Owner's decision: "we need to abandon the perl oracle, this was only useful
when initially porting, but I suspect many of the remaining problems have been
carried in from the perl (such as the weird scoring of outdoor and circulation
space, which definitely needs fixing)".
39 supports that second clause. Every defect the section found is inherited,
not introduced: the two-sided crinkliness gaussian that double-charges surplus
daylight (39.14), quality as a product over a variable number of factors
(39.18), value_supported priced as value_inside so a terrace was worth more per
m2 than a room (39.19), and circulation returning 0.07 per unit cost (hxi).
So parity with the oracle was never a safety net -- it was a commitment to
reproduce those defects. Each of 39.14, 39.18 and 39.19 would have been a
parity failure had parity ever been checked, and keeping the tests would have
meant reverting the fixes or explaining the failures away.
Removed: oracle.py, test_oracle.py, the two parity tests and their fixture
machinery in test_dom_corpus.py, innerloop.OracleEvaluator with its use_native
and urb_root plumbing, the same plumbing through driver, and fourteen
experiments/ scripts that could only run against Perl. Several of those are
cited in earlier DESIGN sections; the citations now point into git history,
which is the honest state -- they had been unrunnable since the oracle root
(/home/bruno/src/urb) stopped being present. run_search is superseded by
run_search_scaled, which does the same job natively.
Kept: dump_areas.pl/.py, which validate GEOMETRY against Urb (4.1) rather than
fitness, and the prose in fitness_cmd.py and dom.py explaining why the
.score/.fails formats are shaped as they are. Provenance is worth keeping; a
dead code path is not.
CLAUDE.md updated: fitness.py is the only evaluator, and "Urb did it this way"
is no longer an argument that a constant is right. 39.16 is the standing
counterweight in the other direction -- the crinkliness target WAS right and
twice looked wrong only because the code reading it was misunderstood.
Inheritance is neither evidence for nor against.
410 passed. The 69 removed cases account exactly: 64 parity (all skipped, since
no oracle .score was ever committed), 4 in test_oracle.py, and the guard test
39.20 added as a stopgap.
Closes homemaker-py-118. Files homemaker-py-bk9 for the re-baseline that 39.19
made necessary.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MJ84Feep79Hhm3E4zZJmnB
Expose tournament_k (default 2) on search()/search_staged(), threaded into
both _tournament call sites and the staged path's internal search() calls;
HOMEMAKER_TOURNAMENT_K env knob in the scaled/staged harnesses; run_6zy_ab.sh
joint niche×k grid (RESUME-able).
Result (negative, acceptable): no (niche,k) cell beats the legacy (off,k=2)
baseline. Blank-slate programme-house (5 seeds) baseline mean 4.80 fails is the
best of the 6-cell grid; every k>2 and every niche=on cell is 6.0-7.0. Niching
bites (pop_distinct 16/16 vs 4-11) but sharper pressure does not convert it to
lower fails — §11.5 'diffuses effort' null is robust to selection pressure;
plateau stays reachability-bound (confirms §11.4/§11.5).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Size each constructive-seed cut from leaf TARGET areas (division=[f,f] gives
left area-fraction f) and pick each cut's rotation for child squareness — both
derived from target dims, topology/type assignment untouched. Area-only
regressed (slivers); rotation choice is what makes it pay.
End-to-end (20000 evals, 3 seeds, staged): harbor 85.3->74.0 (-13%, best 69),
maple-court 151.7->136.0 (-10%, best 126). PROP=0 reproduces the §11.7/§12.1
baselines exactly. programme-house regresses at fixed budget (deeper local
optimum walls off the undivide restructuring path) but a budget sweep shows
it's convergence speed, not a worse asymptote (PROP=1 reaches 1 fail at 150k).
Default-on (seed_proportion_aware=True, env PROP=1).
cq1: n_storeys now honours storey_minimum, not just level: keys — programme-house
(storey_minimum:2, all rooms level:0) was seeded one storey short and fell
through to plain search. New programme.storey_minimum()/n_storeys_for();
driver.search passes min_storeys to the seeder; search_staged routes on the max.
No-op for harbor/maple; programme-house single-stage 8.0->5.0.
New maple-court best (126) saved as generated.dom. 204 tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
operators._assign_adjacency_aware spends ~one extra leaf per three rooms on a
greedy connected-dominating-set of circulation leaves (read from the geometric
leaf_graph, type-independent), so every room borders a connected circulation
spine and adjacency-to-c + access are satisfied by construction. Default-on via
constructive_topology(adjacency_aware=True), threaded through
driver.search(seed_adjacency_aware) and run_search_scaled.py (ADJ env).
End-to-end single-stage, 20000 evals, mean total fails over 3 seeds:
harbor 110.0 -> 90.7 (-17.5%; ADJ=0 reproduces the §11.2 105 baseline exactly),
programme-house 12.3 -> 9.3 (-24%). Adjacency-aware single-stage harbor (mean
90.7, best 85) beats the §11.3 staged best of 95 — the first Phase-6 fail-count
reduction from seeding. Follow-ups (lift_base_to_storeys, secondary adjacencies)
filed.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
genome.signature: ratio-invariant structural topology hash (per-storey tree
shape + cut orientation + leaf types), the cheap stand-in for the 9gp canonical
encoding. driver gains niche_by_signature (one individual per topology, replaces
the fitness-scalar dedup) and restart_patience (soft restart: keep elites,
refill with fresh seeds); SearchResult gains n_distinct_signatures /
diversity_history / n_restarts.
Diversity criterion MET (final-pop distinct ~5/16 -> 16/16). Gate NOT met:
blank-slate programme-house mean fails 12.3(legacy)/12.7(niche)/13.0(restart)
over 3 seeds at 20000 evals; harbor staged 95/94/108. Niching is a tie within
seed noise, restarts strictly worse — falsifies the premise that the
fitness-scalar dedup causes premature convergence. Both flags default-off,
kept for reuse. Epic c4c complete.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Implement a graded proximity comparator key (-n_fails, grade, fitness) behind
a default-off use_grade flag: fitness._leaf_grade / score_with_grade sum
f/FAIL_THRESHOLD over failing per-leaf quality factors; scalar fitness and fail
count stay untouched so the inner-loop 0.5^n cliff (§5.4) is unaffected (0/9
regression check: PASS). Read once per child in driver._evaluate off the
already-optimised tree; threaded through search_staged (Stage 2 only).
Harbor staged A/B (20000 evals, seeds 0/1/2): lex 95/96/106 (mean 99.0) vs
lex+grade 99/98/102 (mean 99.7) — grade wins 1/3, no plateau escape. Premise
falsified: within a fixed fail-tier 0.5^n is constant so fitness still spans
~6 orders of magnitude; grade above fitness displaces that working signal.
Verdict: reject; lexicographic (-n_fails, fitness) stands. Flag kept default-off
for reproducibility / possible reuse as a §11.5 diversity signal.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
programme-house budget=20000: 1.04e-02 (2 fails), 1.36× over Phase-2
oracle run and 2.60× over urb-evolve p128. Winning topology found via
rotate at eval 10357, unreachable within Phase-2 budget. 71.8 evals/s
(~140× faster than batched oracle).
harbor-house (16 rooms): 3.73e-18 (49 fails) at budget 10000 in 633s.
This programme is beyond the oracle's capability; native fitness makes
it feasible. 638 topologies explored.
Adds experiments/run_search_scaled.py (native-only search runner, no
oracle dependency). DESIGN.md records Phase 3 gate result.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
