary passage and conclusion); no clipping was seen. Notebook executed from a clean kernel through nbconvert; only a routine local-kernel transport warning. Provenance and file hashes are in results/verification.json; exact commands below. No simulator, estimator or dependency declarations changed, so no simulations or Rust release tests were rerun. Jupyter/PDF rendering use one-off uv environments and do not change pyproject/lock files.
export UV_PROJECT_ENVIRONMENT=$HOME/.venvs/turmite UV_LINK_MODE=copy
cd /mnt/f/Code/turmite
uv run experiments/2026-10-15-literature-review/scripts/build_notebook.py
uv run --with nbconvert --with ipykernel --with nbformat jupyter nbconvert --to notebook --execute --inplace experiments/2026-10-15-literature-review/notebook.ipynb
typst compile --root . research-outputs/cube-root-growth-paper/main.typ research-outputs/cube-root-growth-paper/main.pdf
uv run --with pypdfium2 experiments/2026-10-15-literature-review/scripts/verify_artifacts.py
The editorial scripts revise_manuscript.py, update_meng_synthesis.py and finalize_review.py document the one-time sequence against the prior working manuscript; they are not general rerunnable transforms. build_notebook.py and verify_artifacts.py are rerunnable. Notebook inputs are exclusively small results, but newly produced inputs remain uncommitted because commit authorization is absent. This explicit exception must be closed by committing and re-executing before archival sharing. The notebook prints each working-input status and SHA-256 rather than implying committed provenance.
Priority changes: complete fixed-area censor-complete boundary/interior time accounting, compare spatial local-time balance only as a conditional rotor-inspired mechanism, and complete already registered hull checks before another broad sweep or interface-class assignment.
Analysis notebook: notebook.ipynb
4 saved outputs · 2 code cells. Cells are not executed during publication. Download notebook
Question and why it matters
Has non-highway territory growth already been measured, which mechanisms constrain occupation, and what boundary comparisons are justified? This documentary review implements the README and plan. The complete report contains the model table and annotated bibliography; this notebook is a reproducible compact audit, not a new numerical experiment.
Assumptions and definitions
Full-text verification applies only to explicitly recorded sections; abstract-only and inaccessible sources cannot establish technical exponents. Displacement, range, width and accumulated local time differ. All material is a working-tree snapshot: the user did not authorize commits. Deviation: these newly produced results/ inputs are uncommitted, so the normal committed-input-only notebook rule cannot yet be met. The provenance output identifies this explicitly; commit and rerun before archival sharing.
Provenance
No simulation or fit is run and no random seed is used (fixed documentary seed recorded as 0 for deterministic tooling).
Code
from pathlib import Path
import hashlib, json, subprocess, random
SEED = 0
random.seed(SEED)
ROOT = next(p for p in [Path.cwd(), *Path.cwd().parents] if (p / "Cargo.toml").exists() and (p / "experiments").is_dir())
BASE = ROOT / "experiments/2026-10-15-literature-review"
inputs = [BASE / "results" / name for name in ["claim_audit.json", "novelty-search-log.md", "mechanisms-search.md", "boundary-search.md"]]
print("Review date: 2026-10-15 NZDT; documentary seed:", SEED)
print("HEAD:", subprocess.check_output(["git", "rev-parse", "HEAD"], cwd=ROOT, text=True).strip())
for p in inputs:
assert p.exists(), str(p)
rel = p.relative_to(ROOT).as_posix()
tracked = subprocess.run(["git", "ls-files", "--error-unmatch", rel], cwd=ROOT, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL).returncode == 0
modified = bool(subprocess.check_output(["git", "status", "--porcelain", "--", rel], cwd=ROOT, text=True).strip())
print(rel, hashlib.sha256(p.read_bytes()).hexdigest(), "committed-clean" if tracked and not modified else "UNCOMMITTED WORKING INPUT")
Data and method
Researchers recorded exact queries, versions, access limits and locators in the three search logs. The requested search-service succeeded for novelty but failed on mechanism/boundary attempts; fallback web retrieval is recorded. The producing script scripts/build_notebook.py extracts the claim-audit table from the report into results/claim_audit.json. No heavy computation, network or raw dumps are used here.
Results
Code
data = json.loads(inputs[0].read_text())
print(data["kind"])
print("Report:", data["source"], data["source_sha256"])
for row in data["claims"]:
print(row["claim"] + " -> " + row["action"])
The table shows which literature comparisons require manuscript wording changes. It does not establish exhaustive novelty or prove an ant asymptotic law.
Caveats and controls
No numerical estimates were regenerated; previously failed controls remain failed. Rotor calibration does not establish equal ant estimator bias. A displacement exponent cannot substitute for area or width. Surface centring and ensemble conventions differ. Mixed wavelength results and the failed high-band prediction do not establish persistent lobes. Latest source-access status is in the report; Eq.5 in Mishra remains ambiguous and citation tracing is selective.
What this changes / next steps
Remove the broad first-measurement assertion, retain the calibrated multicolour census and explicit horizons, and distinguish conditional cone arguments from rotor theorems. Prioritize fixed-area censor-complete transport accounting and the already registered hull checks over another broad sweep. Open questions: what sets rule-specific kappa, whether old-site increments balance on a meaningful ant cycle clock, and how persistent shape separates from local interface fluctuations. No simulations are launched by this review.
Read the dated lab log
Lab notebook
2026-10-15 22:03 (+13:00) · da88e61 · intent
Execute the literature review plan dated 2026-10-15, using the user-requested search-service connector. Start with closest flipping-rotator prior work and the manuscript bibliography, then mechanisms and boundary definitions. Preserve previous hypotheses and failed tests. Record observable distinctions, citation direction, versions, access limitations and uncertainty. No numerical experiment is performed and no existing results are overwritten. Working tree already contains user changes. Independent researchers write separate evidence/search files; the report writer synthesizes after novelty audit.
2026-10-15 22:14 (+13:00) · da88e61 · observation
Novelty documentary analysis completed provisionally; exact searches and retrieval status in results/novelty-search-log.md, evidence in research_notes/Cube root growth literature/novelty.md. Mishra full text distinguishes ensemble MSD, time averages and visited area. Backward tracing identifies Meng–Cohen1994 as an unresolved close growth source; publisher abstract available but full text inaccessible. No numerical run or test executed.
2026-10-15 22:14 (+13:00) · da88e61 · decision
Recommend narrowing manuscript novelty to systematic calibrated multicolour census and replacing claims that finite-horizon rules never build highways. First-pass stopping condition remains unmet because Meng–Cohen full text is missing and Mishra Eq5 is ambiguous. Synthesis writer will finalize README/notebook and propagate defensible manuscript wording; original scientific hypotheses remain intact.
2026-10-15 22:07 (+13:00) · da88e61 · deviation
Corrects the two entries labeled22:14 above: those times were an erroneous transcription, not the actual completion time. Windows Get-Date returned2026-10-15T22:06:58.5004580+13:00 immediately after writing them. This correction preserves the append-only record; substantive observations and decisions are unchanged.
2026-10-15 22:08 (+13:00) · da88e61 · intent
Synthesize the three evidence notes and search records into the specified report, claim audit and cross-cutting note. Propagate only supported wording corrections into the already modified manuscript. Preserve numerical results and predictions. New report/notebook inputs remain uncommitted because commits are not authorized; explicitly expose that notebook convention exception. No new dependencies or simulations.
2026-10-15 22:13 (+13:00) · da88e61 · observation
Report drafted with full comparator fields, annotated bibliography and mechanism tests. Applied manuscript editorial corrections with uv run experiments/2026-10-15-literature-review/scripts/revise_manuscript.py in WSL; first sandboxed WSL invocation returned E_ACCESSDENIED, approved escalation succeeded. User supplied a local Meng–Cohen PDF during synthesis, so provisional access-gap statements must be revised after its inspection; they are not a final access verdict. Existing measurements were not rerun.
2026-10-15 22:16 (+13:00) · da88e61 · observation
User supplied Meng–Cohen1994 original scan. OCR of10PDF pages completed with isolated uv pypdf/pypdfium2/RapidOCR; original SHA256 and pinned recipe in results/novelty-search-log.md. Full six article pages2482–2487 and appended figures visually checked. Eq1a/b directly measures visited area exponent.75 and boundary-count exponent.50 in periodic binary scatterer media; p2486 provides a boundary-return/pocket growth heuristic. Evidence appended to novelty.md; its approximate22:16heading was written at22:15:26, so read it as approximate, not precise event time. This entry supersedes previous closest-source-inaccessible conclusion.
2026-10-15 22:16 (+13:00) · da88e61 · interpretation
Meng–Cohen is a direct territory-growth precedent and a close antecedent to the boundary-return explanation: tau scales as A/B under their pocket/diffusive-bulk assumptions. Retain distinction of uniform multicolour census and measured c,kappa; remove broad mechanism/territory novelty assertions. Their exponent.75 differs from our area2/3 because their inferred boundary dimension is4/3. No new ant simulation performed. Parent/writer/mechanism researcher informed immediately. Other original-source coverage gaps remain.
2026-10-15 22:17 (+13:00) · da88e61 · deviation
Corrects the earlier synthesis observation labeled22:13: that heading was entered before22:13 (Get-Date after writing returned22:12:31); substantive content unchanged. The user-supplied Meng scan closes the previously recorded access gap. Preserve historical access failures and append the verified findings. New notebook results are uncommitted under the user's no-commit instruction, an explicit exception to committed-input-only notebook convention; provenance prints that status.
2026-10-15 22:17 (+13:00) · da88e61 · observation
Final manuscript compile and clean-kernel nbconvert execution succeeded; commands in README, logs in results/manuscript-build.log and results/notebook-execution.log, hashes in results/verification.json. PDF has27pages; rendered pages1,2,21,25 inspected without clipping. No numerical simulations, estimator changes or Rust release tests. Exact source checks/searches and scanned Meng extraction retained separately.
2026-10-15 22:17 (+13:00) · da88e61 · decision
Report, claim audit, comparison table, annotated bibliography, cross-cutting note and manuscript wording changes are complete. Credit Meng's territory measurements and boundary-return antecedent, narrow the census contribution and retain failed tests. Further targeted citation coverage and Mishra estimator clarification remain open. Prioritize fixed-area complete time budgets and registered hull checks; no commit made.
2026-10-15 · da88e61 · decision
The user authorized a commit after the review. Commit scope is the review records, plan, bibliography and isolated manuscript corrections. Earlier uncommitted manuscript formatting and prefix-prediction work remain outside this commit. The isolated source was reconstructed from HEAD using the review replacement recipes and rebuilt successfully with Typst under WSL; its PDF is staged together with that source. The original verification manifest describes the complete working-tree manuscript reviewed earlier, not this isolated commit build. No simulations or Rust changes are included. The scanned source and OCR text remain ignored scratch material.