From e4804adabcdce6d928c5e6b1e85429b2a6acf2fe Mon Sep 17 00:00:00 2001 From: Giorgio Gilestro Date: Mon, 7 Sep 2026 17:13:02 +0100 Subject: [PATCH] SI: adopt the clearer rewrite, with factual corrections; fix two build bugs it exposed Prose: adopted the simplified rewrite for the Reproducibility preamble, SI Text S1 and S2, and the two tables. It reads better - shorter sentences, no shouty caps, no self-commentary in the proposition headings. Fact-checked against the artifacts before adopting. Corrections: - Table S2 said grounding retention used "18+ replicates per point". E2 uses 100 lineages; 18 is the *neural* grounding sweep. (Pre-existing error, faithfully carried over by the rewrite.) - The emergent parents' 0.535/0.474 are the accuracies at the LONGEST divergence (t_div=3200), not overall means (0.595/0.545); now qualified. Verified merge holds 0.954-0.956 at every divergence, residual exactly 0.000 in both emergent conditions. - Dropped an invented run date (2026-08-11; the run is from 2026-09-06) and an internal project-phase reference ("Phase 3"). - The llm_speciation duration question is no longer open - it ran, and found no isolation from over-training (1-12 epochs); text updated. - Restored the confidence-weighting numbers the rewrite dropped: paired bootstrap contrast |rho| = -0.021, CI [-0.130, +0.059] (re-derived), plus the nuance that the weighting does sharpen the level contrast. - "Minimal model" -> "biological model"; "LLM tier in progress" -> done. - Trimmed an unverifiable citation ("neuron-identifiability approaches...") to the reference the bibliography actually carries. Two rendering bugs the LaTeX version exposed, both pre-existing: - Greek and several math symbols were absent from build.py's unicode map, so alpha and epsilon were rendering as missing-glyph boxes in the SI. Added Greek, set membership, superscripts, proper minus. Both PDFs now contain zero missing glyphs. - inline() split on code spans BEFORE applying emphasis, so any italic containing `code` was torn into fragments - visible in the main text as a literal "is*" and mis-scoped italics on p. 3. Code spans are now stashed behind sentinels first. This fixed the manuscript, not just the SI. - A leading markdown H1 leaked into the body as literal text; the wrapper supplies the title, so it is now skipped. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01BkRLcc18rwT2Lysu6PbG7v --- paper/pnas/body.tex | 12 +-- paper/pnas/build.py | 44 +++++++---- paper/pnas/main.pdf | Bin 939024 -> 941833 bytes paper/pnas/si.md | 167 +++++++++++++++++++++++------------------ paper/pnas/si.pdf | Bin 102744 -> 108429 bytes paper/pnas/si_body.tex | 70 +++++++++-------- 6 files changed, 167 insertions(+), 126 deletions(-) diff --git a/paper/pnas/body.tex b/paper/pnas/body.tex index 27c5fcd..476a8ff 100644 --- a/paper/pnas/body.tex +++ b/paper/pnas/body.tex @@ -26,7 +26,7 @@ An operator of a model population faces recurring decisions for which there is n \section*{The biological model, and where trained learners depart from it} -Knowledge is modelled as a distribution \texttt{p\_t} over \texttt{K} discrete items (capabilities, facts, modes of behaviour), with a fixed true distribution \texttt{p*} whose rare tail carries the knowledge most at risk. One generation is: *draw \texttt{n} samples from the parent's distribution, optionally mix in \texttt{m} verified real samples (``grounding'', \texttt{g = m/(n+m)}), and refit the child\emph{. This construction is the paper's }biological model\emph{, deliberately minimal, and its resampling step }is* the Wright--Fisher process, population genetics' canonical model of neutral evolution, in which each new generation is a random sample of size \texttt{n} from the previous one and every statistical property of drift follows from that one step. Diversity throughout this paper is \emph{heterozygosity}, \texttt{H = 1 − Σ p\_i²}: the probability that two random draws differ (one minus a collision probability), high when many items share the mass, zero at total collapse. The identity is exploited as an engineering gate: the simulator reproduces the classical closed forms (heterozygosity decay \texttt{E[H\_t] = H\_0(1 − 1/n)\textasciicircum{}t}; the exact immigration--drift equilibrium; the closed-form multi-teacher union) to within 0.5\%, and these are standing tests in the codebase, not one-off checks. +Knowledge is modelled as a distribution \texttt{p\_t} over \texttt{K} discrete items (capabilities, facts, modes of behaviour), with a fixed true distribution \texttt{p*} whose rare tail carries the knowledge most at risk. One generation is: \emph{draw \texttt{n} samples from the parent's distribution, optionally mix in \texttt{m} verified real samples (``grounding'', \texttt{g = m/(n+m)}), and refit the child}. This construction is the paper's \emph{biological model}, deliberately minimal, and its resampling step \emph{is} the Wright--Fisher process, population genetics' canonical model of neutral evolution, in which each new generation is a random sample of size \texttt{n} from the previous one and every statistical property of drift follows from that one step. Diversity throughout this paper is \emph{heterozygosity}, \texttt{H = 1 \(-\) \(\Sigma\) p\_i\(^{2}\)}: the probability that two random draws differ (one minus a collision probability), high when many items share the mass, zero at total collapse. The identity is exploited as an engineering gate: the simulator reproduces the classical closed forms (heterozygosity decay \texttt{E[H\_t] = H\_0(1 \(-\) 1/n)\textasciicircum{}t}; the exact immigration--drift equilibrium; the closed-form multi-teacher union) to within 0.5\%, and these are standing tests in the codebase, not one-off checks. Where a real learner departs from this model matters, and I measured that boundary rather than assumed it. Real training adds approximation, optimisation noise, and inductive bias, and when trained networks are fit against the exact drift null they deviate in \emph{opposite, architecture-specific} directions: a smoothing recurrent network resists collapse (keeping spurious variants alive), while a sharpening image generator accelerates it. A one-parameter \emph{learning kernel} (a smoothing knob and a sharpening knob on the refit) reproduces both. Throughout, a real learner is therefore treated as Wright--Fisher \emph{plus a signed, measurable estimator bias}, and the drift signs (rare-first loss; the grounding response) survived that bias in every architecture I tested, including a convolutional VAE retrained on its own generated digits, where the dry lineage collapses to a single blurred digit class while 10\% grounding holds all thirty modes (Fig. 2). Retraining on a single parent is \emph{asexual reproduction}, and sustained loss under it carries the defining consequence of \emph{Muller's ratchet} (28), the mechanism by which lineages that never recombine decay irreversibly --- the reason non-recombining genomes such as the Y chromosome have shed most of their ancestral genes. Once every copy of a rare capability is gone from all parents and sources, no recombination can rebuild it: each such loss is a click of the ratchet, and remedies must act while copies still survive somewhere (a consequence-level correspondence; the biological model lacks the ratchet's recurrent-mutation driver). @@ -51,7 +51,7 @@ Selection on a fitness function & Verifier-anchored selection (``reality that ca \subsection*{Grounding is immigration: cheap, with a floor} -In the biological model, grounding from a fixed real source is \emph{immigration} into a drifting population (29--31). Immigration is what conservation managers prescribe when a fragmented reserve loses diversity, and its striking property there is how little is needed --- the field's rule of thumb is that one migrant per generation holds an isolated population's diversity (32). The same economy appears here: the equilibrium diversity has a closed form the simulator matches exactly. That equilibrium is \emph{smooth} in the grounding fraction (there is no phase transition in aggregate diversity), so the practical number is an operational threshold, and I define it as such: under the tested population size and Zipf source distribution, \texttt{g \(\approx\) 0.05} retained most (\(\geq\)95\%) of equilibrium diversity indefinitely, with the required fraction depending on sample size, source distribution, and the chosen retention target (dependencies in SI). Verified real data remains, on any of these definitions, cheap insurance at fractions far below one. But the same analysis yields a floor the field's average-loss framing misses: under unstratified sampling from the source, a capability of rarity \texttt{p} appears in a real-data batch of size \texttt{m} with probability \texttt{1 − e\textasciicircum{}{−m\(\cdot\)p}}, so \texttt{m\(\cdot\)p \(\approx\) 1} marks roughly a 63\% chance of one example per batch: a soft observation floor, with higher confidence priced accordingly, and with distinct consequences for continuous retention, stationary occupancy, and reintroduction after loss (immigration can restore an absent item; SI separates these). Protecting the rarest knowledge under unstratified grounding is therefore priced per item at cost \texttt{\(\propto\) 1/p}; targeted or stratified sampling changes that cost, and recombination can recover rare capabilities \emph{that are still retained across complementary parents} (next section). In trained networks the \emph{sign} of the grounding response transfers everywhere I looked, with two deviations, both traced to the estimator bias above: sharp thresholds soften, and support-counting metrics decouple from truth (forward-KL is the operative collapse metric for a smoothing learner). On real images (Fig. 2B), dry self-training collapses a convolutional VAE to one mode while \textasciitilde{}10\% grounding holds all thirty (the trained model needs roughly twice the biological model's fraction, the measured price of the estimator bias). +In the biological model, grounding from a fixed real source is \emph{immigration} into a drifting population (29--31). Immigration is what conservation managers prescribe when a fragmented reserve loses diversity, and its striking property there is how little is needed --- the field's rule of thumb is that one migrant per generation holds an isolated population's diversity (32). The same economy appears here: the equilibrium diversity has a closed form the simulator matches exactly. That equilibrium is \emph{smooth} in the grounding fraction (there is no phase transition in aggregate diversity), so the practical number is an operational threshold, and I define it as such: under the tested population size and Zipf source distribution, \texttt{g \(\approx\) 0.05} retained most (\(\geq\)95\%) of equilibrium diversity indefinitely, with the required fraction depending on sample size, source distribution, and the chosen retention target (dependencies in SI). Verified real data remains, on any of these definitions, cheap insurance at fractions far below one. But the same analysis yields a floor the field's average-loss framing misses: under unstratified sampling from the source, a capability of rarity \texttt{p} appears in a real-data batch of size \texttt{m} with probability \texttt{1 \(-\) e\textasciicircum{}{\(-\)m\(\cdot\)p}}, so \texttt{m\(\cdot\)p \(\approx\) 1} marks roughly a 63\% chance of one example per batch: a soft observation floor, with higher confidence priced accordingly, and with distinct consequences for continuous retention, stationary occupancy, and reintroduction after loss (immigration can restore an absent item; SI separates these). Protecting the rarest knowledge under unstratified grounding is therefore priced per item at cost \texttt{\(\propto\) 1/p}; targeted or stratified sampling changes that cost, and recombination can recover rare capabilities \emph{that are still retained across complementary parents} (next section). In trained networks the \emph{sign} of the grounding response transfers everywhere I looked, with two deviations, both traced to the estimator bias above: sharp thresholds soften, and support-counting metrics decouple from truth (forward-KL is the operative collapse metric for a smoothing learner). On real images (Fig. 2B), dry self-training collapses a convolutional VAE to one mode while \textasciitilde{}10\% grounding holds all thirty (the trained model needs roughly twice the biological model's fraction, the measured price of the estimator bias). \begin{figure*}[p]\centering % fig2 \includegraphics[width=\textwidth]{figs/fig2.pdf} @@ -83,7 +83,7 @@ Sex has risks and, for AI, an unfair advantage, both quantified on Kauffman's NK \subsection*{The society: grounding, recombination, and diversity make complementary contributions} -Composing the operators (Fig. 6) requires one definitional distinction first. In the inheritance model, grounding is \emph{grounded inheritance}: external samples added to the reproduction process (the data channel). In the society model, grounding is \emph{grounded evaluation}: selection weights true fitness against conformity to the population's own consensus, \texttt{g}\(\cdot\)true-fitness + (1−g)\(\cdot\)conformity, the analogue of scoring models by the crowd's approval (the fitness channel). These are related design ideas, since both couple the lineage to a non-drifting external signal, but they are different operators, and I name them separately. In the tested society (a finite agent population on a rugged NK landscape), a four-arm ablation separates the failure modes: the full system (grounded evaluation + directed recombination + diversity-preserving selection (40)) climbs to near the global optimum while keeping its specialists; removing grounded evaluation converges the population confidently on an unfit consensus (self-consumption); removing recombination strands it on local optima; removing diversity converges it prematurely to a worse answer. Each removal fails differently; the three implementations make complementary contributions \emph{under the tested conditions}; general joint necessity is not established (alternative mutation, restart, archive, or selection schemes could alter the picture). At language-model scale this composed loop remains unbuilt; it is the paper's largest stated gap. +Composing the operators (Fig. 6) requires one definitional distinction first. In the inheritance model, grounding is \emph{grounded inheritance}: external samples added to the reproduction process (the data channel). In the society model, grounding is \emph{grounded evaluation}: selection weights true fitness against conformity to the population's own consensus, \texttt{g}\(\cdot\)true-fitness + (1\(-\)g)\(\cdot\)conformity, the analogue of scoring models by the crowd's approval (the fitness channel). These are related design ideas, since both couple the lineage to a non-drifting external signal, but they are different operators, and I name them separately. In the tested society (a finite agent population on a rugged NK landscape), a four-arm ablation separates the failure modes: the full system (grounded evaluation + directed recombination + diversity-preserving selection (40)) climbs to near the global optimum while keeping its specialists; removing grounded evaluation converges the population confidently on an unfit consensus (self-consumption); removing recombination strands it on local optima; removing diversity converges it prematurely to a worse answer. Each removal fails differently; the three implementations make complementary contributions \emph{under the tested conditions}; general joint necessity is not established (alternative mutation, restart, archive, or selection schemes could alter the picture). At language-model scale this composed loop remains unbuilt; it is the paper's largest stated gap. \begin{figure*}[p]\centering % fig6 \includegraphics[width=\textwidth]{figs/fig6.pdf} @@ -107,15 +107,15 @@ The pre-registered \emph{emergent test} constrains the claim most: true BDM inco The framework's prediction-level claim was put to a designed test (Fig. 3C). Thirty-nine parent pairs (13 conditions \(\times\) 3 seeds; rows are not independent --- parents share task-data seeds across conditions, so inference is condition-clustered, and because shared seeds also couple rows \emph{across} conditions I report per-seed and leave-one-seed-out sensitivity alongside) span three axes decorrelated by construction: \emph{conflict} (contradictory conventions on shared prompts, private budgets fixed), \emph{compatible overlap} (the same shared prompts under the same convention --- overlap and volume without conflict), and \emph{duration} (weight divergence with zero conflict). Before merging, six predictors are computed: \emph{confidence-weighted functional conflict} (bilateral confident disagreement on probes drawn blind to where conflict lives --- a proposed proxy for merge-relevant interactions, motivated by the observation that raw disagreement counts harmless complementation, one parent merely ignorant, as conflict), raw disagreement, gradient alignment at the shared base (47), LoRA-delta cosine and distance, and a cross-task performance baseline. The pre-registered outcome is the merge penalty against oracle parent potential (the analogue of \emph{hybrid load}, the fitness a hybrid loses relative to what its parents' genes could jointly supply), also reported against best- and mean-parent references because the predictor ordering is sensitive to that choice. -Across this controlled grid, pre-merge functional disagreement predicted merge penalties (clustered bootstrap CIs excluding zero; held-out leave-one-condition-out \(\rho\) \(\approx\) 0.35--0.40), whereas LoRA-delta cosine and L2 showed no statistically detectable association; gradient alignment carried intermediate signal. Head-to-head predictor differences are not individually significant at this sample size; only these baselines were tested; and with three seeds, uncertainty about seed generalisation remains substantial --- though the seed sensitivity favours the functional measures (per-seed \(\rho\) stable at +0.37 to +0.53 in each seed alone, geometry \(\approx\) 0 in every seed, gradient alignment seed-unstable at −0.11 to −0.55). Two further results bound the claim: the initial two-axis grid's best predictor was delta-cosine (\(\rho\) = +0.60) --- an overlap artefact that the compatible-overlap control was added to expose, and did (collapse to +0.03); and the pre-registered internal prediction that confidence weighting would beat raw disagreement \emph{failed} (they are statistically indistinguishable as rank predictors), so the present evidence favours functional disagreement generally, not the DMI-specific refinement. The framework motivated the measurement and the controls; their success does not validate the specifically population-genetic mechanism. Whether the prediction improves a budget-matched operator choice, and whether it generalises to unfamiliar conflict structures and real task pairs, are the experiment's open front. +Across this controlled grid, pre-merge functional disagreement predicted merge penalties (clustered bootstrap CIs excluding zero; held-out leave-one-condition-out \(\rho\) \(\approx\) 0.35--0.40), whereas LoRA-delta cosine and L2 showed no statistically detectable association; gradient alignment carried intermediate signal. Head-to-head predictor differences are not individually significant at this sample size; only these baselines were tested; and with three seeds, uncertainty about seed generalisation remains substantial --- though the seed sensitivity favours the functional measures (per-seed \(\rho\) stable at +0.37 to +0.53 in each seed alone, geometry \(\approx\) 0 in every seed, gradient alignment seed-unstable at \(-\)0.11 to \(-\)0.55). Two further results bound the claim: the initial two-axis grid's best predictor was delta-cosine (\(\rho\) = +0.60) --- an overlap artefact that the compatible-overlap control was added to expose, and did (collapse to +0.03); and the pre-registered internal prediction that confidence weighting would beat raw disagreement \emph{failed} (they are statistically indistinguishable as rank predictors), so the present evidence favours functional disagreement generally, not the DMI-specific refinement. The framework motivated the measurement and the controls; their success does not validate the specifically population-genetic mechanism. Whether the prediction improves a budget-matched operator choice, and whether it generalises to unfamiliar conflict structures and real task pairs, are the experiment's open front. Headline quantitative results, with sample sizes, uncertainty, and outcome definitions, are collected in SI Appendix, Table S2. \section*{Discussion} -\textbf{Design rules.} As engineering guidance, the results reduce to rules that an operator of a model population can apply, answering the four decisions posed in the Introduction. \emph{Ground every generation} in verified reality --- a few percent retained most diversity in the tested settings --- but price the rarest capabilities individually (observation probability \texttt{1 − e\textasciicircum{}{−m\(\cdot\)p}} per batch under unstratified sampling), consider targeted sampling for the deep tail, and use recombination to recover rare capabilities still retained across complementary parents. \emph{Merge, don't blend, when there is headroom}: keep specialists intact and route, or breed-and-screen candidate merges, whenever the naive average is far from ceiling; plain averaging is adequate only where a strong base has already composed the skills. \emph{Match the operator to entanglement}: merge freely when skills are additive; sparingly, with offspring selection, when they entangle; and expect the champion-optimal mating breadth to narrow as landscapes roughen. \emph{Preserve diversity as a first-class objective}, because selection can only preserve variety that exists, and in the tested society its removal produced a distinct failure mode. \emph{Before merging, measure functional conflict} --- cheap, pre-merge, and in the controlled setting predictive where the tested weight-distance baselines were not; and \emph{do not treat divergence or specialisation alone as evidence of incompatibility} --- in every regime tested here, what broke merging was conflicting conventions on shared circuitry, which is the thing to detect. +\textbf{Design rules.} As engineering guidance, the results reduce to rules that an operator of a model population can apply, answering the four decisions posed in the Introduction. \emph{Ground every generation} in verified reality --- a few percent retained most diversity in the tested settings --- but price the rarest capabilities individually (observation probability \texttt{1 \(-\) e\textasciicircum{}{\(-\)m\(\cdot\)p}} per batch under unstratified sampling), consider targeted sampling for the deep tail, and use recombination to recover rare capabilities still retained across complementary parents. \emph{Merge, don't blend, when there is headroom}: keep specialists intact and route, or breed-and-screen candidate merges, whenever the naive average is far from ceiling; plain averaging is adequate only where a strong base has already composed the skills. \emph{Match the operator to entanglement}: merge freely when skills are additive; sparingly, with offspring selection, when they entangle; and expect the champion-optimal mating breadth to narrow as landscapes roughen. \emph{Preserve diversity as a first-class objective}, because selection can only preserve variety that exists, and in the tested society its removal produced a distinct failure mode. \emph{Before merging, measure functional conflict} --- cheap, pre-merge, and in the controlled setting predictive where the tested weight-distance baselines were not; and \emph{do not treat divergence or specialisation alone as evidence of incompatibility} --- in every regime tested here, what broke merging was conflicting conventions on shared circuitry, which is the thing to detect. -\textbf{Continual learning at the population scale.} Within a single network, the discipline's remedies for forgetting are this framework's operators writ small. Rehearsal and replay of stored data (26, 27) is grounded inheritance within one lineage, and the replay fractions the field settled on empirically, on the order of 1\% for instruction tuning (48) and 5\% to 25\% by distribution-shift strength in continual pretraining (49), sit where the biological model's operational threshold lies. \emph{Pseudo-rehearsal}, the replay of a network's own generated samples, proposed as a cure in 1995 (50) and revived as generative replay (51), is precisely the ungrounded null studied here: immigration from a drifting source, benign for one hop, compounding over generations, with verifier-filtering (29, 52) converting it back into grounding. Parameter isolation (53), including frozen-base adapters, which forget far less (54), is engineered decorrelation; complementary-learning-systems consolidation (55--57) is the periodic adapter-into-base merge; the recent turn to merging as a continual-learning mechanism (58--61) applies recombination within one lineage over time, where this paper applies it across lineages; and the observation that rare examples and long-tail knowledge are forgotten first (62--64) is tail extinction seen one model at a time. The mechanisms differ (forgetting is largely deterministic interference, collapse is sampling drift) but the victims and the remedies coincide, and to my knowledge no prior work carries population-genetic formalism into continual learning. Read into that field, the results offer: (i) an equilibrium theory for the replay ratio, with the sharper prediction that the required fraction is set by the rarest capability one refuses to lose (the \texttt{1 − e\textasciicircum{}{−m\(\cdot\)p}} law) rather than by average loss, testable against published replay sweeps; (ii) a \emph{failure theory for generative replay}: self-generated rehearsal is safe for short horizons and compounds into collapse across generations unless verifier-filtered back into grounding (29, 50--52); (iii) \emph{pre-merge interference prediction with a mechanism}: where the current state of the art fits regressions over candidate metrics (47), the functional-conflict measure arrives at a convergent signal from principle and comes with an operator prescription --- when conflict is high, do not average; route or breed-and-screen; (iv) a candidate \emph{decision rule for the consolidate-versus-stay-modular question} that currently splits the field's practice (keep adapters separate vs merge them; 57--61): union-preserving operators where headroom exists, fusion where the base composes, consolidation as the slow-store step; and (v) \emph{tail monitoring as the leading indicator}: continual-learning evaluation that averages over capabilities hides exactly the losses that drift theory says come first and, past a threshold, become irreversible. On that last point I note the standing objection that apparent forgetting can be skewed task-inference over latent capability rather than erasure (65); the irreversibility results here concern oracle-measured behavioural distributions, and distinguishing latent from extinct capability at language-model scale is an open experiment whose outcome would be decisive for both readings. +\textbf{Continual learning at the population scale.} Within a single network, the discipline's remedies for forgetting are this framework's operators writ small. Rehearsal and replay of stored data (26, 27) is grounded inheritance within one lineage, and the replay fractions the field settled on empirically, on the order of 1\% for instruction tuning (48) and 5\% to 25\% by distribution-shift strength in continual pretraining (49), sit where the biological model's operational threshold lies. \emph{Pseudo-rehearsal}, the replay of a network's own generated samples, proposed as a cure in 1995 (50) and revived as generative replay (51), is precisely the ungrounded null studied here: immigration from a drifting source, benign for one hop, compounding over generations, with verifier-filtering (29, 52) converting it back into grounding. Parameter isolation (53), including frozen-base adapters, which forget far less (54), is engineered decorrelation; complementary-learning-systems consolidation (55--57) is the periodic adapter-into-base merge; the recent turn to merging as a continual-learning mechanism (58--61) applies recombination within one lineage over time, where this paper applies it across lineages; and the observation that rare examples and long-tail knowledge are forgotten first (62--64) is tail extinction seen one model at a time. The mechanisms differ (forgetting is largely deterministic interference, collapse is sampling drift) but the victims and the remedies coincide, and to my knowledge no prior work carries population-genetic formalism into continual learning. Read into that field, the results offer: (i) an equilibrium theory for the replay ratio, with the sharper prediction that the required fraction is set by the rarest capability one refuses to lose (the \texttt{1 \(-\) e\textasciicircum{}{\(-\)m\(\cdot\)p}} law) rather than by average loss, testable against published replay sweeps; (ii) a \emph{failure theory for generative replay}: self-generated rehearsal is safe for short horizons and compounds into collapse across generations unless verifier-filtered back into grounding (29, 50--52); (iii) \emph{pre-merge interference prediction with a mechanism}: where the current state of the art fits regressions over candidate metrics (47), the functional-conflict measure arrives at a convergent signal from principle and comes with an operator prescription --- when conflict is high, do not average; route or breed-and-screen; (iv) a candidate \emph{decision rule for the consolidate-versus-stay-modular question} that currently splits the field's practice (keep adapters separate vs merge them; 57--61): union-preserving operators where headroom exists, fusion where the base composes, consolidation as the slow-store step; and (v) \emph{tail monitoring as the leading indicator}: continual-learning evaluation that averages over capabilities hides exactly the losses that drift theory says come first and, past a threshold, become irreversible. On that last point I note the standing objection that apparent forgetting can be skewed task-inference over latent capability rather than erasure (65); the irreversibility results here concern oracle-measured behavioural distributions, and distinguishing latent from extinct capability at language-model scale is an open experiment whose outcome would be decisive for both readings. \textbf{What is borrowed and what is new.} The collapse-as-drift diagnosis is established prior work (21--25); so are the empirical facts that merges can beat parents, that decorrelated parents merge better, and that naive averaging loses to interference-aware or routed merges (4, 66, 67), that model populations can climb (5, 8--10), and that merge success admits ML-native predictors (47, 68), correlational where this framework supplies mechanism; the reading of sex as an algorithm for mixability in the theory of computation (69) anticipated the transfer before model merging existed. New here is the framework-level synthesis --- inheritance, diversity, and compatibility as managed quantities --- together with: the conservation law for blending inheritance and its operator boundaries; the per-item grounding floor; the society ablation with its complementary failure modes; model speciation as a named, tested question, with the coordinate-versus-functional decomposition under permutation-and-rescaling alignment and the emergent null that bounds it; and the controlled predictive test with its controls. I claim the framework generated these measurements and experiments; I do not claim that their outcomes validate a uniquely population-genetic mechanism, and one refinement it proposed was not supported. diff --git a/paper/pnas/build.py b/paper/pnas/build.py index d572022..7e54709 100644 --- a/paper/pnas/build.py +++ b/paper/pnas/build.py @@ -106,9 +106,17 @@ FIGURES: dict[str, tuple[list[str], str]] = { "between predictors are not individually significant."), } -UNICODE = {"—": "---", "–": "--", "→": r"\(\rightarrow\)", "≈": r"\(\approx\)", "≥": r"\(\geq\)", - "≳": r"\(\gtrsim\)", "×": r"\(\times\)", "·": r"\(\cdot\)", "μ": r"\(\mu\)", - "ρ": r"\(\rho\)", "≤": r"\(\leq\)", "≪": r"\(\ll\)", "∝": r"\(\propto\)"} +UNICODE = { + # Multi-character sequences first: esc() applies these in insertion order. + "⁽ᵏ⁾": r"\(^{(k)}\)", "⁻³": r"\(^{-3}\)", "⁻⁴": r"\(^{-4}\)", + "—": "---", "–": "--", "→": r"\(\rightarrow\)", "≈": r"\(\approx\)", "≥": r"\(\geq\)", + "≳": r"\(\gtrsim\)", "×": r"\(\times\)", "·": r"\(\cdot\)", "μ": r"\(\mu\)", + "ρ": r"\(\rho\)", "≤": r"\(\leq\)", "≪": r"\(\ll\)", "∝": r"\(\propto\)", + # Greek and math symbols: the typewriter font has no Greek, so these must become math. + "α": r"\(\alpha\)", "β": r"\(\beta\)", "ε": r"\(\varepsilon\)", "Δ": r"\(\Delta\)", + "Σ": r"\(\Sigma\)", "∈": r"\(\in\)", "≠": r"\(\neq\)", "±": r"\(\pm\)", + "∼": r"\(\sim\)", "−": r"\(-\)", "²": r"\(^{2}\)", "³": r"\(^{3}\)", "⁴": r"\(^{4}\)", +} SPECIALS = {"&": r"\&", "%": r"\%", "#": r"\#", "_": r"\_", "$": r"\$", "~": r"\textasciitilde{}", "^": r"\textasciicircum{}"} @@ -123,19 +131,21 @@ def esc(s: str) -> str: def inline(s: str) -> str: - parts = re.split(r"(`[^`]*`)", s) - out = [] - for p in parts: - if p.startswith("`") and p.endswith("`") and len(p) >= 2: - out.append(r"\texttt{" + esc(p[1:-1]) + "}") - else: - p = esc(p) - p = re.sub(r"\[([^\]]+)\]\((https?://[^)]+)\)", r"\\href{\2}{\1}", p) - p = re.sub(r"\*\*([^*]+)\*\*", r"\\textbf{\1}", p) - p = re.sub(r"\*([^*]+)\*", r"\\emph{\1}", p) - p = re.sub(r'"([^"]+)"', r"``\1''", p) - out.append(p) - return "".join(out) + # Code spans are stashed behind sentinels *before* emphasis is applied, so that an italic or + # bold span containing `code` is still matched as one span (splitting first would break it). + codes: list[str] = [] + + def stash(m: re.Match) -> str: + codes.append(r"\texttt{" + esc(m.group(0)[1:-1]) + "}") + return f"\x00{len(codes) - 1}\x00" + + s = re.sub(r"`[^`]*`", stash, s) + s = esc(s) + s = re.sub(r"\[([^\]]+)\]\((https?://[^)]+)\)", r"\\href{\2}{\1}", s) + s = re.sub(r"\*\*([^*]+)\*\*", r"\\textbf{\1}", s) + s = re.sub(r"\*([^*]+)\*", r"\\emph{\1}", s) + s = re.sub(r'"([^"]+)"', r"``\1''", s) + return re.sub(r"\x00(\d+)\x00", lambda m: codes[int(m.group(1))], s) def figure_env(name: str) -> str: @@ -192,6 +202,8 @@ def convert(text: str) -> str: emit_table(block, out) elif first == "---" and len(block) == 1: out.append("\\medskip\\hrule\\medskip"); out.append("") + elif first.startswith("# ") and not first.startswith("## "): + continue # document title: the wrapper supplies it elif first.startswith("## "): out.append(f"\\section*{{{inline(first[3:])}}}"); out.append("") elif first.startswith("### "): diff --git a/paper/pnas/main.pdf b/paper/pnas/main.pdf index 023c6c6abec94df74910577c51d932676b4fc23f..f7cb2193076cf3e32b3f19c82f95fb26278f5938 100644 GIT binary patch delta 86214 zcmZ6zbyOTpw>_L;aEHMH!7YJ6aF^f?!QEYhI}8L8TpJh&?jGD-g1ZI{u7ThXEZ;oO zdw=h`_xq#Poa(BsI(_Qwz1LLrOneD{L??eWJ|zMtM*|43%edd8x?xLZGcs+9}5I)7s<%{j>@;J*#H#3(N$x zUlRNt>8>qM@IX~|MgG|WFbrPwi!3u-f0T%75&X1$iW*yEd47-5U2^2CSHOxfC3>$P z#%LS06s}@=DL5|iVY%q!?HywG@OGp~K)av$u}~CMhdWepH5GQ+gVpA7r{MV)-PmEP{e($w}KH0*>i|pfOHY|khue?o%{?YfoBC}Zw{NFlG ztb2b3VIdtC#95HO9iVA%PWi@mR^~xuauoX))pPLScdmKKX(sXJnB#i42cZxG<+D-j z8k#xMU+PF`e8r7ech&nbRBKjY7bnvemfRNyZj#vw9&zFhKRUks$iBaj)Ry+xT$n~e ztgPk}l|b2MqMJrblC{+lb9~G?a%a3N%tOU4Sq>K*FN$Hx+~kW96`OhSTKZiyAf%m05k6~E8kXzbTGuYhAAC_NnXs1fH#0^ccC* z1+OFl)b#`Tw;6`l)Un96RCj8F%)_*%lrn9CcA8YLE}fw<1+=(w{9IWbT4E)*ibWyp zs-FdzY;p}{#WKw-@(e|J@eDIJV+%ff$1WA6e^b_Vl8NhhJM=yhq3LDLR-3RQ*R*m! zqMFVw9#kYl^6dwD{m*GnXAy_;A4hfigv?0%^~N#qUZRcP*~SMkDbbFH3*$n#T)V)g zm3h62(!;u4v$SXRM|iWj-ng1*A_D1M19G#mCgLwCOm>I5tO6jUTVcfbb*kKkZ7Uu4 ziv;hAUI<{6T@nA7I$*o>oUJ9aLX40pRegh)H3^+j(@Zk-mJG9|iL5$|-x=#?v!gr{ zLS~CExVCMVn+zEMK~kTi_Q}wO3}4eHj&-x~c2G04eiL6ZozhL!v89pq0Kdt{^|b9; z>OXot7O(z(q;HR=n3KM(f&4l+2^GjcLku~T?v1p5A$sA(sgN_`hL6g{57>R#e@BfQ zRtw!4m(zK{n1GW0P9$r&d>b44awI}h?2FR~z9*-aI3)YNReXuH8m}vDpgV_v*Y2R# z-(Xo$?POtDe=w@eOVj-`QXIuec9*Gh-p_i$$nT5%ADp*VbL-Kwgi7cN*nhlB)g~+O zY4D9VJoLk1{3KW4ggV>Y^3b2uh@}p5$(e!j53(x17Z*qS$k}1IO*FWNdDet+tSI0< z)1zvzFIVSY_>;eq%t!+iOSe>F|CzV8j9agasZT7IT0@Xi!1+CO$~?76uY@gUmb|r} z1z8^RexG`9)%_(Yjx;&7G&950`Ft3gChc;mW8Co^8Ob5U7OWNgUba#hG`g;62>XeMuaLpfD1((Ab-k zf5cGl$&q+@0DcLh+S2vxueVG`uJz2DH)))ivBV-Mw!i&rUgm#?3y`Odln56_1<7bSN4XD9zHECF=DGX* zml<=F#U5Rk!_&yY5Q0hfl3l$jV6Y^yPx6aON6H{Tot3oXm}g08hAfume)nU_`lVU= zg#;4IjsbG+3>{dw`$fLI^m`qa{qM$OsX^31F4&(7uyr*=KN)6(tH1W>MqpI?lYdg(JWx( zH52lpr!)&6Q<@;Tj54J?L77i}5O+BpTt*oikVS~Ie3ai`BXDt8mVJc97Zm%9kV)|& zJ)~d{*`&E!mrdr|-{Y_08rCy)P;>m*AmV<;AyB-%nFT zBd$~JM|FUyq;ZI)i?H@tql@a+L*x@vgps5CD&lC;C#o3iJt1UAKQ!qU z#5-MU%9YI2Vy_y*qdINmtV3(d5?^lp;!GP3lqZ)9DGL=S{v6s`=eK~G4_$p{g6hcd)qksY8o+Dh;H6xv4lNpOd2#wzyCsK={VBqyMM z^GUPp{`~rS26;5pkOJ`JgRV!qV+?l}@{3(t9)ustMXtUX@yQu|^HX-n4*wT|?h$%} z+yPofIn;o0p$$_O$Y8CM2uQcYj~e{sBcBnvd>8#W*>R+&qVu!yxXD##`4xT2w!gR+ zY3+u2!OD&%*;n4VZ+fm0Bxv4dVUlFumW6|;YLO*_*=pFrroa}jDWGm6Dppypucp$a zPF;rR_a3?Ymeo62M;y+qZ>YBa6iuR0=F3 z#qIcIh=^uUq)w2(L`oCtw<1%8Lf%%g6CbycY|#>gNaAVw0=aYE*0QQg>vY~$W>PdsLRr$7tc>9PpvvtT5 zO0pz-b;OaHum~(BoR0G02XdusQiixqd&1;8J0WfQjOPCKL5(f1+%2PXFn1AV)K>R= ziLIt27Z||qavFYIl~aW2yI-P93oy2=#%`Eg)qNYHv&S=xm69e-8m^4+bmo6(D$BS4^ifVZ@C!K#j9H^x;H<{+V_sz|Vw8YW+Sw)TFPb`T;+%ru|J+g7A{$YsKCm1o>0fNv4$Ymv>``sD!}Q98PZ}wZTE)abH3r zw(xV4rWUi;Q$dpBpGf) zc?u=9LOMhjw=28%X}H_BTn+({PA&3jjiXHRqRz0cIC7GV*LfO7_R1}vxD8x2D_6#a zqy#5>d*<2Z|3u{lyOZT}83fm|O0c{>TUpSxdhGNUwJ2T3Fh#H1p|?fE)Oh(Vi-pJ1 z?xfMG;p9^-RH7V56wNL3guOJUbIZW&^t#4vEiEh6a6?8b7OfuxbJxs!hJ`H=vH#J} zxKrDp@=!AGTTIIJd~S-K!FRj!ZE<%E*+^fOFN7_7Da68rjWr)7MOi*ihX=fjCUv+P zn5InoTA{M?yER!_&OBYF@`t|m*i37l{zoI@U>v$-OQqNv`Y7VJ*@hx zjpAzRR_m$BSI#x%`q!swXdUGd!no8j3~d2%iH90#Kw7PvrNbr7tI^ZqqV9>T!*wr- zw<0C6*@VzE58oErs0FfewL=Ei!LP0KLm_!jzH2se+6V5YXJ_L)TOU8fSs@;qx~L8Q z=%bNJouOXq+a??bnBqJRotJ%CB76G7^C(Ohh_4GeWto1Kyh&SszhS@^f1DvN9O0k*Jg4-zU1S`w?gadiUceu?W{>D`GArd|!mHw`1Nk^!ItP36pZX z2312Md9_|ZcR$Yx80P!!6SCZ&`0h+pC*coHf6keJ})UxL++0i+Rq#y z6FqJ)$fE68M52_;)tXDkyU+d2KMv3vI+$QkI)`q>Dhmz$T^bqba6|IxfTO(#eY>^z zIp_Pvb<`TGXy6x{nmAj5@ahyRQdgkN^)by$gH;LR$7rD)p_VIi-%4|v{aZ`m$j(^IhRy7?$h6$8dW#3c=A|4zPr@&={$dngqF=SB z&tL|-vfDvAHfAHU_x$`N{3AhX02Cxm9_Og;)5ik*t`&)h#O` zhJdg=?nLa6m5tpie=UjK5oRk+3QD~YXn{uQ^rOxdqLI%nDgv`P<$iCV-#dYBWxc;? zZzvaeX(-pV8T-FSA?suap1rMFkPRqx+T}-Klw2aNoe!3|Ae8)`jL5wjm0Q7nNc}xQ z@2G-a;s&|r6CT=1YDFQhzA@JN+*?|@d>XN{NP}$ypV8nv8-dF3uSuD8RBkL=#5k5rY_XBozx7JPOsMrw8A}8wl?IJZ^t2lz}qKHEO;3Tyhoa&8O`9MIkETI2nJ4I8@m1U6*|K zi?ID1ex?P((6OQRnT!flF6B`7jP)j9zvF~l|N0PFG8s{~^psfV*EhEeu;jrJn%kd# z)dK=#l}C|ei8+5x>vQcA))-zW$$SCfr+|fJf&xm_$4yr8p0`uxSKFS#5A73|=_Hr; zuh=ez2jVb)j4BBCmVJEV?f6{6qR66TvbWF@;!nZ3mSO)UzV46uQv2(rmA&VRFl8Cv z-eaZ30l}o50Pz*YDdL#VjtwzFn}_UKS?g;t5NYZMY0>+Oo?dmQdb z`2`fKQn-AzmJEJ|4`n#ueHalamv;?ARJ+zZ!? zEvyZh>_5+~)#aAUS>kP;jL_yGU>x$fv>AP@$7`Lestp=xs#(QsHO7Lm{#ndfdy_`W zJFM{*Y2;P(4;#L_YZUA~{-;^6&+AS(F-W0~M9BsL!pNlGnM{O=BUC3NA;ykZVqzDH zyV4qpim+1EGMr|=F5>u&u906NKRTM4$Z&hN3D;R^CUITKWOm#L#PIiC%A=Tit<aE6t zq!}MpSViA{d5)$q0GT^*^9*!#6nm%YRC#fCTxE80o!^Q+60gnfw874UkRSjiH$E)h z6Eg;g=O$?UjVS>F+IS$aoD0k^fF2*D?gu9)2-xC-02JJAfUy7sb~1uN4SeUm1-=U6 z*G=%mfPhgUe87><6L=EFue;{60|6qU_&~A@O&uVB0s`nexK$*8kpggU zFJy}WB_#r&Qf>rzsf-Uq$j<}JD)@EI3gIANOa&i!@}q^tegabgHXjIJRjin_u)R;< zBm^Elm?=Ld1Yl8q1_V^`0b!L`1YR~+a25t<;s+ZLFD!u?69S710e1ku)gX0>ss$iG zUmF5|)f0gSZTz}obx9DAtbX;1NKtPKrK~q2& z7zzU9%ppy<5ny!?uxtSVN=-t5dP@i(YU%@ESwmnfF<_dy6H^2D9oCR0cr6eF0JacV zS1foCsIi084O+^804{rcfXezS@ZBE2Zp?ZJ1eiJC*A>}ff`C;Ae4yX%BjDsj03!y# zfk4bh0@xW2d`bSQpu^?>urYw(LI4Z;i9rMFUcf+tIVFML16eLab!QGiAXowxCP9-< z3K$IpV7L=BdBDJ?AmGHE03dRS1U`D;Hyx#d13JkdGk@zXgn6$Lm`U0 zfc7Uy9TSZZ0GW1PF!U*SUNi0l`c$;Yp+%<_73S;sdi`t{`qeI{X60PKrfP zrx<|+f^qnQAwX;fHS8}X7B-L`c>{pr{!J1X<29xOkQ7G%>&il>X&S!9Bmx1r@dPlF z9~d-1T}%^D5swdZiian8W~?RLjW?H=8~}HmDKL|O4||`9P6Ld@$pTi1{{teacxylt zzy~_xT>)6qfAY{G!4deL41oz;!XE}H(FCx7!9kU{3jBsa>Z}2F5D=J(55pLQPdJuz z4^XB->XwqDK|tD9cw}KdfL{haEUOMa6J4qvP@jnpPwqHiBMZMNI~Lsv1boaUs4M?E z2?CzsMr;N>2w=}8Xo9l{F$i$YBY>ly8h~UQ0q^twXPCllJs`9IzX{G2lpuh-hycJ5 zp>1*+$G`vqD8&STi4q+QI)#x0gcK7r#eKnGgGZ!9uk0^ZdU{1fv~0KJ}|DSj1$4g`Fzhc8hq zm=?xX4K4tX8VG#US8EBmO@~u?9Mt2;k}M1ZcKE0Mdp~ z;GzY;uDanh2)4zC312xJtXbf1D*-IF2~5-En~Y%s0<_u)0PB`WfVLe1J57Oa)pd9L zGXyQ{o&{45Xy_z>XS5_B*#-GGLX=hLNx)_oK@-bY3^=U@cN4;JGch&*jUIx3R?8Q} z4%_s_fcyu@-d(nUQ!fGRZ9RH40tefFcx{6dv-6)y?l1Vo>GKmg4!1R!>yf+h1}S_7TK1aM$F0m35?7%m^?6o?}w5`zHtK#tW8D3AURWoQoD zgSdd6;Q*L24hREK@<+-407uBb$TS<}2El&NA=K9?jDy8t`SJ+eVc58!CTv`UQGzD= zb_4|&h@(lB1bK)a%)!A6n?A#!gyC!s??CcEALUC6Zgs^gT zFt<226&sb>I~^4>J4Mg8pefl4YzIep02uPG%E2rG!Gv|V zA}k`}PUm_CJCJ^9_Nw|D#x74lcf?t|0JN z5KK7~{9jq3hVuoe4f0=MN(OQNS9v;N-JjuJn1+H)K(NP1@P7ru6$A^50u%kC5WyfA zSq$81xI}1xU=^`&u7PU=>`NTn2e?SUOcTHa|Efe&M-;jWaf%k)pLlDdy2DkpVI+(x~Q{hQS34$?Xz$b0&<8B00iJBne&R5bM)*wsu;xFUfWXQAA4E^pf!RPX zzb1Ha|KS7pb%5Os0{aK`AWlHcAJUZFfQ}4;;kLk?htp692zImq#)M^1W1<2=t1D|kQ5Q4JAZkl+OwUQpl#waHNptwOblx(z%`kHEwJAKd;f zLQ+KF;rIu*_$~;XJTN$}|2J~g3(y24{x5PlIk^6R$mQhuPmlkHTuvUI|ASmk9^R&@ z0<cD+uo?lJ6K{p$khEx|cEtcvc^L-ivT_oLofyyNem47$B~+AibjPq2&Y zivPA@Op0{g`Z*37@MlGeG{JM%aFv-gUv!p>QXBOgLudq3WW03dsCCu1GX{{V=tr4F zT#ME!(7jNHLjTODtBs74*!$@%M7{f<{zWMSu*5jJ4Cgy9!5`V24V=;3z33bv$ND_> zC;d@AdzE~D)=~t?eIKhfzNI`A%!HJdm5G8nBDwHT_@-xVWu&Ls zN4yzqY!{lh2-{iwrayDnV<(7%jUhKh26`1q(RK|v9)oP6o6 z_iR#ywrcuGK{g8L5{wo{s>NOilbJ0VueVSAAs^;c#UQNyJe3lIN`vt{-MJQwd*iN- zlCPuEJ6`3W+N4;|^CbU1c`eRRRfn&hQ59}$2Bq3JuBx?_F+N|^y8;QEC~tZ_+{JTn zzQ!}C6*Erwvy}B$O}B4bW&r*R_VH*I@<8fFqKD73nAN&es%ggsI~G4tTdJJ zQqZ8W>cI-+)!g_TK(w5J>L}f#aR{YShmzhj^mP?=xc&+vKMK~oSO)rskGitAHeJom z)X;*DCRa+#6ZN}H>I)(%%$FOZ7GHJ>hXfBEmho+qV;vmt={^qDYMiF@#V`HRvB#K> zbGWF!N_*%3sq1RpM&{eMAC25|YKedRtbRb7p>)Rzq$x3`X0~SXk4U&2Kj#lG`?2Ho z<0rH&EQcN0#|ioRmXS58;!FN6HS_l%W2^BK4v7oOuoV+ubnxro+v^*;_(p4poM+V+ zE9w2l9_U%(Zi7?~3*{(pxlNkjwxe1{3x#kFGt0jDMO8na@JorDX^La2E%H!yn0yEN zb5fg2jSE#aP{HN*BDGG>jmol=)r5FeU3}a`rl1_|ggTDGWY%lCKR8%XvX)vS;L0eC zpp~J$-g3W}CojL?&Kv6QZ3MUk^DTz>5w}WEFuWPstjQVlP#8T65HB8UQRi%J-i=06 z4Cv<7vrrf=I&tUv_=FfPle+Ne zF2SZ|xUdjJvseMEcn*>``swaRky!C>9tDH%hHo#h;yFT|=rC3I^rs?MeDK zFSTDBq&M?OK4_p@#pAtsWwEJ%brn!-<$_qTwQ_H8ZTK)2Jt7lLZj5@f#+x!DC_`NJ zLJX%rf(T)uoH}48gz&o`xuwD!BJ`H=Ra@N3V0G#$hddSo# z^yhMX_yLpFb=y3+007P5Qj&B1ByJwbR&SDnboZ%><@#aQX=$QX1D|&WbaP&$!=CHd*yhK@qNp#Q- zL8t9|SpDx2v=gRs#-XS=Ria8LW=j?>r4~NVZg<}XEen^mzS(Fkj}if`rFp?MSAz0R z5SI&!ywJFbfSdX2ra(UyP}EY^Yp*m5N`55q7ooMHZyXCUjCSQRsm+r_=^rvLMJ^Vo zld}VPbU#YsW=){`MyU_8T4w}R0^sAGo|EfUk=SpMdnlMG69`5>NtGs`Hc{y>GkIAP zlw-07cHH@UUko_^G96j&sl*V@_YTjp+E}y7n9eN}PH}i<+iIL!qi3vG4fRsK1*B=W zbtrH$4wQvpz0$3k?ExBjZ>5DJ+kaMQ27P1C!d~F7EicpQFWL@dep8Y}!JL~XH_XrS zljS|t7wRX_cX8N}$OTC(R0BEAIA%*oFoHHsvFwNoEO!|ZB6p0-|1P$iXpC^ete86Qe;L=BN~Mxn^PGtBOA6Od3=cln)H zYdiB+FihL~pxo!apvbg8M&^y;t7B;dr?>optFFX>ctIb@N`A^aM?+N>!I#Z6qrswC z!5hmOX$oG26(|S(CY|E4Z}sa?m|+{<({mNV>gLuw)_9);y`E6s7Mc94$VujjOeWVL zZ<^^~@a3{IK{vE||HKq_(Vf)9+x^0ZT4AZWmAk(fD+4j7KgczMbC%nCY6Rqeg;f)$ zB3o;lZL>HO`{&h(-cQzM;}RTMT&5MuQcIMRQ>?9o_VV1$zUKEOGmQrg>JbPpe;pJn zC_cD$)DAR!#D!7Lq7g#h*f`QG4L$2o;UuPcm8l9Sb5fa=_qHMu>O8L3uvquqScxP$ zj@jW$^_s2?;~b%EDwSA|!)Oa~fU%pT+{zpl^>K7{W$Wg}DF5by+wx>bM)2jfYh2*0(dDOG6W)q`vamH~Qp^yayA$DT zll9N6+zwl2BDbVBfp5!;Jx9MLq&M|Ve>a7_N##0{ z2`gp3yV`GXy)_9*?1G4$*EQa9e2GjE@%;&Y7OIqrtD{h4u=fB`%+wTKjz*6HX=`fh z>O6tYj!M#sMP;>n`3eQL2>Ohz8eM-p^L}ZrhQ&+h^7_aZewC{q{)zj=?==dk3yJeY7$YTEJD*XfAxfK2Lh#xO&;Fk79()o+o%{K$T7w8>RW7e8^p(3y;%I zh_5aRS~AYJMkCSD<+;}~D1eD3u>1>cdm;V86=L%86XzLCBEi-RIcebtYUl8td!{;? zuBWa`m7voX4`0I>w^h>5Q18o>N>HP-Bh7fJEe-9v?75te55*kBnJ#4 zYcR37l>Q+L_OYmM(CDXr09Y>bx0W~%$R#QCp)=kC(}br02x@pM%ll#;7 zLuHrhRt{vkH*K@B>o3p*Z;ifKuG)w&4>+5|nhZtNdEf86qD@`weqh(yr%qHKQeT?Ea>I`eJ~&?(^BzDW!H|?1s9u zj0g8Zjg>Ch?5dc#-uI)}86k~<3bhpCR9qd3Y_hzK82o&Q`K?@2K%S9(&ZeOKcDWnD z;G5DcV~wR*^S`C?UIsBGGrWCP8{?v%Dsb%zoZ1UJQ8GN#q$SQLs`FC(3f1A~i#~2Pq0s2wY$Mu@5yu z(?)l+S{x_KcA1;ALUzituQ6fMVH+d+gD6#SD?^m(W!*yDT9-7HUx>HI)8)%8lkIOY zE6{o`)cDm1Tc4{MwJq(nJPSkW;r?;DJEioTC(4AWq28XWEv?s}nPZViwft(}j<`MK z)7MGdiVRKNsBfif{HNd3@6Wz!H6lPibqX&v(&n7HguuHj}*Ey>(o#!Ok&+XZb+9JMRE zsLM-}VCTB*vZ+5K3OvhkBF)SVTvUy8Bx@cLj4Dc!v2s^Q)~?kD?}kHXNxTiA`DknY zbn(WwOJ`3@JffPNNroMof7AEMAZFZ>#>1D8C>0=+exa8n%7IxgyORWZDs&@{> zdqm71g-K_ov=_>Ini6$;mr+clq+~tZ>T*ZVRj7tWepZYO&wMN0teCJlPI)+p2)=&s z(W>R|WNjSCuIXi6y15 zvxRQk@+^?Eca6^eZ398=0}J+E>9_@apK7V=M=g63y89r!xo|9pqI#hRO=E6M`@!R7 z$BSSyj!>2|Q2wqwtQ_kR+o$=0{E~n2We;bjtbA9vv7C=>mgb-GDX`y0gDlcKo=V|g z9jYVfTiJx%{ijMJ%tK;l(=}1o^V$OkckEY*7T;n9gvHCQ9TLHrDdk8pxjCdDmLZZ<<8LF5 zFGlP+oU+yx?{-!Gf<615GyXl4L*V88XTRd(3R)T>$fpTy15FMGk>XfL5taZ#7l87x z{h!T?OauNZx)v<>=7o_BN|l!O;S17tY#ZDgW8D^wA%`acv9Ip_GNwqLtChBv$MpPi z&@QKoEEtLlyWj4jyy=@0y`gQ1BN$C_lY1Ea__uWO+3@f4_4cHwNZ=ojztvXEMX4F0 zf&Pp8*(aA-ql0mSqEM%SA3vkdz3zCM?{7^NEB+2#Tpy}e&>B4PkeYi6cDe}7{B+S6 za_nC$9^FnBG5yl@rqR^)?SALxPvLsYBf4+GeGxZo47SJD)HtaH9W!5=8dkhyOCL^- z-EMEbXlzOb9R23$@KqZusHwiJPtxJUt!a*FlxM6R&vYScSJQ{WPv?-%n5kZnp!zAk zLv{Q)_?WyF)3mzY*cI?}JovNq#ZzP0)|VH3lhj{jvS$m~^N9`?t*bYYIeo%tv}i)K zk56sMVobbQu+qS#Cx^-gB%`PaeXaoy5gv@I=fbN-v@HS5iw zwq}vEWPVKTbUL1emYw_xd7Fe@ra`+p8|ndMlE2}Z#^zW{g|+b0p}6*XwJ{~-Ds`VV z9O#^dv=u3DDj5@{SoiemVgwjzWKT61mEp9nT{XH1fqULM`!jB!=CT=nlWbDqI0jTO3Z108 zK$-VVH{To|uS5Gd@yBH|_smRgY>#4vLdX)y)G*JG=0XJDTMI&KjRYKA2v92s~oG=b-kQN;J#Y1*`aO`SM7{+>Lk79Yh z%<@JvlXpmSWi>P9Ll9bcpm+!{F=) z@whgGa#RKkFWI)J>W{CM=nG%d?*E~3^XjHpFCL%&2u*7`K^>g&U=JyC=S2Oj?UOP- zn6drV+ah1X3ytp6(#Ky7>73XM@|I+J>6j1X?nrCReGfB%*NhIpk2l`61`q`sDFQAU zH1`9W58g4}f_lqNJc{pYZrI)LIii2(z)z+YWNhhsgwKk-f0)xvQnkN+EBW?EPc{cr z!jxEzi6c~5tV=}rPz`K;8aN?KQp%>bsEpK%T65F0uZ1p?b4X~BYFhP*EUBg*RLUE8 z-$r#oz2Fo^GqN;Ij8G$s_O*GV-lhEBXEXYedwvmwheBxY=abF%o>H>5K1+4v#Zc9fSUC6P&P_ z-;Y%6!((Gw z8+G9jUsFx>=IX{TDs1|wufpc|u-g#UsxxXRNt`7O6|K_D#dHq^HDpWxrPaLHOC&CQ zALZ*){j}r2{V!|BgW67v`e{?0_l0ySu55`2FRnKk-@N?FayC8x0eE^DP4^1ly+Rl} z0b=#HrszTP#GrbmvyBGV^nL&1eC1`xqAX~7B4Xjs_&sU#jDeD|S@bLH*znGT-E47k zsEP4@s{fm}hQxtxLkXeI2CY%+yVc(_#P#y>^G;leLq@COp*X8m;)JLx9aPKIsK1!U z-)7I?gdpE*y$1PSh{c7%oWHH37u)h{xCTa^Dik_}d;1)a`c*hbGfr4x`E+`;I(WOu zW4+nCGwf&Jj0`Uz$93akJ`E$R`f?gC2IbFjqf~kSJ3=bXX9Wm17aVvd+M|E_*y~>4 zSyyP6>$yu9f4sJ1+P?!^>U_S+F?&vy`;1qB)Q!PQt|P+CCn;_`G+UFC+|vJIw&BI> z7ZUBX&yDdae>}4u(Y?rPhQb5s;Xn5Q^|)?PBzdyxkHaPyjEd4k?T zOr%r{i~WaqV8gWXj6vQaQX((n({md_WKBeV(3ecsb39zS1Z6yWp7~?NXMMsWPiVI^ zZNEiOHXD2Grk&}sj>bd&-URj`+eW$6lROVrxbgvywf(_qEg!v={rH_|&XhTG_{vD< zVu&cZ@Wex9C*E6Eb__RsJQ)wt6E-MrYFMR}DzyFb zv#BUS7@teQ_vXenC2!MmS{9Zwb$n)}4UHJ^#GA?K>>w>UTAS1-5;hr%&T1%A%T?)Z zu$&2XXg>IHxc=4HUU4eHy2RjDieE|Ar?>{=YFE>?3JsA@8!=ZVy9aFr!ie0bLvPIv zc=r|F=$_G##c+wqpV-zG+P99wl!=t~LYH4Q`=gKgIok;E7r!I#Hy@Q#>K3EN( zt!JS?3my2iY?xK=&Sdhg zZ@AOnti#NGY_bknb3G0DR0YHm0cTC+xGD9`0E-rs2IM-#m-53UBPjdW%=r~x|7)xVyg|~?uFI1t2)@p3;t&-XF zaCjDExs&NNG{Sp_>jC1ID_Uw3yaQud%+{z6_p_gh%EyAqCZ6l9`v`aAEOcc;?Zoel z`jRC!_I_fmMnAu5ti!5nieA1v9%&=P4Y3k$14CrBQ-+5qpam+hA$7dQ}I31M9eOZ3@z6?L{TB~hyN9>^16o_6A zGs9^{x)=Q&tqooFw>WT2LTNilZsS;WiPelS`+%3d$hsNu4VW?7>!l>+EwdcblX_U> zji&ane3c$-qBYvNaC#@%Ce!UsGvtBIV{kC8=1B7#1pn&fadsZr-zw*zVdYX9(($hR zXR-9zGul!z#R1NZLISpRl(h&OV{j_~eRm|>RTRm}1^-iNU z+F!z0I!)1KbCBC&kK<+${}6(oryzrQ51peJJMzx+FIfnjMU}`(lnr;HTx?;4`HFTi zYCJk4+^PNTA^qF^ouvlXr1e|RVWhCy zP8#+s=X_`y(_-fS)jU#!lsH9|jq{qxncR99e-#;~C~s9?LkRHNWI7(&;>%XNXtvrr z!N0K-F7bD_=l6*%S_SnRl*(w;D=zhxJ-^Hj(Q-!}7;m~OQ7KWJCI@tpJj6njV}80< zXEUh1^6XRqtxeX)CqaHNO{%RsNBT^R#^}f+UKu(q7|HNubqK%xs(TfT-ufNBJ!1VAB(<660;#h#2Fdf{H{%DH-dJtCvxKZvjI!oV(o;m z#0M!0(if~%B^cz9<)iitxSmK(7pY3j#;*?7A6RKDZ76gbCu6&hCrZ2P)^les8h-JC z`nRA+=M|C9CZjU!B-Q~fSDE;xRWdyAuI41e*QaPX&HRen-k=8#6r{~$|T zY4NB=9BsI9!>k!)F6RCBS=78s!)Xu|`_c)j;CPHbTAtWq@ZRYC*^)u?Tf|Hz)dIf( zrhVSrl5-EPkJ&{a`gszxR}MTH^guXrUB}@mp?v3*_Z(kinf2O*dd9#wyLbM!&nL-P zJ>&ob!hmH9+wwh)A-vGO-DktC$>vM-WbY{q4q%Us?7f9qHZB zP_()6)iAsX@{eAo^Q8|5Kl!axbc^R|COgGPMH<&KcCX^jm>CqGZs(2!%7P~f9jpEh zm%E;Aeh5z&xXX}S^!v`OjhCb-SBEEwhx~Fy5Ph&px@bAQm@p}p?35h5ppT}p z#uo4B)H#2WQ2uU_8Y2Hilk;0t`2~GQOI8SMdDy~yIr4{dyO5&AOYx9=x}2AJkSiy< z+4IOHcTV=WmA4&UQ(1IhwX}P>`9(Wo4*oDi^if*A*Fnz6NR7uicZ%K0bW6>gjmice zXfkihd>*^}jQ6={k*m4I?iv{-9N6sUAp<-5D zhcVf)w<}Xt({XYBb>zWbN!PE$s7MQUxkXy0s!QxmnLfwwZW3WG^KT(D7<5BFd03=2 zWl>Txzy7xDGMPh@&THmloe~wj*OC|R7&d&}2YCWn6C(WIC5`lxTl{MHps)9&5w8^W zps~Kk_|ostxAdf6-o4UE=SCbFP@bv&F0#)9hjn=?jw$;08>6gO&TPlw1V_<66Ca0v z6~=67i|+_0pXXit;5mFQaw&F*;_r7pt73$)=_{zZ%1ZLq3&Tu4t=X> zc}L2S;RvFkc;opb!B)Lmqi-_4=n`W2qngbCfrtW?I^0#I_fd*X-IMwQw5#9E-@9P) zHSwh78H5*#p17c#J!^q;s-`~RGIFD2`C?+9kUxdhzUwv55%22~TosF%wFm-X221{q z-#EJ{c`AtSAIBRAp8N;*%_!BXOb09PNc5@S`Z^YTe_YND3$j5cuiNIM%1>-$yD+A3gj$iapvQYEVe8 zp^nP(y!f3oPxhHi{&ZY7PP%VGG}%OFLKc#&KEWCoS{ZInDI90X`c5Ne{@o%~^Da+E z-X2@)B9b6WvFDdbwMNo~Kj{7j52pQe;)E5D$0wY>;N~E0#(y@A2>yNTrD+s|&WW!Q zETZn(Z5YEJih1`kFQYb2Y>HZs6noR6{exEgUuQQfC^pOSi-?=YeEzII2qUS5qsoCu z)w0NRW@D{uLwykiK~v~csh8)IqED1s%S<2s9Ut`1jun)PkMIBdc|oB8|3^Auw}0NN zSq{FZ)fEc}53*BOf@;Ez4Sm#C*`|3m?RXb4}d3n?I@n+mW z^oFISiC{FrO>XmK@vp(;^TYkqViWZF{HCcZjMqNaEOPkI$>9F=V@Du%OZ$`eGjz~4 zT4eyy?Q!`7|le8>X%0taUmxnYij=PI+9wRE3Z-iaaYNlpVKZJE0@jchr zE2Xbl@m%YOJRc8sHEf5_-fi8VH+U&G;+m27;cQZ?Zn+~ z*kjzqn4j5U#ZKRR-vEvJk@9*Myq{D?nIRdOOZk)3bXYuvl{s%vEGl59^Z=!0et**w zFxQqVed?!GTIVr7nf&>t?f(OBK#;%DdOsD@sWXEZPZSl4Pxj|kaECmt!sb3t8}ETT zMXV`r+swJI{^C4vmBon{nQClbCbM3Jgl z4bW(P#f>tehUlad9#;%bja~mf>-espf7iebMg)80d{i&EGwCGJ@50@ffF-7B=6&1j=$>C}7soR_S_TpHOngHNa^D*wSp;f{{N9cX^!OrSd)>~&He?>;( zcB>wp%uWpMeLd?Cs;G5VVEt&eOz^kAJ3E@ul#Z*ltrqy2=pX;{YW+{xa{o)h<;9S?%zp4&z0y59TDbxoZclExqaD(0j33estlk=k4>33z(Xw z{H^eN6+@qjl7#m5A-1+| z8cX`J(Igl7b-H&*;c&@1*va=-#(dnJuZ;Pgkj83Hk+3aKv{UOmvE5Ib%3%xS#QSN? zen0c!v!`~a@JpyC=qA3}&IS1>7Qp`TRM3AzM1LZcrfTv>cFGwPwkDN8fBKzmnqhK{ zvPDmESR*fJ(e`ARWWB71u)_m~VeNc@^B`+R=x1iV({T5?CBfSlJm13Cb)pr~(B$-S;ItcW1jRks)a1#miS;LLublD#% zS*-x>HK&c>3$qb;qG`~Jf1&>$``l;?Tf1&Ii6%Q*CRY0Ww1e0EnCTqZHQKk%3&kG= zF5P~|3#7^cX2S~*RVYP1pLQBV3E*abhjak1M^)}ZH2qm7YUZQHr=;@e$eaJh)K|Gg z%A3aZ%6t+ZkbS^rgByrT{uIX>pHwT%Bxk)K<+^T2&>T?zR_j>Re@b&gA#G<@N3iu3 zw4^lcOm698j5>8t(ndG%D&$hW3mEr>L~4%_$6eUTz;MhxAo-kwq}tL?jyzea%kI0Z ze90!e52BBHu*S@^PB*KXktj5qtktN)=ja%dCL?3GF`1|jtQi%;~Vj7B&{1jl~lRrWvlHUZPx|i?1dVI<|I`;vs5UPYC zBmC}pwy!mRM>`@)t;_LL_y>-x)Ydv`+yNa4*ePRkweTj6^__S@TI%Sa`}_9t;o`** z8x_2Z9UzoHW60j~W$cG1-CUWb0~;JLf)TUFflu7;tKH8Ue@;jMGDIeXLLu_N19g(q zQwUl~@5l{?dU>9@ZL1ONp^c}--ylQb@;#Y+4g^0!I5mpY6-6lDP`47otzp_86Eh3M zXlE8}PhQ)=K3i2|X+}$Y_b&U$6pBAI^3>T9XZloNA@D0Vojf+y+eLAM`y4?yw|J#O z$_t*@L7?{Re_cvH%qF={v9Yt1H1KZbaMo$B4^bjYrHT4>HMXg4Z>kcJta9Whryl&` zDtZ4`ET(5I`;cB+=c$bi9k9cB8{$H?RK>I22wqb^>$Nh=x3O@SCrp95G0N%DPm@6! z`A&GOsSm%nP~LYWO|}og54DB+L*jtIe-prBsKjd5f2)5YeUn6Qc!edR!Q{w9J1W2R z(eZc+Hi!!RSzsIG()SP!`y0_CI=T|g;-e0R$YCJ!*8mOnf63I#C=%C&o$t$j-d_Ex zK+W$s#6NFf#-w!CsLh++FyXc>#}P^k9=EY7Zth|Sfz z0YCM0f9-h}d3XLTZioA}*>BPAY=GQJK(0E)3aP9o2I_ z$j!5KUTWc@IG2R^Tscswn7eG zfBPj>bDDEVx%DT66Oang-nVyIiF$NvLQ5a2H)T=w!=4jL>Wzx#6ugcgdX7hM zrhJ_v`%p;sMj1@<8sJF27+@+{_qCvEe@6Ss50)HwJjIgR$>r8*tJbmshLXL-%x^vV z9u08@dy*3p%a(r{62W1bel$=^vR)ynlC+j&^|)Bc#;N2c#+tlIX^AuAsocdSd8^ai zoW+@O3EGIdUG4OPJ4%TQ7EW$-^YD=Sy~3^^l=NFf4xbopI~ADS-jWg^Kdu=sf2Yq@ z5;l>?ywCsdbN|?@B?zq~Vww4;sCP{A>h)6ZJ*CWrT~ys9s!)y+OAsLl?j{Fh4iXwJIGL%H#IwQvyx=i?kRk+^TDC`wWu zQ>oK8XQozc=!Vf*GNCB52OWwSsUY5#RSc0D(()8gzxOs`N~>kF$USJ_NMi1y+UK3S zP;|`d%sS5|m%{zb6bCxia zzQ3ZBG65kN7Ws2QlzPTK3N-W}0^($MZ#W$YHL zMX}l>FksCqByYxGC?)`Fe>dqRm5%HKk5yOpRDAPPnxSr$jiu(wAFF1kBBaZqX;P{rlkcjX^CP?FFqDU~t$Rn&gYMCf=3h7ini zpcPdX3TI9;c=MpA6dZcD8i1C!Y)Lc{Wd&&T{$J5IhN&Z+N%45&e@6uXG4`C(rlb&` z4zQ!q?d`7@l~)>gryQemOCcf^4sy8as2vG~2Knk5n%C=KQG++}7A!?)y^^q@gIYs> zq5UrB>9GD9VbFf!TXdMq#|6hu5^i%X7HOLmBG!RE_T_54A|eKlo!VC^)LCOCbA(cI zoHSP6N*6eS={Ou`g6)3Hsi(%#M!~{^s-I6=Rz0YF?!Q3PY%8XT+c6$c zF2{U@Ca0VZIT};Wh%uZZDEJ7*>+H$keA+ab9=5~%M6iVlp_`I-eL^!A(28)~dq&Vbl2@mKTyo!|q=OZ)qRS8ysf1`%MKWWegtT}t$*S@fVa%NN6 zBgsH&vBjP5?Y6IV$F3;ubW+>b?XE-%Et@MAQQ{hHe^Vq)+ml6tadJ5W+IJcAyH)O- zBo5}hO`z^;oSaWrSP~X}jxcAQlh34e17(Psmp#|Q?fi0Vz0}%eud&EHR}m2fuS+yo z1+)?2B{o@Ggl|*{Xrae3cV6us#dpGpZ@;XvGb`thNuzd}a8!V73#op2ZO7bR+cD#{ zDDobre*%NO;Mi7&AEFpFs(AshbAc0BK1voQ<>1y9W^3$5-~cy8i4GVYuY!{>dYI)d zlHQ?2IOcl6fmR`vvG!#-n}!&yN|W{7ncAs9WpMlz&h>69N((XAEgmv~P7EK1`JTmL zv&jripzzxvkBrqIH@7FyL7q!sIV)~oiiCx8e-qXbqarkb=*^r9SrFHP4!oYcZM7h`0>)gDXx42sNLjF)N=X8JK zp=vSOTdBiaFNUN7(o}$@%FyBxvGKKR=B>OGo_`Ubj#~l$F6(V_k-C5eto{(U*dNsp7`&m+Of25Z? z?_pxNo$oHYvyM(FLAp!PshEA*EHwjnr^taYu)nEr8J%Xot4wzx4&yArCukAUiAQY9 z9EhLl4kV^W3qRyj-*DcSc@Dun(-NPnqVIHFdJHT{L96xM&JzcdVXC^hPXu|;A;PpgE?PAM z4kTsT@7BPPcPLv`gH>eeFYr~PCQMl_sN&NBE*bCgm^HzHH)=HW+wbnY*uw?x2>GjX zALtG0T9_1*6n*V9$jp~eE|;8HH2|=t1jR4F<|@#j&`{Y%%;FYAhtU} z84Ad%Z9^X!6bv$jN3stCb;>I5p)#vf1KqTioze(CVfFOYN@_ov=xvWu3sO=AKMgdU z3I#A-MrlI6DqhC4e;Jk@r5E@JjpGcPO@OIh7b?nuUCk@sJMHH#7KQmiX$fIseIU`O?4wf{!^{cz=AgE zeZKs_7U*ROQn7`6tJYEiZMQYt-c5dH=M(*XQ8#?%%}!|de_RVT-W4|i*P<-cY74tu zNahui;kdg=Cig}17dR@m;H%}i@I{aOmx7PJH(%g(y(H^f7f6W)<^VtrsGtS@>+AcsW-i;E5+Wo07cF-K#}uHfZ{H^-y2r^wmV&7!I*l=cd4vccHcM`b^Yg3 zF?}h$FSZc}NMO4b%ZpOg-u4AbLzl53FPX%&-j^@uUoXm+ts9aKaaZn(b5b&YA9S>X zbIu*bUe@;QvVzCu;4{MQf77+OfB%R7120F?o0su`1rq@|my!4lC4V$IFg89u3UhRF zWnpa!c-pO8Igczka`yEr+SD2^m-hh_p0R*dfII=ulVnUyP0Ny@|L-WtU|b|KyLv|$ zJWN+6&tRNi1oPv6KkSD&|Fa+2edJV+QfiJLzyHtq?7z={|N8N_Umt({d;GBGsHU1y zK7M_k&wsSvvDA-WuYdn|7-N4oCcKPspFe;2P*Z1HA2T8a1B$8)!jeIOw6*M}PEkIBMwo`T4_@#~_9x zU0-hHU)YTn@q`|+{QR$9|Mv0Y@YS8?{KNVqhU&)QrAF*em^Re>;C}82N7CNn5d$%x zP3k?;SX+COe}kUM;8=~k^AYtHRTIlUTaO~hJ=Rjhp+AuV=S&G(AYJglTg$dO92r-P zvpF@>N&db^*?+sOm;M%q*=cefSM+_}!E9|cxXKvtaB2Otn=%f*Q}PLIdtatNJEi?0 zwN3VoIappo4(-D_e?EB)W`Kz*AF8jlalgUsYX<3@-TNph?tSzqeOx5TlO;3^PEIrz zkwy?NI6kUI#9+haoQ>yUzA#uY(_Y`!7eM|Oj@HKR$A4p3A{(2+sM*12$fkj5`(y^B z6M)^e<1+I}#W3Wx4b>-n2`*OZ0MsIJ^^5r!`PBO`skktfgPa4h*jmO7NB6CJFONjL z#n$~`UpOw#ibTcDqb#9HKeft_5c*C(xn)1dFV=!=Mf$6^&o7qlc94xpU?qwTds}3~ zF4u#;U4NrE?9=wj+UOkZ_qJNpNc{}b{s4g`&qGXkDE9aqvpA-E7rDU>{oS+m(4xnu z*4P)`@>#C!>B-m~YKT*_lMukFT8o*V8ogYxC=#dIPb}rm4Py(;B4iKcyc-fX>$dLNU9cRDZGH-ogMPz=b)U`Ka+K=M zvzxhG@IykI>Wp+{9DA#sL(6bX3_6egDSX)%b2+p? zqO+9z*~&}RJk&F-6#3fBe4UPHYw^s4S${g`L$J)xXdYwheiHZoMrvtjA+>Y>8$4U- zF{;$U*{E4)UP+if%a#LeDu=N#GV?c^6-y9GtRG=+egkkvaN&sl6s(Ck*3}B09Y^O& zm1^iyTYK_Un?F#ERIB|Lz6W0!N0W5pc5Q5=F}SGap`Ui@iI6eTh}uCk(vx{k&VN?7 zBCp{qJGtz|4JqXf2Oj#91(8tRi3%aUs%~1T@{QEyp2Q?SOw-BvDMBV;wEn-4huam3` zcWfi&)g3uyVHCh2Vd)Bn1CsXKK7Tk209hmjw~g||NIh5?BlO;WnGfFHI6_@AUG+!E zsiyhF7w4?R5xm1+uo0N72c6sCd#^&uX_T|6L1~a+&6t+06*^*Zam~`0HoZ9H@rq4a zE7VD7UVyIHztO?x>+wm)RBV%9?7u@B(o4_D33H2JjWH-7R2CKAqRS>3*?)b`tB(Tx z=k@ZEV@2u!O zEOv5)i$LBk5fQTs8M*AI2!AO#wimeU1qIBf09pGSe<|UF_c~ z>#!X||2IT^Z)xJ`YCD%$dwyzF92Wg^P8l(nO)tTp$;{qRTmZEM$JKl^+?#FlitU6iEea#L1MhJf z9UYbd;-NE^Ej6WkCF}hPDA~`X0HUx*7Ti;g(K#>8aSOf$z)N_JRUVAo4h8|Jh^5&X zt5tp2oe_xU1cnX#)_)UBSh`7~EJk$0Jayg(DdpLXN zjew|og(af05ng^b6wq|!TqM9xk+5|Sjjbr5?fv(QNDJ;Ag%8cKLA(t`oG=Eo9I4Rn zkX`%n`Ys4VP0KX&|;K@+G6qqj&OY|ZbU>9D8@(s&)) zG(x_WhDJkS7k_iw`^>1Ao@+5qz0aCkTgTTR5%dEH1d%SBrz~zLwqNqkRs$TvV{_ZU&%BwIRp%g z>||7esy_>1EXX?{&LcwdfO7qWeUO6pv-AWBIPpYbnFv!FS#D)R(sM&|N~)T@?y}V{GG5qpx-iXxI87SB%Jy zoY6Q$=1*;792JIv5ZkHB5uPXzyk3G)o>|sI@~ zR60^!d61yajW_J7&{yM^GJ?|+p3ABcC>LGx0@$*C+JR}X3k(Zec|%MSO%%*0^OTES zx>~glgoihl|i(g!-k& z60f_8W2|wq8$Ol%c?*zG$r5A>&?evvGIr$IqKGVKu3kXf)&o-4)jn8%OZ-ZKGIh65efxzXjO}U zx5faP>OfSKWKsG+}<>JUHvLJ-Ksc@L{ttznIvehT3G+md2i&c0P@QCEP zpCJ(oy){{bwS(wH z>!GpW$#h;@1aue8DSte56dmzi0SUr4OC}|UJ9P*3PD+lwOp}u#&CAuH94@l71oRep zmyD{KzPSIQU_{=Myfs2qUP>hLqK)2K;lLJ81DY`(M*Rpd#%p0YzH^N41~$&VTU@R# zLvURQ-DbJTH|(@7WQCVhKp*RyfV9xOE+GNomd2;La3jm)=zrSJLSVP$;br?l-4eNm zyhY`VD?a(fq)nv&yE@j8QR4$ZuF8Id2B&+pz19Eg#YGtM-Qc!EJxfXFSeCdsHMWBp zw-?X7&pRZNFWw}}(zarUCq@CFqLUmoS9A-uJtQ6a&t8#^rM{UW==h6UQJNTL{4oYA zJWNUpG)pyJ2}D{bmw3)!EEiSmajwYNE$DwPP0JFo zC%*BFC0k~i@2X~0-s2lBH&8oVd+^Jt&tI8(pi-wv=eZkE=eI?>GEdMkQ|C7dZ_qukpGz^d#~fYvEy-t| z4Sk?&c{Z;Yc(T@H;-+FO6*&f9g%Qv`D6#1hdN+w}FRsSPm?I zm&1kjx4Fm`ua(AEg(qaF8bD^3P~nCBx3xQQcjJ1t({h#$*!R1d4ii6#V>@Dz=!EK4 z`*Bl9HQwX3Hyh1yYWTYg8St@}H*tQ_Goh14exX8PXBEfY7c1F#>*Hc)N(tDdrrgS_ ziQzRRO?gYN>+d~@IqG_iyQ?je=L_C{?rtoK&gv8h$ke}$Ymv|Gw=%dI|GElrDD$!? zm+0p!^H7cd=9hYa^DuRj6cR00x7PbF5kD@P^x0fr5#4t2>(Elqve<1gLH_vEXf_f{ zfS_}w=naZ~nHIYGLW3^}LEVxT)xJp#mAHc`N}5ZF)oA+LB>0w+;$^9hElFs96{^)Q zNfKRP0hE@3o6lB`iNr^K=+JUqJ_$f8l8!&g+w10PtiXqc z8eEjwU*o@0*pBY~d`Hbgue1;deCP>)DLMm2SIjtb^^_m*!*MQL(&1tEvsM+0Wx-(S za~5CtaIu}rEmS}0V)EOd_5(3)QDvh@ zn49>r+$etdcELYXlr_Fv1);;W+50`i0E{0a1Elar5CwMN0Yx(xa$d@Zy(yjOWV`U&31(Q6I6gi76l9$ z$M;@d^f1~pruBM%$?n4@U3_%V%kM*a33aXap8chwqMPDtc2_(815Q6fKFZdBvfk)letrD=Km8xS|De8?@qh&r0Xdgp6bdGn0Xqt60lv3vI|@P(1S5gc z9hdQd1rq@|kq{=g>`V$^1ApZiB)hm3h*;{z(^i8GqUz}PA>?+HA;rbBhLrl@*0p^P zL;9*t1INY|@nSGdK8t_vM5W5CP)t{*!^*SYl5-yGc?r(A4xABoXV{veV^fGgR=Ng% zik5u@96+o%%~!o3)id_l4mj-An)+!Z8zXh36Ca)YFblXVjUzv&`hSrxJx0`{D_g{r z;32Tz4Q{wwqvy@QuILv;1?|Shh$)~JxM*00%M;fr*8Vv6C8K)C915H|2z{$kR8`=o zJz6pF0HY97E*0$=c*`(nTAhlx|IF4cLxSDc2~`ESJdL77IS0;hUlK|`$1U;-LDSIC zB;wegL$o+Wz6oR}qkqe@vwX-J8!|~aa8n1kJ$)gXkBXyj(A%{&&FL<%bHo_INR}L1 zbVf^_p?a7IKHzkE`bzZS6<;t}GPeziPkueJ*@S0kAdcB&G96SeLH_7C#1H=EE0;Ok zo5Kw-x)Qa_k@JQSauQa;{WEmXH={&X7YiG}UZaAq0V28-oPUz$U{W0v$yUDAp#tZL z|J%GY`4LCz^j*eNEDr|mk6MSsP0xz1v0U9ZID$rC>0lUREDbVztGQavVP+_3m>fXV zW`ylsZ5Wa+)!d#edR66^EtJ(%XAj&dTU^t5q6L1=wpuo05qMk6H#kM%FvDT8kHyZXiFf_{h7@3 zS#Zn}9B{!yzIl7EaWmExQsKXlIk53+d?qG8C_=Nzx^WTtpqpv~sJQdwKj)b!bLCjk1b=h2&QV@9y5YScaPc9CJ*EoIW%pfTZNAqE>Ab1 z4T6?z5)3si@*vA?C8wr%P|hOS$rj_D+JE<2*JQg+lgKQfJ!by}$os?zIB;7v9G^_d zi4CxkU}f#*|4n$f4M)+PFPSzTx`AWIGRO5eIe~U)Hdi>W@>S$m7aLBD0%bUH6S$R_ zd_MIO6eoW=FVVPCub^I{(Tr#T_-kGw?ppv4s(@Kzszd23d$rJa5wetWDw*jsC#&^8|E zg@hPd$|slp#6iar#Pcb{AFogB_$d@H=pOx=b38L}+&9PpM@hL^ajy(?+}#eeXSVn0 zjS2d?7k-hS7-j*owKxnDkBu^JzJGcj=;O$XQ#l|=GewJnRZ|@?E!GOX_VfhXzhn_5 zZuJ05hst3p1Bp5Y7xhNM9&JakQtzQMdbGHiC$6|T(AAy0T3AOPcaKs)r+I74BJqcT zep}qiT)w&LoT({9)n5&Huw?IBWU(g^Mn+_VEG1^mZ_w0DZHhWpI0en_@_#f+$#KnA z@eHx1u%BSl6ufwe-mxUx=C{sDzSlfQP;A_CCbgQ9%uOaVia^HO9|?vsBMc+di-b6} z2q-_*x3Z7jHOm%{!A%5*_6O*<3{|L;1tPae6^=6XL)JHSA0y$`M7z;wTYu4wAb-ZK&gmU_GIgn7`2Y*{o+4n(HscO(;5mFO8hsCAOu)~u zB9ZO6DSu1y`LNvuy3pa<^w5nGmK!P3%MpV`cL{0laBN}bu599BG?NV-uw57|o$OBaVt?d_tDje`E&FYE zsVBRjFcu8lMCukxXUC$xo2pv7)5YtD2?*PxL@~m*@3tz9wm=;bOV?=vmBJTE>K^7I_<$ z;&~ZDU;+$KvJiuE4w5d1TgsUHE5%g;O1S&Y!R{V`(#O&zFn^_Ooa+u>Fy723GQ33S zdp^0uIVhXJ<|+;N)$UW43rd1_!RC;k*;d+eiclR|2o*IX>{;SqO`}efoEb2g!o_4{ zrez5IQby&Jq_?FX85;D&WbK+mS zy#CSw#=G)AV)jS@;Lp!QzVX${O9g5)}JL&{KqK+m! z3d>%#j9j)7AX>|qTWQE)-Z_P(pA|VSS0HtWQy=$?jOEb0{i%^$OP~iEkoFa|Y5N~~ zH1%q{nm+D*hFZ|+v=FQq)%*Tbo2hI!%x!JS)!iI7%75+V`aGU#9544H8aD*^Bm+M) zjh`z24#PoDh3(fJbo5~>;K8cl+(ptW*4}pnWZ(dr3o7n+?}_+G0&ODBJ@H@)-=a-& zJNu&Ea0d6O-u#i1#&MP-o;P$mdL>>{5zHGzmnRkZILaJ>!a{P#Wg3{Y1MsL!Bhl+9 zmO9fsB!B+Mb_P=L#O}RbFMg4`JFii`y-H*0<{WY#VQnHapi_ zp4ESIh12a4;|B`Bd}pb*nJB0y-qr?%){?Ye7H~by?IqcV6+Q*pyh@tUR=mwW?>GbA zH7-`CR?z)w7naxAXq6s8jL?zuOGcYpM{pf&jDL3SZRnl4^otDrmFRRJ%a4%c2PbO4 zaOB?Q=~e%weGKN8Ei7lfS?N{@yJ?XK{Q}dVa&Wf$7X12iS^jB>AGo*$Dt};`+O8hjV?lp6=og-l$4&=;ZzRASdur%>#fFVt1WLXfZpnM zRoS3!b)y;ZO*T%RSBfhO_@vulq&@QYy3FOt_xi;3C+`r;?^x$#h;`aHm8C*$X?cHS zxYM^s+7Q{8$~29=ZbC=r2%1CUwnmt4`=qhpN^`5%n^tS1l31`3(YQ7-Z=SK33uZ|s z=CRXg=}pW`g~jR`z*VTPH0viAi6sjqUe=(_p|<4Lesc&*;gkwtR~Rb;zsuiy-^NmP z$qWtCwhRXLCf?q#jmHIJ&Z*C#a$kStxR~i%ao;z@#u0bg&xX;=udcYQ)Cc<1MQn2h z;+6VacBekC{2enrbGf0xxI~}!<}g+xIvS@eU2ESc;{GJW=y$`5A7j0Umyj07Hfz!7 zoW@IFyr69&Yt>iina;Az@k*Q|{8sUSLp+W?k>!`6d2QOJ_e322-q7jq;CCI_c%1Hp{RYXBpThjE`uNrZv(ewn;NER zvY917#g~@P!sHrb{Fb`tjzu!!Y}U_eL`;V=J&(L!@>H-O0H4f2ngFY`CZerUE_Wei zD81f?@z53O%Cy|2g^MEi`!e5@8Ec-ZrfxoF?h=HVyZvm3raphwRo*Sxp&z1!7H85u zg)5>KIxK<1?TJehaYKT<(TJ#8Q=!*HUg7+bt9s=GB_m4k<^AK9xfR{JTS{x=E&c+3pjjihgmVRz*RhHh0^X3`AmbFDuG-Z)K#oADAm-ap=)(-1oWX5>K+RlUn z)^^r%yk@i!1Ff-!8Nx|6jry6b4bl3(%3?%Yw~hV!_)p z@UgmRTxwZ(SLOir?{OuM=?K1EPZ?x$g>)5B=I%qoHM1o&YRm3F6!*vRu^v79uYvRk zL|~Ty9#2)o!ET)(C@+N<(}1KZ0`rw9e7~cj9 zHe@~V>9?w@&7kc~AR=%bT0z$EDM}J77>cw%57d9TcowK*Gy7qnuCxkMNbcw3aTZ~5 z#S?lEkh`jYoQYo~ASV=3;DF)Jd z;2Clf`|8mM)NXQ*U+IWJ>pR`;jzfRC2ai1J68|iYtzvc|ZJjG0e z@VkGtMoF{!tTLl(APEyHT#5dok#wNs1^D&>)rvK%lAxUD__-UZxJ`!!RN>ij4LeBrF9|6>$G8Rj{XV{)6S84(EzGYVm5!1^kTqPooLh%R za>j}2)Y=DSaZ>LVAheyQlM@5M0P8Z<)9Zh$I%*Ufh<$lsCgIFM!{idRHRJ6m)2I_C17^SrqDs*QuI?V7w#awL*3s#G= zG;C_Yg!|2K+O)uz!>L3gWDJS(SA0^U_;s9myCkqy2Ri0Fea%-lyijb-4~sPlug8D4 zdCnB`FiIy<=W70n?3FzgwK4j1lNU0Fsk=+DDFpzONFnX8^bxa~?!;%nf^FE2?+8z? zde`-c=Na6VBr{_7DgTSKu`Eark+c`~5`(Q>J37=~_HD28o`89tE+~sSoa6`qbTSPd zC9U=2^EHM|n3MZN{#<|8zO$AcVM~9B^L$48rno==m|{FrNd`tK!a#!-^<5MYVMO}PcB|Cqf^*8@sAV&Ahll6Ev{>SH;u&|`~Xh^>qnauW)BRR(Zg-_%6QCYRZ`UIQJ9FD^dhW&XZXN_YlQp6 z1|CLvej8?mjLIr* zf}GwWwL3sTrMdJ_`o@udm;D$|k7=I6Ow%fC*dAP75JJeW)q8!p@RCa#8~ZxyBwq!* zPm4wLr~rRl?&2E!(T)^ZEw^e*Eu|#_`UtvIGWerX$tcwdxGR;J`bd98MZ2S`oF*fE z>K9-Rt%!kvWK5MkDvOCRFyq9L7f-6B4+&mxiQSqY>3*@qxhsA1S7qF&U zhhEBdTTXUfRZ^g7(G49N@R+R&9TB3_`NLM;hTxW>Tyx%)I}d+>Ka%k+I`S(e;yvWW zKuDn2?6}5c)K0nkT)|5vN25Y3PxV(*y%tP0T60a8-zb-pQTJ`eZaz1*_+3ZF)>O}x zq;kL8Gjdl%tmA?fE}lB6Lq4h?=9N^=2`h)m5v8?P)(;et>YC&>pK9SC zS}zlnQ)Ykfqw37ac@*(y!~ZZ`)=Ry=l=m!DZCOZCOHDt(>!+o_F3DJfx=Ji-Gf6VT z&gydwA_R3dxtmQkX~{c|dT7aO)9#?5v^_FnnkcfUo(2(6yQ3u3?$o}A?R&zrdLu}7 z@jlp3IF(IKA3K%iC8yGS(y0s%z`f#C(y@~a)qsE61>U}peAK3d90hFd=IgFz@NoeI zo2L@Uqth3dQKfB=iVvO1zG-JwqrV9~xbUe);X2f>&J};WiSrZ*X`V;(6k(U`)!~ z395f;1i2y~V;)Dk_(~_6>5e#d9;Z6yZ6S$1Y|;D9zC4{C;iw^O2AK*?Io(`szu33~ z)giKd=38DV2UbL{;(9)a>oIXz!SQ${y zhnWXzGQa96pV+iJPMB3j4%g|3EFH)_m-bjSK;>XgEHDptQLc{cM2{O`t$fvkph|J^ zAgdbhe)xI)#MuUHkp`(%k)t|N2*z=m;^5?ZrXAug{_;4wVlyr_gcXOvQurB5yjzl2#&vv?PFrj(Bbrs`E9UvaTCOgUy4D$KjeX zZKEYQ0I7A8j{!PH(SP@z!bIA07BhmC2os8hB0r+}IXXi%slyDnj#OV^K?az}O0f>4 z8=p0D$89Ww@e`EL=>GuBWsHKv;+$Xb?=ahJd`)r+%nr1F7ypPHX?(F?wURu!_j94K zXrl)3W(0;d)pSgQe*px?OsS?bsOMT;2{?2o$M8Kux>c}nbbnK0MN*Tfz$%alD%Q93 zaQ+!*<~h5uVn5hG+9bCszu`W2;Djzr=!`L^e4aC|@?}*2q-n)`7eRfE*D+0VOt{AU zAM6>c7cMf3MU{*RyZ(azG_Y}bB0kC!O8UdXF(Wt8P00)#VU5)IU6HzG;^cO%zJ3m= zcAjGL0TfdGqJNQUYp{tP6ccW*of#?RBg3BQ zke1EKRyRR;H!pHw&?7c-Fk}m*{b@%`;}N;Ib;9HWnrX9e?Am^DE?;z7FtDgLw%*yCu4PKI-r%^ zjyiz^!N|K6Y&t}cmbk32<9=4OuJv=(qlS;}I{O@jiOE|KM-DB~f5ek{^h58=18GX{ z>tMKaOMg;ZF~4{Xtr8g9sBl(>o_j=LC2HrEzGAEqoIWmzFC<|%Fa|r>5*fb`7(W_^ z;%VVa>cI^aN1mbbwCtt_qIz4>;lqAqjYzu zNDBgz(%nc&m$(Abog>{XA+-Vm0s0@`ZJFkNx{5DzXDaQZQSz85f@Bt|c>eyqNVYD(Gw#0WRHj7GTO94 z)vyYLtVR9i(~Uvu1r(ZbCnl<&K7W{(V_W+vdi1BqscWVdY<{f3z9fgq@rx*+bIb}+ zu$sKjdFW~zdlBb4NpA3v#~Lrhfk$_#!7^XGYRFB*-D2v=b|rk3$Np40d4%P+i8s8# zETtg*56vqtjm$ZPt?il7VIktcSq(w$W81`6!!7mQwd({|=WCj}oqU^aLTwM?(n{GuW=_fk1jjXQOWc)(u~ zX+A-k_|+Z#Vfu0oLcdVAWj%%bw$BA`$g7NtTEE?wd7kB#!789gP`CUQyyrOcEkUC09vQT&)BK1GI;CGuf&SsaWX3AcfMrrCbECF_op6C&XepXAJew#EGG zUljLDeg#7Jj<#QWBdwl&FL}4a4o>s=o7YnA6HjiMHV01KKg2?$&BnkvTZ*j-#NV$l zIohcZA7e1wGe#sTb1eDcK5+Ln>+Vvv_2DS#mY@r*5cLle9K(369p6<>W&c8%nDm%t zvB~<~=3|c(K3PF-QGOdz>Uy&7$>~>2XZy{M<=d0*oHd{Ct#FCT2?z_fF_<(&1tI!h zkZ@&0_`D9h(C&Y*@BjBr(c3C~ci=il2s z@5$&p>~cnWN@2T8r@FlOg5`lqGZrIa&viOL;rFW_k7L)J?&nu+HWN&;EsLBc*;rp` z7l9wNg+x9diXweE)|xaT^waa|UW*KVz*UHs9C=r{q!y8F!9lrVQ%wHywU_SftR^bU zc}1#**ZV6Ufw2AeXm85kSp(z3^#{+^P@7bqi&N>n*&J$n#^XQ!Esbw(D;Zc{VDkO; zN*-^tjoAB;ssg2{^3Sx;Qz5@6QZ}QgY73&Di!uoA;U|)Bu!uO5*|*0fpwx~9`MmPl znS7*DpBd_-t+G~g%Gr*FDncnI;?oJ4L=LS~J{}Y;@CF}{X@SqTqHI_FuRD37ID{L?9VN9{=F)z>k~ll$^_w2Zf2 z&zI6#UHuWGj2Gv2 zj|6S3C|)0Kh)!_7$=Jcwx`XvSlPQ(2SD`~l+hz(#CDKx;QqBWU9~x40zl;hK`mhP1 z`QSvlZ;vXR$~EdwD|xSXC&xqX*E6vfA+_p!Lz}<0{RS_JBGFzc77BR@;5_-1*R+bc zZf|IR;a@#nCC#So6Z~8e``;rTAYz*R5ux+Q$AzY; z>yBz3*9(jQSmo7u*VUcT0XyT{REP_tT5H2fL+0h;K~TFvUqS$iAWQ-T$T4{zg2W3F zlx2dfWba(CsADsVFag~lOcOvM925VBe-4=!%%DESAxDEXxo@e=G!l?8AwXyZ*o9#@ zCP=+U9gBIx%w`DzRO2x50MaPTOdt+i?MC!OVbVhYV*nI4e_)Y7fIR?gbwf1?G=9J& zLMZfLu|fc(Bv2dPi&X~!@<97WD>Qd$9v8JO4Ptp6o`v&{kvubC);$ZW7J$n_}<0hn?y zZzw3o0F+!zJdlIZ4Fb&Og2gu^lt5iR*yfsr65uNUrT>vn0^Q*5zkG0LMc^JtKZzjB z!_0#K+eKif*94S+-$$^OYyqY}fL97$t+@hBkbE+x6fAzq1RVfYO2OiP*(ZU@GE73m zTL8-)0w{k1yZlQ$3B*@m-qOs4RD$Y%Xy)Pp(kk#0-*8U?pR2%w{&_<;7g`PWij#^p zkIc)@3v%-D0ADLHwSoK^OuTDSzKB~Q)jlL%e#Gze>z_ySGdOHr4Q4t7q}y%w2T1{K zKVvq69HIvJ2$4EWu*NSVEb40(NGQU$9#b+Gh-QJY0KW~fOd*1RgDNr|5VenmcB256 zjIrpicgK);1%dt@EV^5gNGcO7F^OAJ$iF0!yaKmGa{rP<@(SLPLjI+i;}zujKU8zP zAd?_SH7ASMG{G8zg6Glv8jAtW^Iwv=c#W=7G6CrHF^hR(J$NtIlPH2%Dj8`Sb+>dY ztFg2a)A#%;FGuFp((li;K3uz+S=h7B4)$y@Lh}_hF*LV4YlIvpnrRSBG^^rLc30y4`b)bOaeX3RTBgbW{KErOCuFR@g<$n4fx zyKo$s6Xdf

F9ISCZQ!VT>BJHsvI+v9+QXm6Gtz!dqXc-hFkqsJ5Z*>yEVFv0Tc7 zC%Civ!I|leoo@lQ00OmVSyO0O*bmOqq^_bGiW0vPmZYy$_w%}cES`iE178v7XbPQlp+ftqusonWH~PVynkBflcC3 zpie%z=;5XB<`gLX=v=|X<3vKr%`9NZq5koy%#VBUE=li@(ufE=O5q+)tf7YCDZ^h6 zP+Q*=#(o@7?f!(nZW*AUVLF%{Ja1EIpZay%W*{Lz;gRZZ{9*e}x!xWtd}cs~aEt-C zn%DJGN3zTv3H;WI(iz6Xq>INrIX%afDp9(7Bm zVc3qzg>;9_{7MGwl~f~`7nhAi?d(q#F>{6Sa1ON|l%4nGdHtj(U%cD6tJ+1TqY@W1 z_nBdWx5=4BkiA)l{Y*qRz+a{OvT7vPd_ySo5S~z*>ie!D*oake%2dy7v>IVVJClforOzB*PX*W(!GHV`7#xfnfd*!@m;ej28 zf;sSwPtA5%=05s^gML`YYd3F^F6}4>fOjQny)rp!emSlEJ#%=GLQ&u=IzbYfWd)P| z`|#GYPiiIc%zU*XsH2DLi5Ld9v&nA~!ibV2G1kL>kVp0h3<1+72_iN_ROYs2;Px2WLJAH9*c~o zq1tX5-&GfT5qW+v2-ocwki;~zGj(!yG&2H1+_5Nb*xfkTY`6>!~=^65$X;l z1tC=EnueMe=yb>Wbxrww&CpH@%wj_?0B$%ILFHI{C@5}p1hDe~ zIm)kLzzU8f2W0w!EWseFGZ>K<{6Jpe8%F5dbWNblA8Y~kN&^AB13(c3SP-GV27=W= zfUSoBWOZ8;@gQX|{xzU-Lx7DCtQ&^yXuvEKYzG2r5Cqr{1^d1xcm_hlKyu-0 zhGzg0j&)6B`!AqM0*Vn}Tae>95D<<8r5k9~jsklDp;ZtBeNeon&>jO=V!-i0K*fas z=&@i;5KsevB~ZHNb*6&=`f*^*zg(+1fMh&4)IVISg}_t-*a8GoOW+j%9@Y)(>d^kV)}2^=3pP!9<3IT<|V3oi-aECuT(oj?G)sbDFHmti1ra2h!L zEnbqRgXJJ^^8!TRFIsAXY{1c>{}ML?7@6RRoatC!0mwBW@eNb+HDhojK7t|>D;vCK zIpAU5Ff#MS_D6#(UV!;pAYQnnhZ?N zX`CihxHr9p}R)YyEUh*Z-!(mF)GzRS$J&T+MZkCkX}*_!;(z zP;@%7h3qWau1$MrOSQ9V>lEa;cja)=u!4vBc!p}aG0&7Q7j8PcYI5NQN_}e{>P>pj z&5<79n@HcBVUK_<*N9YCPWS|o&yWoGo+;~S?{n|aj(l)_&tu#Uj$j7d>-ZMYJ7B|(YBJ_2#Kdg!2jpnwx9&q<$&4EpLWp5 zqFkZj$b9^Qw;44OpFnIhALy&sIW>UR4Vu&~Rk1)179j%r?X_`x_JYO_romtZ**>rW z=%v&k^>RO00rXOK2%t3pYHz*tV&Fes3Vav>TY_%N1p$1A!Le`h-iu+-gh4kA1}$g= zG`VXxO&A5u?8Zyw$NoPrl>h|C!KOF9*)jo+dBeaCEKPzfZhX^w3Jd=>?X8*ydj`D} z2LeRRfNel8^#^Qc!D9fu6wK$>K=D6b+B^5JmlgnK3t%DWrB=Y<0(ew6Uixkc9OOS< zI<^GXxh7QyvX{Xlxy^IMR={qqz0`jd^ujWB=x4xd4IJ|}tG&Ap9@DLtPJ-fXRtqxo z-1uqv8kRP2ya5geI%X3xKc^rflNU+?i8XQ~0N9hTXkzV@F%f*5Ag?=FP6iLVVna z7&K@i1a63U?>h369px2Y>+qYa3yJ8--)Wd>5+Ry72_gLxYrPTL(JJ|#Ecx$q=&WO` zWkcy|Wx5r+jk@kP&xoa^w!T4a6f|I29h)U(dQ2?fA4h*aY!5;jTE^)2Id?PE8m)_}x??Dw;RZlNc(F$8G^|@j zV-n!wruV9+1=o$CH*||3N9Aj>BJK$p_dJ84!FNqi-s2E(Oeuy zuYXDyma0arUp(3JAdGu(riwIwf0+BLn!>NB@ziD`%ObI%gvw&~_1$VG%^YKjSNAh< z2O^y+B0eWa4L7(l(kzfwe&`JDXS>MQU}92+yP>Ou6Zh}jC+Di-r<^N06%Va!fB4%u zRYMW$aF|X_aiJdi!Re5qU-d4Wg6#9B(T%8r;Hi{Zdvrmx{L#$ffvN2>G6vNLxhzp? zjB9FRI8yl%-B$KHooqYB6B9Z5lb=X-q|b-Nv`gxS}t}rt0W@DuMjGBl3yP*+_=rI8ql05orkM=O8b=W(G)IpuIi+cFj6^~V0X7O_-~ZZn;Cp8AWj7;Z)fm5RG<=!WTSv2HFy#?p)8UH?BTx&hdAI# z3s$=cWvCCKe?u7q9jF9D*(lHiN{9$rD2V+h^kCn&Q}_V}u>58UU&aWQgRu-e6=o*r zzwGWgz&&PgtpBmQ=d*w@2te|ug%Uan$>2E9p5t}+-1LGfdW9B6=e0jTBmYiwx1eULLAcr6U}cb#=W z_+y(J@MD#JTmt!zp|IG7SLgtE{hu%iKug{_XM7#?y@PQ=q)YlPj z2Kc!PPUgWJLrIc{^Z(=*{6hc#++Bd{|HeB3u3IAczwu6h>z{a+A`hKsf`BLLezW`( z2IP1^32Szsl8_s&dql81l;U>nq1FSGfa|Pz0#FHMUn#pz_9^POhaJdkTmFr5k&~jD#A4Yfdpx zJm75<^bJrH3pT`#zP4hSc<>j!#>UVD=sy|kI{++gh=qa~Eba&Jy4+gO{0FeN^aSW4 z_>2Fu-lGYeC4$Op`vDhq#HgV7i1cbqGzgLaKj8BQ1X$c8s4oOKNCC&aPFs-$`1xX+ zJH5d2*xzsQ5$LJlwFg_?*uw5E7V7OXjzJpqvBWLZ{e?aO{@Z1pf1yu+|27%@3w;6t zx5o27QxM>~g+7LKXt4|cJ~>8*!o>#~_Sp&)1qsgY8LB)|B*e6GhclEAe!j?%MV9gL zaU7W*b&)@JWa-Y}mCQ4Dyoc2(nqm}#8aR$4+8Q~|LjL{BqgfYtV@muGukn(v3s9X= z5@nnWdP}ryY^8T+{3fjfDeC*N7kzd}Kir9tQ20L*rhSvhbpdY-3w=UMi3=*=2zoTn2x%>R_k@#;J$Ik$pKrm@S?)PEdc_)_ zS%TSS7z*OVO=n1$qxsjHtv`F}P!?w2rLKd#|7wFeGBj+u#KvEg7E{VLv_Z8lrDa@m z0KdyTa7vg?^K2c)pY}1)5fX*%^PAGWqK4?oR#kRc*rvITKlaB=?j~}x)Kh#jiQ*=Dv?_}f&oMZ2H z6j`oyKIQt&N^IBdm}yDmR5&AGE|gq8bplVW)Jx#bM#hN$}^d^>0-T_$OC!J@BJzP|t8CXb5W5*3n+6{1WiMLVT_+iMYna z=dm#iCSc`>oR+w5=bnH+JXr*>z;lxk0d=4r$k>Ah0Z60Zd_s61>I)bPTz7B{m%IB= zWC-x{7>veIA7}uEL#QnvdjhrqCsQm{0r;JQXzv~~36z|HL)^mEqw~LTWp@E3 zyk58=h5+T4V8g!)H}P*_f3fR11ok(Nia-Ljx7fvl4Eqmu^`d|Ux7hU=74{Ey@uGps zTkL8E#oL7&ZFEq&$)ox}>HotsatyE#T)6QE&@n;j2HBXfz+P^Vtr-*nP7K%>a0-Rp zEZl(06+~9(cgXIB)xG=nb)2PCGV2j%{3LmU?bHgJXAJqLP zjnX235h5Dz!9cM4LkPPCyGO)eZ7^Lz0GuS?NVWtp5bVB^!2W`rE-6^9L=0Q_57;3K za)Wm+NMNae92pE;iXw$ILxiqpgTxSIK^`uIJQ?gW1i}Y2XJO3)omAjKfa^0&*FbkY zDZBzMXkb_XODrZ6a76{XM4Uc=Jw*dVAHeDnG0ZS=r0eOSCJ?~}le$*cfgRnzVmT{p z_Il0c0=x_I&gCu*0OF(vtIKTGzIwtZb6fc12#zu=6m414>|<93Ty@y&d0;~-+LfidOA+{9W`%fn~W)2 zg$T`0XBwalq7Mfm26%+cf732TvtPGv5Z3z7otU3A2Q5Vj*sS_OH>-q z%Y0N*fmSA`tC_Ey{*_`!$-sn}3?)Qh4rf^wVoqo$qqa?OpGLWMu>BqN2K}{q9z;JK%Lnvi8M&yw1-w#DwA&B?wc994%vbT$Kll#V?z8ibDIvj#^i@ZVluZ2~$|Nu)Lmdn%rq^Prf%KUc zx%8OBEON)rU&%>N6s_alE2AB2_aPdapNQq6L`D-ieB-jL!;s9x z4!7t>b(dCUsg^dd3f@yfjmNVf^Hi2o3ivaKXHVi|H^RHsH}%eIZib)4Ih-HMx&S-r z(@0iH68#64ITcm=cP(Tj&wb!xq8_VcO81YyJj84*Rjzx@OI$q~o#hq5_>n~6Vu64y zR@9zkA?>t{RGv880o`91b{MrE-9pwVW^9dh-;j z8QnHhR>2kLqU0`#m>(&N_#FwY`F%_OeNn z5lFYa%2La*TqEla{ahLkZx+;^mc+D=dL#BASe)s@#{$^CzVS#Vw)XJ^zD8O?ZHsQD zI}ujebBu=!qA^P}Kkzgd$KBWcM7bXrwh6DpSo%z`D=7#(N0pm2lRjZ@q%cmk3J1nX zGDV!;kry|_z{}K*;(XvEZ&u5w{v;=o$9*)@;x$gS(5q+(tnw)2Gp?$SU}N-5lv+;zan%pqIP)R5i1ty{r$XJ@O0wL(j+VKhj#0moke}j` zX{P4f7wxYf(h$oe){A*#S)5Oa+1!taqQEu3e@xwfpSgMQIbU<1(?eu387##g;yE!i z@zgQ9Y|^*{pZ5Zm@|(ZI-6duR!m>#*F)!9^@tLM7)zVtY4?htE?fy7p)a&n{EGFV6 zC^=dcB3x#=FF{a>{7NlGT`x%GLyn?oB6IJzca0jBcqoq@%47~vW1>Qw6{)L({h>YY zJaXci$5hT@X0Cn6)BX_W1Kdux!f=~RHWi)KO8wZfjg^H={2SG4=;9} z>>M5)?DhQFBV^kAg*Y|xR@#|E@?AoBVwF)0|Aarz_$gqK-MMCS`V$emMkjK$@9HSr zm^9s#GIO?Ff7_*G8+Cwb9soP5^KQqSBq{hR^ z*nIGtHa%{Nu|-PiM?s zQjQsQ`uIz1OhG*S)nS=)>|7#@$0^UB7RxhVkOv9jsWSG+=!)*p89NJ!Jf%qQ=G<8k zU^n?WrZ!ARB=)fMjyt{smoTLU|I5G?6_?RIv1oUkHvC{XPQ2W(_xl1xrzNVgyQL5@ zoVUE6N0l1{$j+juIm%?iqhem|L^J2EP&)!M8W&%pJZW958omfA;#Og+jzon#tZb&c z45B~Ga9-%Z<7CJnf!=rVFs8*NYIgPuIJY&59-hoDcA!N|7+KnT zRkLsBEZemK{~csSBV`dDb3tgq7iw4}Hq1i2wO;+)fGWZqh{Y+}7Jw>zM7KlW*92P>AJ z-l~D~7UD-#iH|?~J0yNkGpI!%+>zCn>oX4d8uHCEAD<#$#?f!a z`rOY%`M%1N9TEL6WZCCmeF?uu(iK~xgsOk&Vb$Rx)BGHBMI)=xpW~Ug!mGm?s@cvW|1^$N7PehV) z^dEEyO`kZ)YXsL8ccNJt#hmZ{HsTeIiq^T&~N43QSYU>#tu4!%SaM!9-9IQQMoT4C@ngX>U@IIQe&0JI zj$fpN8UD7Uz$K=L6{C`MK%`VB=2g{stY}aLFC0??kCs(EOp&`T>XB^o$fHBM26QRW z_4u47ZI!aa&ir`m3k#JJ2T>v^lT2IGWF{vU7E?{tS&_4oxw9R_PFSNq<}19>#Y%zf zVz;GjdC%<>j>?!&?0v?#pA2V{{TQz*Jjz4dC6LWu~&k_ z8UiPjvNu4W5gix|inikl$3xRIxxcmD_-CzIzkllOAR*qFj)E8l zan`J^?fue|sQDzrGw0;2gleRgBo^G()*V)B?2*Ii)oCe7d2Di*o?sTRvg9gov~Bd*7ZB6Qzo; z?|p5DQ7}=-J|5&5X)=!KKA}-OBPYSE(VF$egrcMBl!T#6Rp`lbb46UGo^O)k3a3J7 zir-gm42~}?5QvAM@)1TCJx<5vH2wNG*C@SJDxJ+~zCN#ZSFWbkkUE?_#7UAft#>yX zuIxfA!rk5|*wB7%o^pYHa>nCpniFnMYEFDN<`e$|8MG;ld}VB|b#rKEpO^GKUpxu! zgGTx@j>Yzi-nc{(py${o7my~`TxcbPkG6zw zLRS~0Ln@?%+ePr~oJ;nRd8N=+6Y&JH;N5er27YAmC-AAUNV29^7az(bFzcR`94x$K z>W;B9k2H(4+h%g9j?mU&sB}6AkQp{BR4;f(^DIJL`NQG-$ipPH+FS)a1gqPdh8J6O`F}-wRXn)8I+)|bZ3%xz>V08 z7fdKAynmM6B`E+*6UFc2OXNu}Dg+)%m?XZT$&ryKIFOZ%qr&{OaM>rmx%OS;{bpUU zQiW=P1>q4vZQ@JD@rx>YZgWBCJGgFGulP=$J{ww*{{Tb&vZNW6DwncgqGPk}>UqXwo9^=-f#ZJJq+?}nHJrj7 zcY@{|<&NIMXURN~;H#v=XJK>P#!~xNBJ-Wz=ey&re)U8H6QvK2_kvV#NZ`71L{lW( z{AA@zV>W`>-%#4CV|vgGc~TovV|oHup6f7*W||H{x+bc2dp}YTofeB=8D%z|wys9E zh_88mQk*Eqg?qM&7rrAP(k?O-WaZ_3G77`$od4XX)Sn!T)v1FT@Ud-(K~K+Lbj0+5 z@w*O5GYZVvSQ;q`vM(9$6$Ic}4|?ioQX5Jv6l*A_8=|mL-gQP3TRE;uC2=3o+LFGB z=6`08(=$(O<#aApL3o*f$2ne=VB=qD#WqfmuEc#Wc_e1*HKRjKD~-b#%F}$cY*z3) z@SD*Vm%sonSM;`@vL_R@&+u8g$!5MF!J~{UJ&Nmw}&%@PaA%yU+lBL*hjd20ciU; znjh_qxI~JKoh^6HOdnQs^)1A_yxb$~S*=7%$_;LxsS!zj^LZ7fzioJggGPJ-AxPr>8a!S{2ym$X{)S{-v}Pj&HDVnHfkHclOHEpw|f3k z0j;x022)9hDdXkKB6)0?Vs*Z8isG?5<{ft@o7{bS3^3bSq9t}i!u8d9(d?a*nLlMD zGv>PKO;?dG=wp{$ZHZ*zHxipuHgY(tJ{S9Lh33iiEPT2;L@y_|7=G|{*!HygDa%ye zlRa}iL?N`*;c{6;V5uAf@%4_DAxzw-jrwhccf~l*PwRbx0X$gSN0~F#EG2o=J+8KA zxIOeW;nHXw? zJ$Zo6fT&9Lfls2ml|7>}V+e$Mz`u&?!+Ynhkj`*J4z9x0$93SUKNI{plodbC(~Qcf z7if?bTIvspD|q*8u8se+UvZp0#91j}=t|i*MEkv;tahN9E6}KH?&s}vjHOqSo9=zJ zKTk0pyZFmvReR|hJzK>K&o;8hXop9ls@lu8J-S{xlP_nrq3r{K&u7BQbVA8chExyg z`WBK(T55?#TRv{Ut7gW%_WsOm3zbl1!N$c$xz9qYyDyE|CR1oE8ncY@PWL;>EhN6;(L1@kOr>&95b~=9#zdnLr)#k-!k0SZR;K1Vh4Lis4dK90E zY+IVU-?%F|?#89su=t8Tk$XEm7n$HkD*PnZuAO&uz*{|SDrzvFkkjMjqkI(E#H2E5 zGRh0>jjfSRACj1#_a06}sGHaI6IHx`%1&;-fZ2}1rwc#rudePc@7E`GuM`l~bD4b1hhd>e4gNPo#eC!iQA)Q1KVmkcE&(feH+wu8q z^e@9pn$!=iUihy8hu z+~zz*J%DY9M>LV1>U%TMIzgPxy~=2z{Lb%oFJhzRgcJfkVVan=aw4%-`Z%C66C}Gr zb%>RVE4`z?zBeL6_OA$!c3O793Q?Fby8DBhk>5uD3;HrC0a**ASrE0uA8T-ZgLp>QdgFzesq06FeFY6;qoc$8K zLcyr#-dJldWo%n(j&)y=v5g)&ewp$r0%#2EOzBQcI2T!H1U6*pjNQl5Lfy|)_%S!sXGw^&(OZ6dhh&M9 zKfT7pQRYMrT_}vtfZ@B*$BQuP!1;sU#+h;Siwt{85tLg`Kg*>Vw-W7de6IY!#col7 z4BxFyskRi1O;u*oX%Ve%WKv>dsAN#vn8aUx~gqo z?TkZ8BoVb~BO+zkMbslTH8}vO5#xYN8WO#;{a$-Uzcl6D>X8Js(OS4yUJX@p&qkLB zBxTgPqtlgu0s-kmzVR`(ejTZx`+YAi8Q!Wj`De0ae4wMAKc!(KI3ay;Q4eAfYWib- zsyMf;t~7GAoTpge;#J`gj0Q4OT0wKyS_J1yUiB#xjva9wyRBb^g2tHkr_%E2$upOY z<>pbv)2?SZOFhl}=@Xcu>#Z4(>h25eMjaNke0{~}eHirsh>T_-o!tPtcbv!}>BV*qV zvVYX5(`nnHD<)C=f~Pg8K)>kxXoMp?`;eQmc6%O{X?i0wJf^@@iKyE~tNW-gFu2|v znr)DN`34BXEp~DZ;S+k)mXwletkC=7Q(pJ&CnfJ5>|R&m{eIj}DMi0)DpK;qRN3MJ zf|ehLV-&o>Ev`PHR=3_4`L|iV?7Tj3%isE zUNX(Y_9P!YZb?=$n|@rQST?#$P5#L-R--{%B7qZPbJ>~q^5(Mju<@G7D7IuQ@% zr^;^|3Logw<)4sp%3Wd?G87-V+xD@pX*F&n(W&F)tslnK(byDR5cA57baQIe$a}`h z*OA4K@Tr{Rc^@@-H$O-l-n!SIjXKgQ(%l$Ug(tWY!&%Z z=21Ir?J>qy|NivxW)~O{kt+AC6c*`EM)p0V+U`kXou%4n4$K;HDYrQuI2&{r>={jV zIDRvEO3_(icvij=VngMs7roW`mSs#NVp_f2oBZq}Htaj^hrk*~_~s{JXXQZ&-dXKN{@DBS ze7sq&rMa|cz|}UP25wzhrI(z$MO0CxCz6%;_QXEICzYAAT$L30Q4KLe#-5PiixqF! z?Z@dgIfDg$or{f|ZeJu;!q?%oO) z&CIoLXH<15z&!D#K=)WsgEDqqwVdBJ(bsJJt)JyW!eVghL^`sDnN5wf%f;5A!n!Rs`## zoJ4fL3sJGhdGxR^eK7|Vi;-Ub;ya^X;XBLwZVf+Dh#uWcPWk8K|HXV3tn}XxhogQa zu~6A~sK-CwcHp<1M~^=15}#M}ep17CQ0|^D0_b!Pxm#ay$bDX4248BI2x>@6Jrlm| zhzX&UrqMh@DQ&#Cm&lRase6giT%`DWA6ek4;1S%reUZWYqa2H;?VRwNC4I$0GG77# zp;tRt<{SvJlfC96=eHLYrkuokMN(E%YXc)BK=QN<7iv;w`w?3DPgItL&?sGDT|Ay-0XdGd48w%JpmC^ z`;&>8eJ01QQS4E=Ny<^WS;#2rtl=P`O}0q7ajmzFZ1lJzMa^!wqEl*Fp{{{$_DQLs zPS!~yxY7D&Lq>RNYI?A@oXC+|_a^hCkpwOH*O@WN*949+b<4)Y6wdhC+JZtj zriAaWhBrdLy1ui+4Yu`h+g7b2H;f0?-fj5WtFd6#z>?_x*(e*xyl#wS%jVX}Y$a;M z$Ra`Y`xRol zYjLfG!yo8}W>&6BXjKho;h6QShl;34>9Czj!wGGxSIJ&h5gwxQF%sLDcLlTS%*hd& zSo8-LeQ+f5E_mJuN^{AS9AmDQ3GCo?5(}lNN%?Zz)48;Dxa@8vwOE`H`U#n=?t|kT z?b_A3$goKl*+C6T@2U1+&T>fb zd%9@sE)K!^&$;j$^LA_Js8vhII7iG&TnLYPtCQWhCl?b-yZ1Tc73ax)EkBs zO?Ds7&n5WZo4R`OdZ81dB+@$W;IqG zZ;GS($C{d_@$4i!e9`0>)${6YZwfG}3|_Grb)BS1ivTi%$E~Lj(*=x`wv#DRYY3sC z3(GI=AJ3=`Y0MW7(_dYlEl%- zb>SQl=z@4;KDfr%)jn#pwf6Xng~6}MMeB3V`Xi^Y_ROuLaR6we*BjS8y=0>_cI(~i zw@Juj2^u9w{2m%rAYZqmdNz1|nrJq-ewr9=jJVjjlx)>^>zVcZ{6ruASeNRjlTJ4t z8=gl$xAIyhv$7@~GgkYu5=MNK5mCQ4?eQIj3_4zm6rb-4Ql?x^GFn7^Ar3^eiGg7; zLP4hblH5)>T0yMsJlHz;Bn&OhmPF)XBe4?imZ();l;o#G^>7)x6c6Fc6iVgPY`&+f zCKlTMR8NG?(b8OD0e-<_{A93y(4yaSDIc0v0o6aF28~Qv?mpmX`QE36{ovWLV3_aS z>=->ql6Y-Mc8suK*z*eTGesf&V#A0)a-Eefn>oVnP!Q!Fc0|O1AWF1n@ZlHF;5sjI z==9?9S5N&NF=l~NJ!V)z|9A7ag1aRE$U7z138|Nq!wbn~lLPH4c62PVa%_8{94SAX19a-?xfjfA5< z6Jr)F!%yL{x)3T@uh`;%y9VaM0$YkZv7fX1TuO>Bw5qUh= zA-sj{^oJzZ)0e+=5AQl$dYmx+{CTc6!{kw`=TwIr+yjqsOobuoEYh;97=3G03J-V~ zd}qOaMKH`YphtQAe)^XqGrEZv)#bInxfz#DyE%>uq^XsY&>qztJgoZGtcGNps~v_X zt|GJdI9#a{ZXd@==knk!bA2ASwAT6uVfrQSokY}Lq#dGE;k?PHrb$l`l(*jd-NYBG zA3Ug}ab{v!mP6sc#-4blKLKg3DE<&#a8uP2#8%sVFaKRG^3*DS_cxO?nE_!h)!L?S ztg;S1knaz}21sIHOmHqPA))_XyIfUU8`YO2jfkzfE`H{aaJ{&rENPMVCGUI;0|GLz zY;2STzM`TfIBMl@1R^M^r!#WdHahAjOWiKUbgdsT%Z<|pNx^9DR`wfaoktlCq1=u6 z6K{E5nKGG*F;fwAIYPA}wDN^vMQCR~ZG{da5gtY@tI|JquA-{5F{qvy<_^KT@73HT^BRXA*6Tdb62~>&_cPPI5+zO7eMtK|?k-{36G>L; zPJd?4S!3wD=JVZH6;-N_Agl&NhkNwal`O{0(MCS@}jCR&xLTie`Y-yVb}R=XrC z)KS7qT=i1b>s>>~7e2d5bBvZ{jM_|nHX>|LFP@gM=jXOT=v!pzIP)W$42^$Nz?n85 zapU1fZ}`$|SgDm(Qihn^auLq1?(KGcf-oku-Kim61>Y~)N@Z$KC zeO}dOQ*{u6G064{8ETDJt^8i{1!J)G0+%tINm=m=J7y}iCaq$Q%u`ZoKhq$Ttk{AC z+=ny|)Nv<+SeBkUs#Sm-8yI3l8FcgsT6$fP5M?$60Yh_T}ndif6hxB*PrtO}~BVn#eLQ=b;#HGZT zKS~=g_2-;7(moE}+vEB^u{F1=v9jY2Y_`>X$ zbv31TF?qPON08VQ)z;!lZ}z=mqZu5r4BeRu!_i$)V^NZhjHE8ZaxZaxLG_IY54`nq zM(g+a?^QoLHO>TP81T4h&%jrAJb{6>g~J8(1?te;;ei3M`IlxR#QTL8bH4*^F5&IO z#{$h=eO$F0LlJ|IUI;*5B2N5IXN>H5DLojnNb5Wj!qQR9N0-QoW3H%S;@(#fEFwc&H|b!sDG-c zN82bbZkX-WYwPw80&fn`4lUBBxa>omPjsd{jDBlX2u7IawHYL;oMv*6yh9svYURuMuE$svanv~LEyd#d+{+Z<2d}O~!u1kmdm%AFs)fG;^I%Y$k zPW8RP0r(pbEZZq#>919c^A}#bO4lgZw5|IRl(1~Q3`qpQe^||WU&PF5=*1!>!-RriRjrVTT; z)v;~c>2z##Y}-zr*tTuk=-9Sx+sWkpXEx?L?!9~ME!3*3R#lz)F<_Z!Wta!CB43~B z?gh#kq_O-lm(N4}ob(yx_Wi_r0zPL?F32Ulk6Sv$71x!VYxg_EY2iL6802p*hCE{S zuWM61>$Bl6*H?CKGz5UIZaKqqfYIFo?d9?eA0EF&nW(K?>W6(sm3Uf)1)zi0#cIPG zB}0bv_Xqw{H|-ZWe)FeOqrg%f$F`WIVY|)AQxiJ2GjxcrA?IWN{EhR0dR(0Fg-Z2sMa4 z`gG_HBTacxG>9-4wUhvhN>F1ICUF0DletjySfR+UE=(j(eS@wX!X?l{=u8ztRRBaC zH>$&QoC2d96FKKR10m;loONswChEHp-_vLy%eX6HW*P+5%Y-G8mjyW|tWxOC1S8c! zh!wh5@rZb@`WVcOfxK%FKuNK*F}Pjdzg@V9O?H&A4kXi@l5&9^mvqudxEZEtab$_H zj@!agEi3F%UbBCdSOC;H-~(;}-omK)qi5qAOUXOtCVcZ5YEep&qcE)S!YIq{CqN_* zK9_51!F^TT^toiY?_=ro_gVyXep7h4e{D3N;rZM&`^Xu+>l^3_^aG>p|MtaP1Uxx_ zh4X)1)%bCX0710iXZ}H2CCM9h3lL&bY;*DX}zmYn>buFPYB-?otyQSzCKK z^XKoc8(~w;OWVqFfM>1hnv{hnbqREQU3@+!v3sUb{C@FE1!X)~kpLz~7t3G;{uPVA zE*u`}SNz5j^~h|R+WXR*MoQb;Wn@gu^JjY7u`Q@GxW*-M2 zO^akls|s}jQS+BS&Oj3&gUD!*f#(GPw;zx+05!dPAh=cOD07kUn9~p66 zZ-^#fq76%2*^Z#U`KPHXW^@+Z0>z;7W?K&RbP$1bmrIw+egD{40mkOv_iVhXS?-~~ zIqpFzm2d#wFC9$a{e?;+*EQ6`*VnXChB=a#mc1th-E4trQz|1&D=CqyOq-fV0F|__ z{<*`s1krDr6j3&KhqjvIv`Q-T`4Dcr7gnX09?vyJ%FIy^Xbf07%qdsBB7q67eQesl06-91j zzan78u9IDOq-M7i_D*(wJ~JvZbwiu%$0Or<(V38Jrv(q69Fjbx5Cy}uOD%e4pAuanhLF_N^A}A{Dv^Y{ z@lk-x$>O&5^VgG>jxCjUo2#=2gkS&AiE04AZsWb!VtI9Up;SmD8`byV2;S86FQ-=E z)h*ReGX+7KMW_(Z)gg%v%=-vbYRE)1(MGA3q*e@0oS|7}HsVn#I>MTB|6G*sEgcR~ zfg1u7zWa+b)(#`^(T3do1zr3cb}yLLH9PcA7h^jFWL}h=h$f$wRHYARHF;{U-{y@_ z(B*DlK#4g{|0hwwm>8M3QV`RE^mMlg>8r2?i@LOq3KT)(!m34MXOIp6J z5z`=iZwX8WjA;iY1$)@VtRgwt#g0;{(pUav4y2wkA~6^n-j1>Q7aR&IFJ^pt?jI&n z@~f!zpzj5uAytN+&g-uJcYMDChu=uQ*Lq1p5YR8TjVS3C%>k%?vL#?Xx^XS#LoMT4 zqkr0x$H}6w9G)Wy*F}N>=%sG<@4%LaiQa2LsG$6%E7~wA8`Lih7eI%|Q$1&s$7b+BB-NSFMk=*ikim(Hc6FeyB_zoYRS1((KXCjOy!Eyw?-cqsU-F z{cD@pA>;LiA^>)h6*n5Cu{N${@TmzjZfdt9Aj{)aK#t|t>`pCDh1=918qY*#1qhL_ zEHT}f2)O;KKl2QA;U5?*;#k})(VQ?L1D%JW;j!KBRz}HzM2`61Fg*&+&w@NJc?CuN z%k*__25i-2_Q1ULa8hl`sZKl8gL|Z>HsIMqKBqAYgn%fmJ7JoaB{$DpT{_u#j=I0Q z#C&tjU9u98iB1vHzr0oQR?d*@P4QhPDZ)x_D`Ae(k1H7>6+VGr$!bgPB9FB69f7R^ zv@hE|c?b@sVNe;tYk~=UiiU{sYIfnt)C<*az6t0zCz~*n@OxlZs?Z(Eh7$f^5 zO(`o%-~cYd$QS-yTc66>E@hu+@?lRIYESULg6xhy!o{}bBKOXo$G~qH*~{(1zd`~R zD0ZBx1|@e3MtjHiWef#(ji&{4&L5mvXRUATC;9VsTWb-e^*S9KTA60BpG&1?)TE?P zNAb-rlyrdsxm^veJ5LLmse!e1<1F%59Q{m^uK=7%`8|`(qdQ5GiyxE-+q_@1peBR- ze!y;AT>8zbhh+j2Lp}1i`5{QLk7!-IY>f7><(0vt_MU{5qvC|`vGS(~5}c|CUl%6X zd4&*CaBdDpkEEEg_jdxf$1SY-v%qP=e1&knKGNR@mtgVJiN%B2qPRqPR|^Z$)bwH< zYWQtzX>I@Qst?%II{D!LyZU}!)ZkJwU*NF;%q*<`|FfYj|9=v>>z&5w0+U1*!VnUb zk_t=xls)~aw!n}kO}{Q`5BKglV)I8KC(q7%#%I@;bFnUgjdYKcVSqjiS@exg79w%u zMjE>Vua(vsQzsXnvp#Y30gP{<29r5LT+%?I zqb~8FdV%Wi*KA=L|Jl@Nhae)qlSUa`0CiX{k{4!meGejA15J%FH;vokP?gxw#`% z<^j*2+z$<5k0Gr6h*(1e_0j?dsesUi5oSh4USzO1#8;sMZD6D_7*b5sAWSt(EG|jv zHF&tlM>x31%zIes2OGnlsuZ)LBdU~d zCT`Hq$HBxxyd*e$Mr3?E2#xpW-JL_%z=lEBN4^t}3|`RPL_Azj?!w+ee*FR(c*{p8 z3V6@JhBA0;!lPiIqmPY{xr9ZWzpam?g#!MJMC2dAVQ3acCgytOkCUDEy3Y5j=V#rm zYJJ|DL(fr~Bf#}}H$~`Fc9ZkuMH)p6F3-YeVkgng=o5YuY<=DK~v1#(a2lW zvccQcnPb?NgT1>RTp-ibdpd{P9eDkE!2H6pwc3#(2jgGAqc=NZ4eqaqS1+Rn_ zWzSK3|7}@d=KqQ<$pNfv|Fc~aCVwJ2NP*WrFnCImw=qUkL3-ku*dQ{Rdk=FNhPH|# zbU1@t2bY$X$@muHJM)q(9W>Ud*UdcZ(ZnBi-r94NQ&rZu+V2?^5}7A+YlO1D=a)zl zzH2%xp!&RY;cZpKh$ogsUmxFI+65l8jrUIg=JVgIayi4qOZj;$&G^@#VS(-~bl1yq z?OA-!22C)5P>2b4&Q7j%Y8;Q1F1`)5;G70 zCi2n|HnVecjfV$+M;PTDT=71+62g~JvJt{!A*lhkW0p`A&R?Jgkfo(iZB{}T&v0|O z@K8Q5>NGFr8qgtVSLHy1E2Dv0q})0APp7^Vhog)|rNHP#Bi#Cs z2WkJb`=@t?SaIR5#6^_yrLfacHW%fkBxdEJuCU2BI*ibrgzG^qI9k#BuVz?F*@JR+ zwNm8+gy4g1S?=VLw~s#Dj^o}lTHlT~SV4T~KkX(RXEfe$HuDIMl)j!}6@Oj-zmepR zLU>958{_{5>Jnuv_Zg7{ufNc_UDPE%sMwI;%yC8ZIyzYC_;{M**;tzuGJiDLnuiDs z8|e*e{IU7aoIwMPi#AXBpziXrCr`QII@^@aTSY|`_lnXwxI$%mGk(wqz`HtFug<+J zczbcRTvXwrnSek5(AD__I4;-Z?_QgKY;cIVwlywPF^rM+>3~-q?CUT9bVNRNo=10S zRDl#>%-26Lw`X|#EN<^t>hM?VPbyw}XVuP2@BfB_6ANw$yJ)0GPHauWJNkraAo-EiY z3VWMu1dpCwEF`D7R3tyjO&JzyjI`PmOI}PQB7G9(#jG>|M`|J*e?6*%ZJ;U`?+q>C zk{7wc5*UHeM`jG5qSp}DA`NJTQYdnQRUm^mPAwt}OvA`WoJW%@qaX=1o%mwrHGKLB zL`wye4d63RzqRqS0&{_!?A}hGY{zdYe3~qB_ zPq-wi6)FrCE3y?zLc27Os3Q+0*bADC#qu@zYCD0d`;XD!ljVRb`yB-t0!hz(sWhR1 zF${WaaPpVn*l0$~kyp`%ZQ|S6hSR^#uKk+m*Tj~rE?pTI5rK({k81%Y6x@{EM0gC~ zpH|qUpv-J6DenwNu_I2bCs_|eG z+nFPwg%8>*Dj2I0mw>~#%<{X$?HU(JJ+q26Wrpyn>l_axhoSe^umgobMF{%f{ zs~xNpt#3L9n8A2)J+0y{y_e<(yVT+aD-@(#1JTs`H`%3@h9*# zd3@_r0rqx>dLh5l(8(u|Z{IM5?NrPNhKa2aSY1ri?YQk-x>*v>Td(Zn*P~bRZC=hl zNSM+4!BQ!gB7f<7tQyuCGMn)!0}RU_uDd?fD((Ov#kj-hjuFDAiN{d)R~Z9dg7-kU-8T^~ufZ~v= z295VqqU2lmOoZrW9x#r6ZZK=ClI`6r{6A3h$wDoJZjiEBFc1M<2us`J?^-p^&oe?X zJH>z+MxP;$-pILQciCJ9F128v_?~BphTF=YT7LeusTgK{St)5J1JD)CR4E+P734}t zPf$ziUc1PD`772R4att-X7Wdix1vr8sycD@hgM3ia(K+)$l}8BnY5x96A|*&CWTh| zc+T`Bl}B37HXvaLBqZE^?a%aHfMgHoc3D7ux!-R*YD3=V(E&Ol){SpA}%A5>5Q+p!|_!Wb9!xQk)%xYd!-{^xrdHF+HcjTP+$P6 z)T!{-#7x?WeUh9Q(Tkbw!&}05pSeq3t3@dD%ap;( z%N)*BnJI+PP^c6lOS57GI5sh5WZ^;F>B_T`gz{B){+feh2+5j*J2)VZc!66KVFvQ= zXGG?qbu`ldQRi5A zK}Xjd08)_?ZNm@4-d|8Mfy4c{vDY#-Y?LqM(QC;)+Fx(_`@pdR(})xDj0uq?D1}6X zG|h@-{%WtWXOw)#HY8IE4AQ0LA{Qk;($eXVY3)!rU-E#O;dLnbsfvesGI4F9@%X0p+uV4cq85_;Kl&$p z6hqd!-;vN3ItCKb6IqXUcdyg_*xL~dJxMvFB`cVRaumX4dot|pVGRMrgY=UO835%-dUFWkxXhr!1^~89?$$Vpz;Aoz3g5AZ{kz+Eb%j__;8e04R~E?VN7bJ<;pw+hig9jg7oxdyRi# z+|$?V(-iQTjzlGJW-CBWy0X>3KFv2X%s(*fQ}B%8DG})E7T9I$7D~!$P#N&~N8lUE z(_Gxmm+zZrgJ)))uWx-^2j$`r&xYXi*t%z7)V|;|K6Xd^UF9o#s~36e36)}PUChmM z9q=ytCwMB2l}d5%90}!$XEdaT+(1!pS`;~fHYcOi=rTPD^)H?7!zlK!94>JI$)&e& z-(^=P$>t|03GSD{+A^r6a9_X%DAQNOQ1DH&ex3e`ofQ~61WrU$|BxYSZCENOJ9LJ> z%wVzpd>vwK+6t`|J3Hh%(mQ5G$mB4F9sssZbj9ET)e5>9O*5EoSXGavPG!Zy3ceXZ zGsgNKZe3I=YBThDq;+qz!FugwD$7xLoBmZQ@abyxU|2hz9KDlM6W@e8!d3fV5~0&9P$Zx4~!4$2mL`y(znAerU>K*YEuw_%XsqX-)&h^jY-x%=lV8p=So-SlI-pi+_(v~N) z*?I0YWQshPbK+CpdT=D>Pnvp29$-CqiODey!v&=$fPq__(jK*rZ9}bS{=lN~0zmYFkKlI# zmL6(}rqSjP;{cW?O_*o-0di3^3zSOo2L-v&pTk~LmWpK{U2pc21fruIXXbv4jU?@# z1kYeeKVI15uEVPAU@sg5H)@q$jxB1#w58>&i^mhn@u84$oeCS7v5LF-BM+;U$fJTc zp7%rKvhq^MO8&?1HtG?NZUErLdg9B1NoVaqjtYLBOGIrw)-AD&EJyXnFit? zD7lH;{^)*XZyTf8H{fh8c4$(Df2nx``lM073jp(NC}v_eyz67{#dK9?tu%wN#?sn4 z5krT7K>JiE@qSaAND3`Al|n9ITz4y59WWBdkL`-y*ACLUV82K@F{&bYVzN9E(FtOtQQ4p7pJ#01$Bd+v%__O=>hlS&Fe}*?_=O zn-ki3Te=${V?}m`#IVKdNgD7Lk*T8oNL%TD4mEcsQPIzwye*TqD@?;{>dRl{GQFPS zDrgPqo=W5AXCNN!?f=_xGt+Pn^99`R&)ALNUc|+Mwbe4>pZr^5eD#&~M{fA{uQ$x_ zRdwvT7=VR!Ei5YLVb^>Zxwn=VO$PDGi$g9=UM{2p4C@>5(L=Y=YSKgR5aYFE^O#6W zZ$Nzzd7-5bQ=U4|J1eX{`&?MYa98S0DHi;v;jh~vD_tKKY=L)!U)X1_Co$0bywI8r zF#|@~W+<6HLqS&<=@fl#_Vgo?vB2)}=|Uum7XZ`yHfQ*?ddTM9eY^W?Sos+)tSEZJ z{U$+m!C%1IT(AkD2C-yHl!68@4bH$@Ck=-yZrD|apBC5-BH1^ftl-stECeDL8ut*I z6fAvjAu@k@vk_Wf^I|g~-oqUee3{jYVk}m8l7(z4#)BOKIW5&g)QA{)VGH$bk@tZOKW5$EFoh?VAjgi%GOA_jU z01{&V02&etOBxzWET{Q1215+g$ryjxX>hwXAt-IOy>Vo^?4KFZ*B{J_3Ib1S^Z4_w zFoONp)Ur*hN>#oLD?GAOvxTtNv=mN{82~E{GP1$*hDOedp3c!P%e3KZ_QR|8?AU)q z>5U--9scKh@eUft=uvgePOjXp|7qso6a%pL472&=McwWuTHhMA*+XwSq*cB?Z&W-_ z?iT*crlePaXWKL9GM7<-Yevj3G0HvW(1T($i(?ZE5;mmQLQR4Fh{cZ)hWzRV2SC&2 zZ!MV`p{E#mmA66%RjyYH-e}<)5#9UK@uGSerPr^KXsS_4e)+l<5LxDB`zg_$=bRD| zwm1xuM0R&{7w5X=G!15Y=Ye<{xE^WOsbGqhewn9(OD~MD8wWe(DwwSbFL^ zddbj5E}D^C8DCPG8t0=Vh|;y_=&D4D;{?{op;#+xG~;(iAsZtPFyb>O{g(Y?M?jf* zB1B8BAK%qul0@mPOLL9#JTYwxecIDjqAFwRu|#XcrxAci8V=k*c!_IO0JzL_B!$t( zjBBr2eHzct?q4qSmzk3H2$=ebL@pIArp;2Y2jLi2EkjP*;flb&8?V|Teb?=~nlOEc zcJ^AnYY0F*c0^sTRIMmUYrix;o;8hU{9-!PHn{bF%`lx-bE@07+B~#%SL)>XRTIXK z+ZfzpM}@lqhnEgsGKm`G1<3cIrjLNQm0p(1e=#FWO7<5{eh(?FMN6zROF$QBi8y@= zG5b5M1w8&qN?kj#njUh?UjHZ6r;CXPAXWuQ(KUy&;L67{EDo5}Hc9gpmzO zj=<4->e>JgVz&5o!$ZWHv5n60x?uZu5Fpc!`P&Q44R(`I%SaC^Ml@GEgo6NL%w!2o zlve%}NBspIElu9mOvVdvFaoSiBu;09g3biJt85h4*xp=Q8GpJ(w{m{_g3!7Mw0-_^ zU@tC{C`8f3^F&&H4+x-;-;h8!8RqK6GgxtbQVO(ojkv!z-ZAlSq)$m%ELobp*<`>! zFr{WuRIlfNNFgSr;Y&24F>K-RR z3@n>NqZdDJ7X_3`&KZ(4=gt}FI0SZN?G&FtC;CX7K8ilv>EI>{?&4Fz6U zDkRSdXtg;MP*qcP`ftv!q$&s}T#pK3%#>lJ5z8ct4<1oq944EnFq$#-4dk8qYVS%L zNE(Gw0Qm5Dt4fy^(6QC;>S-A|@{7~H5IPA)`*gIu2W2B#qVc*4F9-6vWFVa;xMumx zz)foTPCoOYuTDqfhTZQw;8{hH2sGKlD@d`M-f*OgBnlrifjZ5vUj>hcAyiQ|>q1dx z>{D-AQa;pWf42vp^Tuw!qsz0`)?oMs-EdNS0nW+1uyz9+Z^ltTJhlC{{*G8`;pvL7 zhXia|U~r>6`%=}$QAKa7H)m&`wl8r+tMuoH9u#mOOW&UF3_8z@^>HPS{u9EeCidy_ zc)KI(numZG?6HTpfxYf+G6nZgBMxk`t9qfOR-f)q0Cdylif{+rYzZo1-p1V(1v$NG z0;H2fKV#e;yZFykd%|0bJj=QjpO8^aW-l>n*1I(~+ixGYas$2e3N`L{yMe_|DqA=pu3g!~EC52dBo z&RWBWLdH}48@g$^?%cMn%7k&?A&Z)04IC2Duh)~0l;y>FAUXN^EF@xwL68qs{E zcuBcJHsD3K9g#QF-T1BEz1?|UNgh_IkBq>6{P*2I(Z)Cz&?yI4H1otlKA1|W03B6% z8I{4rpuhp~e)b#$uKUHoaYRf9U-#BUT@~h(FLQTCQd`oJ(g#o>+VB8@f5vt~E>B_2 zwyXg%fy}f(fpgtBzEe5JYLh>3MI&l@Dk`!hq019agqkm1t}|kPG25;k2jyr}@?%m` zMIjwQi@Xwe3=fNOEKh2Q)Is$eKnq~8yC&;mr{3M%n{+3TBSO3t?XTrLv$OKs;(god zL{YPVR!NL$fJpS3{s_D^LKtnUbQ1=7!JGPn{XhJt~H&QoQL|$7yV{WvBv4Z%rJK8{2 zFTB@gc;W2E@{O=WAjokX2D0b5o^6D?n1BBw?Q$tm?nh47O<-i?ftE(i_)^KeVE+d~ z^=Uh`$0e6c&b^z_wyEIi0M4gJi_{1lL&C3bK0PS+1+P0>=2TZ&z`OPbNEd9^%&N6B z)H{xlPeAY|bi(+zAw5YaE5mOU*BxySbY;dRs&)ZnF84qN7QzFc*%*`(uNa|Sd}pc zu`mm}7p(b&VC%BkvPaJc|AlLg zmxq7Iyz9V>Z|HQ1%?8H)-jG}C@Al~u`S0st;R;x~DpQ(NAP}<1l{)?k=o!881J~5pK$dwZSGqhW_#W(FNd~Ya(2_}Nu zH(2>ZcXVdXloI^mM&Bpmam`@eJY;P9XNCHq^I26ZftC2_uhSxdS~K@T zL)g?(ZJ9lPz|fu%zIR!LMSA*I9<_H!7E1Hn!YF7>#&^_f`_$l#N8W zw_io!&Vo3a6S-eA^gz`is8p?X3|dgzL7!M&Wn=3|P#s|}lTb;86s6VfZh zt}}%>1Ge}8wxaY|Y1QK%`$LS;J&rUBVYXv{9UdH_DW%pX7JixhLzo~fXyqcNi+v{3 z8P)@tJa>NH(uqC7qYsY zdLN5W$6A3nkq5V^M~6pxb!3hdMMcG9X+@eTLG*ejs7adgqq)Nr}SblG7Z1+LwI z%=^VAmaLg_Alzu~sW&eV+N!v(_Q}NTQHup_+3!q$3JPCk!BK|=?Jv_m%& z1Ud84XnKC`Fz0Q$uNF@xZ697%v6UmyjH2=TUEG)5lGj6uH4He@)+X)i9;V(KujA>_ zD-&rsilYHt940Tp>S+JV$y#mdof{K?)bBG7Gy3v?$9xcZ>Lh=&uO1I&xq+Vi=GbpJ zG{pSzF+p=+&~tsf+xWBOe3|KL%x5fm>{g9}&C`<3=Uz~`eYiqt69jRTV~{Rd^{n_p zlgxuFA)7|Wm+Wz5+STY@ev0lO&>gokhSh)I(Z|q}FiIjOvUDtYwR&@itgm%|i(lCp zEp`2FHUb>T@Hp%#Ty)T!&Cy#ZCr8NX1XT)kvc@B4*pWqg{Egp7L?VCmVs>J&C^JM! z6goK91O$olS_*NgE9q*sUFUYXvuo-zZ;e|c&#%JJwUHf?%o1KyngmFQIw|(qpz%}q z3*#Pd?9B+zzRg%0zw(U%RequXsck5nY?3Vu$bmp{;?$QwUwKoH-E>*;f8DxHT0NWG z-l#pwgl=C_XuKbK?*Y^96rO5dUhDVph@adeVVE3I@6ujN_Yl|-wTn`J84qG#ztC^d zlbgSUTzzz9Q>QSR=j4HsJ?aWGthBYHrQsU9;*937$AXRr?>fC=mHJo#NHbTni^0^v z(%HJ&s#`0PMDNx9Etde&m{OUWKw$jqQR$skAkF#(+_x@H)Du)ALN1j_+T{1a!1;B> zHJ?yZRGVR~E!Cbnr<;PYj9)g20z|Xg5BvfAo+O@ey*2lL_f=pGTGuvv-I98kiG#MT@!mvyy%shu~^xTb0Hf_kctQNA}bzb&(efC@#s&#k>W zCu5X0MJm$Cz!eCOxnfC1x;rPMOewv{N!w+g`a@5ZA!l}ftQxlh&gx^8{;}b}JLxeX=mLexv25QTmg2XDK-I$eGS6Vz9P#&@c)xkNKWllP3 zSZ~yCkfrU!8aqR`4$~b_ExD4dx~-!}9N%n9(CBoYo!D{5660Vxm!oE=z^QUvq!N_4 zFCz&So%&5HzG!L=XndoEIGH<_WBe)}@6KYq@P22y7|aNX!)h$rs8#HEBbnJ3KR|ZC zxbO!ZpzwF0IAY>C@2uZ!sKR?3znfNsSD|~>tBIuY)A*B63F&*5{BmsLH?K^{{-l9nb=iBn=MwYWEn2st{pz8dATw_sh+L{* z*~C;m#J7#+7h@S!a;A+r%y~~Id==e{Al~3+hr)15=pbT{R-}6@Ce`l|)g;|=y?#-i zI7l0`({KlRz~QmFsmUhTE9(Oq5e+MPj{iwkD4ysKzCNQar{nKx2~b4 zamB{=zl%FCdcUL&IMEKrrz}@KW>|PZR+IupWr(U@fbW0tCAl}lfPX?4Xi;uFE7<4b zmGEnE7!oxw+`5#eh6n-cWDAq+CC?EEev9PA1CUm$CUN`5T)dNA6bDP)v`V+(g`{Kw zU$?`;ymY{Yx9AiU>bkMn85M>*0y?w20H(& zk;OFI0SWSGQKX$!JCXi<9-P_qWgz8R&04@lacBfrUxYrK`{t)XpCAacA=%E)eY|12 zxq2}{nTdY>@lJm#Y9NYOadlf!Cn;-{ZPSo_D1)fTTO}=?&y(y(3L_6-U+69<_))v# z)Q}D?8{l{;O>&{58S{`aYI*py3bXNy1XT6u07T;REv2e~Tc1RAw-%F&<9ym96}%yM z>cH#-)x7e&}2eyg!V_kEe#ws}N%LP0 z^{}hWI99`tz-8+Fv}J8nb#lzQ3L*Y^>+^Z|%6WBzX7YZY3t~YJiD4;(|41g(1bCR( zf*dMiM3Ef%kJ|N}g0lUxOrrwh%K0meXZ-b#l#<# z5#=E@SAeRED3O?K#5k-Vcdo@W5Q!miVz-oA7m|6~+q~k=oPh1n$sHm-ANuA50i_#DD!OW^e0QM)GDtRm0zCW>xJx%*MWJjWC) zxuw(iyHgNrRepv&D|*nW)3J{fq3(E22iIE`vUkVlS+BdDOHxMkbTG0baN!2Sm?F|% z3+PL|HrIkRBvk29KWaL^pGq}{p_^dEgR&+pZbRFi7y{m9jGNq}Xwjkw+o* zSr+-VTqw+c)50+;Q3wQOeJlqH1E=wZ_nMYjM8me66JvU_ckAQc#W?BJRSu%=i7!{Vn z`Vxk}U2;`xlV7W@ZJu&w(H?kpd91{=87qotbV;MAXbu>Ye-6#Kc>PQ6`i$idhaLhp z5!#X59#fkFUwAb!R0o6!9!R3uQEdlP;s%DtHoR$_h;vat;Cg19=)0y2gb!>e+hE}` zD@Lmy=-qE>>Ht=XzXpSqBrH@kh*_}9+y~76p|_PkIF*+}Cw^(t5>SErcpYIblR@c* z*d_+js!Jqr%n=Jouv$t^v!f}f%!)ot<^serYvYL4XU|2LUVcGLil#r)ouvsLCMj2?O+WAt$bo9BgjV zUtHF#djRJXay%x#+U6~P1#`L80!z&u`G`--<;*e1q&0k}KTvw#o!=24);j-Xw%N)z zISe}vD5_7I2-L>$*D|Yd9?CILTXny=!F9%Jy!VB9qWh#C7GG`|j-)bn#Ov);;Z1DY zoVS`{gvzbPyJ5hORQoTe{B~)tb$nJOZkX=~E+pTP{fz#zXsBZEo#- zpV$`IC6z8kY$WH7!R3EhqR6DIQ^hSwBoCB#ka2TnE|3Nq4MQW26&MH;bJY;w(?EsS z;dmolT{X)iwL@dzbT5Bb7DV(zLm8pRiczXq@Kvl(Cd}T(35^)3rg%t7rnElAm}2R zAqcF&{aebW0j`O(18nqQM{0+AE3h{Itb25mwc*9IX0A}Uji^h|J{7COwMf_&ETcL= z7No$m+7W&@ks-Fmoxj6sqy30SY?kZqn!`Fx*kM!VRaYK~)11!)eAgU<{# zr8augW|b>3L}@VQDx=NxwYueq=6!(L^1kx(y7}JU0p$`OvVDnJa6rWMI%Xy7O4`H$OG6SHG9fX-cAh$B`_rz-%?&`Th!_wFJCt0sVcfyy)FT z&57YWc?wraMn5nyVaBdf1@&duu@nL?q}OOlmyq_Fq~+-j_NJxzz6OMrv~4_O@5@wt z#r$nk~P@DK=0S{E*2Vis9VD{n8#UpV0&zvTh5);ut_ai zJfmM(!PIn0Ta(U`O^Hjhxu5z5xvt%0`M*RV7Owv?fEbtIQHbbXbU$Ced%gv}KR>@C zfx)mb^#p-J$uQ-CP*F5tfZSwKZp&cNekxIZnAQK6L)5ekKMDTB+We0jlTx|~&j!NE z_TLB(fRp8aSin^tkZ!1dm!6=Q`h54nB2jojS@I!JG}>^j-yPc{!Gd?B>h<5Y0?CKB z8*R_o_u0>#$6EHWyZy)rAuIrs{o>2~gV4m3qe-0<{J2D^OlL13zdSOX2Kfu5V3jE{)V@(YmZeveU&R%8-} zy0Im$8?J< zQ3s+>Nl0pvLP$dQCzCuBHG)}j-ed|naDbu^nv}Rs{@)kIFzMj$!olCL%!>Jj1O5*$ zn7!$Ey+r;TqEx5;QP2PdCMt6#|J4R?|Hf2IAukV)v-md0d6vDS&;ff#!&T3@BXao7 zMD}G4o2Lv-_K=CKR-41K$=-_NLaBN1I^sG)ZgM_iA}Z^tY50%*?F`YqgHY1I&p!1I z=|9DU41^5C4Y8{uxJLAMD9i-vft``GLdnE6L^UMiap)sBBa#3k^gHr;2=$0cF?K?n zIGGVuBOH{-Kti6_tPy%6hI#^MG2&QpQj9P$lvq+yEM%gxP)rFiW@M`PsF66WDX=N; zDFIFaJN1@Y-SJpsen@z+aISeyTXTGSe=O-Cbivn|3hUWgus~0YoAqVBB9z(bXBcR}ScQo=Xh<&HAm_?-10NbY%0;<_9@t?aBImi8@B84?adSX1PfeF+JtnDz0GFMPOBz> zMnC79;f4>*TL2?Jmc&2v-@3e4*mbf>SYX&_801*PxdtDd%z){Qp-)q#;%_h;8F3Wy zq7+GN4ZcbIC53N;WYa#pZO4^cEqZ{oz?f~^GIYVfEmfa|v5Ik7#XsSL&Sua4Zpt2B zjC1Zt<_hb>8}k=5Tp2JgU|t;r*ndqety_v7d<3%Ucdo&5jDNGyoQ~b?aNvTW83VL7 zdbiX2tkfTP>%6DSLyKM(hHJs0+YzlDS0aCIE%zL6lp$U*2wo7(z9GZ=njwH;SW|GI z(n*CJRExDV{n_Y`h;yr#*UlJ~Q?%B3gUz$3VGkXXD49qj9;6JAtsw%zwHnEO_S*Gh z@Z|i}RpZ!2L$uRz2Q*l(>cQWOw7$2@0&7RUQ&*)M0DpqFrIJCbVm+cz6f0PhU&~8T z+dffz2eL<(`le?XtR1ixtsD5}G3c3vH>IP=w>oO+v8-hu`2i@z(yG%JGvMKn zHYt`iRqb)In#~FMhlg*#ZQM&=B5z0W{`iqSf~tz6kM# zV0~bVVkHnZb+lUKsG3BqOlLy@iW=tLFH-q#3+O2yM30)1h(QR}Bf@nsE+M@t;nb*5 z$9e`{Ju2z~W<2%~HWGjVdvvc??Kd`uQc-WTH~7N5Wvuoxb4TJX(3-F%T~MgZZ62E! zQ+Wm31?4(e(zpO>3eY+{3Vq)l^r7z>#~HZ}tBlU1GRc;aM@^exbjo7LB&C&>8j0xS zfcSG{t@y`+IRHhuEziDH(SlwfRwj8bU2-DpFfnW}{-rj1;tBA+$@r8TMWzAqnfdn! z?9l3Kf;{CmeV4d#^NQOUspapfv8YkKYBqT<^H@?*Q&W=l?|%Z!yz*ChCI1g$=M)@B z)OPD+l8J36lT2(onRsH`wv&m|v2EM7ZQHgzG0&XutH0{}SLeES_eIyPUA5P{*0Xqt zBq3?`)}50)i790pxDZi<>sOIn@Egb-vf^=-&v4y`=LX+26 z_l}PbyWRQg8>@?ewDTUD$|;Xl9OC=nL3@Rd??Z>P)S)ecdq$vbHf8rsNU)6(O8z`@@;=n&4lz@>g*jr!Nl%6Ha+mjK7il84r8jZ zQqm6BoU!W(TB)^y_;e?&Qr-^o_d9=Yz}-*xm~xT_oJyte!!-d;P$5@z3^_9seP=t0 z8VN9Ac9G$=DVy+mJ{NTy@(M z^&+BFXs}JdERRaJ5o3A+QXM9J>&P(w>WZKYwz<|}F*yZN`i(bzbvTprRb@x)!hw+8 zI*jUJ(7Hv!rlOkKg3^MVzJFQ%Up0%4ZzsAUE;qMtYUzWBTI{Tn)~^PRv#YE@Z8~fB zG-452(s<0_sb7bTT!G->RiQj?)!)11eu2ZBow|Ah%2C|v6jCTA2_|f8>>g?A;nI_K zx-Usd9vn|y`|iPRm^WTs#3=hE0X1GnAFke8{snR))Q+HDoTnx)M8m_7T%5hn@B?|{ zh3l%pJF&Non^bdpOrW@1yEc`T@o2^hn!T2mGGf|YPm`0?n?Lcy$#db^(mOHf?cuR~ zBaL4HTR!T6r~c=^73gz<&n~C(G-fao;kEgryuHkSn5N4in_(gfRrVCDmj~lrD=A4N zMa9+EyzPCWE#Sl_6m56btac`d*R_WH^%&wvDOYdx$qyiHwcFk5{WkdqIvu^uzq&g9 zJk;m?NFnU>j9Rpt%N~i;Zk=O&yH)aXHg2p0dd+^&D)*=bXlrYC&k87Eswyvfyia9b z@XQ6PuKWeT&RrTRuh9x_nA()vl22a5X=3Ifrt)3sk|7J9{^aJhd@x_)Q>`>jG?+31K$S-x>V zS?Eij-B5Ju@L2Hh!7_MTdZMn%v1!Zv@qU=#_wef)uLVL9k_ma)9yb3tea>lXmL&C2 z#9SnJELZW}$2&y(tIXrRB*>u!3!fz71p_DNCKkc_?NN4<36MY>Ck=6groVa@wNYcQ zYs7gKmCsD=<`npA58Pj#iRT~p{53a#UX#(#KMEHcpZL*L&!FBb&Dqkq0L~QnM_d;^ zyBq~#)F|vcWGPxug(tN!NxH^l))CtnvzjDES)|*+pZKmSaAq(|0ZLH+6NrV8c7_}L zH0y6?C>1qhubeMUE@|9cZKaQp2e76H=mlH92%p?-d7+9NWkpHYbH zO$lW-^uEu$WRU~?B`*<&%`Zof7&`f}AO995iz&FNN)^qBJefSoW;W4Rp%1UlKlL25|7%WwsfMZxZ&})e_INO+BR8m!(4XQ ziKsqFS=X{9=}bh#lzpRSP72k_!y=9E?{Jj`A@A{Gz%cm*UweWC2;iYMI1=>QEl+Q^ z;bmjKbK-dwV8K-NgWPSER@-j$@_PfY{=7&q{;YPp&UDs+1fSW>I-1WlIhpR}^R&-! z2mPZ~=G|@agjtNrEr9@(_$qo6WOUR+$F?eD>tyjL?B#G8*X=(1TXUTMppqDV$3Rl{ z^k?X~bZ+{~hb4q2L2*{j>pSA>@i&${J37Ruz~}AV-wp!2x^c|RCj7|w-I{IRO^5EP z`d%TeW9eO&W${*ej><9#I4nmtHMb7unjkNucz5lgDF;?NY6O8>;Jy%IWPvzt7e1gg zPD!cZ$4IiqB1UCog)Iv^OY>`w8cPY?LA%eQdE+8_LLB6-6U({^t{X})9tX~p5A*1- zJIQ)N^QB1=*X4nH3%qrs?)yPy_Pt07F&XvWY)%w}jK$$V)`&F3U%qSN7tdPGZ(I=M z8Tn$rSmDUS!Xp7uW|R3;Rv4236q7qawyG3oaj6iH^|KYVu5>lc_4-PzW(+hnLg3_* zkB&EKYmz@1pe&DeRmJmGiL!L!f(PMBkK_3W^P07YNUd?!1J5_JN9w)*!BvcP&Wbfvzn?8z6~CtTML8Lzhh>@#KxTSH!EVB zpBCtww;0%tXHdGxB_%u0DYm4kfXun7ukluk?K4{^EBkWJWPZrbg^@l=meeBqa;oKo zb3yBxLkj@IjS84yQj3xkcDNryJdU0(r!aymK3Y;BMtpHjwNKz@c1QfBu0acgx}FhV zmEcUIMSn1sJyHNnm-O z5A%N4tjBilxHA!;>!r}w2astFj-3|2kT18RYy%PNnLkKmDD$XMzNo1z>sZ&M2hi30 z(jq%!zP{U-gWO3l=?FkK`s0*sC~k;@LoxAexxwSLJW*{-b9KQV{l#6}giM33C7f zo&rSCgFmp5&$isy5*rk3q(iS-Jl{w0xRsL*kK@{z_G;Y5v$DHAUwYpblicgW88qZo z`DPAIy*vrW26EgRee)P1MNAv9rDeExsVXf0V7y$Kb!YSkU8{`*V&R>K69&Fr%I;>A zd(k8vXZRX!AI+ieO1^WCkCEU9jNq7F_W~AWRo)ugRS|!X1v9CXLn$j(S&f|odBupi zml;V&bm6=%!#V_n@KheDT=e=D3Kf}~)MZ9OP%-SV^w3*HQcv?PY9>whxBTgydVEN% zK;Q$dNWGFJ5nUCZt2~1h3UeLqLVf$cG(zjtCHGO&US!%Awf2Zk`oFY7>k7lr^#D*Y z-(V9Syp}ocwPi@DBlcd_3Vh#x`5)Gu193I%!XM2{mU3+3=PQut-y8)%8PcKSN6nm; zUV6pFaLb4YZ4Z^fGBqlvmn}V|(^Xjb_;p)Bl5gwK{o&nV-S!e*21e%>pmo=a<_f_% zjSys$jPmSa)ivcNj!ZFDzz0s)2my)9A)@$Bf`r|t^22mBHf0;q2bk?|4qhNEspFFu zQ{4vfiZPHubpfnti~09L_K0xmR0g&x3wM{UR3)_dXO{rT+swKU)yguH#+YR=GlAt} zcw%ZreSYfqle}Egw}$KDjzlR;YDs4%&Zc(Kvx6^A#sU+y1-{`m+KVGY3xJF>Rsx0W zGc7H3dWNDOQznfa7ZKk4W6j#Usb#c#b{-I;-Wl{Yl}|)6qF@~A1}1s2M*6PQIV!TA z(+D0>ANnGIn1*8=3eNt*>CdCs;kO!GR`o~><|Bo~-*JnSRJ^wk^W~G@Uf(x}Vp&Mt z_z9ALmP5gD)kalukHo?j;M=+6eZX|6sfihZav6cgN48fYl{1laDpfQYVg+9fF`A|; z+u?obQc)3(ftXu#(H=SNiVq@O>Xt|ax?#t(V`qcQ55f~|7-HH)6ms%3MLYO&=Ro`w zc%#x>(afMjs-Ik}y?8h=7kT9~MT(-Pw{jvv$hGoHQi2_v_A~cxfR*~J=@@4E$X_t9 zk_DrMq`g24!`AQsAb0!Ry?Juc2B8jjwRWX{9|0l5CHEN)?}ap3W$yF6iAPqVjf0nW zxu~fuG!4u|pNwp67>$AC55=p-6x((wkX@6YA(=ITBp&`#^HKjVU2%qgHG%I}a}?AJ zU&`M?SP@ejT_5Uh;J}^zW2{2HZ*(F zLTSZ#6SIO6EtAn0qsZJ|W_d59Nbxz}ji<~C6^(4u)aC@Z9TB=@4f#SLa^8_n8Jw*R zHP@{0P?A$&R{u2NnRj)lRomJ>QO@8JBiAn38?L&Iv-njXAWUa-Akdd8BK?eM8EI(!;x-P$88~5(d~t%U)IhycclzuBlBgxK7O&XP(Rq|- zCA6Af+j_u&qN0*8L-jz#N;mE{^>qzM{;0`M-lV=)g$jwIG2(BuZnJ4SGgkr4SHQ>S z*8)OO^f(BrIVlE87s@TQUS~?OCC27PK96%#6MIE!FfzDtip59s3 zlnW~3uQ-&=?QtOlIm9K{&qXXNWF!=g#wc{bz&IgvbLhCeT*`zKSV2IN=A|A8i!e2`?2hnw~_?# z$%F|&Z9n$TLNFxb7@KXv>BDI;k1%I@-=d|nmsq%7*W5Gslo*)_(2Q-Dt;jE2}!!g7@nvUbAW zMS-0x<)~MB<@p;F^FAu0!u%y;K$a)&KTNg;XQ;k49VPwvAfefP5vaIx;J!!P;oBF{ zn+g62`f@la<-h4E+DF}?yYq^(X~NndM}Z~H#~BimF1Rqqqi)UB19*c0&^>;|9SFs zYT}~8voOvs^V;<~bKYAdP}C6Aq{zyGQd2bAMASrs*uBkS9S6E)cXivo*|sZC$h4;_ z_vHk6YxbS$GWRCi->~{WqTvFf*~T|Gx+zFI8A2)aAA;EGB{6l8Dlj+l$bq zr6wxk#>(q(L^-)yRe?$kq@PGcNF)BH-=GA3qC!QFknoTY$x>2Ps+6v|SL8O9_ZoQU zG(KtFHt-%Rd0bt#eyvC^x1oahJym&_8k4Tpf2Dq0+5%sn-Lymi4-b^@F9w(8tIaZ{ z#WEjsk3(oK|ppZLig4CvND0ux(MooiYrcaI3do~hw7-HE(y zq<|id^@ooat(#k&iLZ3p2k%+u{h}w7K1YZjL?+1fp)Ui1d$!;hP~VxeqZ&A6(5Vq4 z&80T&$^lv!_hJ+X@@)-GFBOkBl`7A zdRSHCI~P=A`f>&1S_hE`CY_?NWRVfXF(;f>4H8MYawqdN)puj$FyTlV#X^~++ir_i z1%mw8;Zdhjdxu7;@|f`^v8(Wyz!HVVI_=QmgHoJUt3enbRf;5e98JS|tu2wdu?-wZ zr(2^FL&?D}#L>@SsP)`5bF_QNNXJF@nxquw zHAp_t?H=TTFnyo(vLv-Q!9T&|>gHPcSmz4cp?lx;nt{dkhQ>{d_b;yemW%3;sf4jS zLvpH8|C+LoekcN!u*V&{y@wf1r{7%AaMZNQRS5=|C|R>LC;V{q!nZ8!?kd_C+S!SB zeOwbT|qsli4_A8Q;|eSh@(^CWy&MiZE+HLTvMwHn4!wQYNmnbr>2YBp>uNz704 ziRL6AEw;_xL)V=bDf@X*Wl@<`lGbXOG2qw=J<&OZ|2j@%wX5JKTPedLo<>p>!)@t^ z1%+{>lZj3fUwKi>)QHaxtyBl;82b^rsf6Up0xZ;#Cnc3Mdti|S>}o+5U(@uO@oPr| z7$?fUrlz*!3PlgRf1@F{zO-sCpWKmS z8dXb~cyCjzsNnI&5IBkNDxSv45Paqky>l&YDSlQW-s^v`U=vv#Vq>UzU&g$-GryFp zs~%e$zbXIT?Fc2BnS+h$a7z^9R8%+EIxNW?Nm0#lK4S73SxEx4&OsaGXinD$M)Uz2 zR%ZF@8o2GEx_bF`GXxzstQQ~sC^z4aE}|nNJYKr(d`5O*@plOK;6T6J_>>4t51nJC zaOBWb#rb-h11FU0#}W{lvRVVcMym=|g7R4~Xf&fJgQ$ib2|8YiyRS)P0APe&;ObQjX3*B>{q-`0VV%h>6lM-d{zB~UI%VQ$Z_tALj&$te<9L`LcCaUUT^vc1m;B>v>9bF$UahrI)4^tOI017XMi}MX>(fn5Ir8lB z`^@5AT+>>RmWFM^@RadA)^Niy3lZXEtD3soer^c5RFUE7dQ9ynMoGx{3nBnU*-1_g z;#bCak8)jODn5wAarCp>MWYpE)}$QR%5ib>uW#PTgbn%DHEMq8qqDEFd<41G@a4J< zOni2F2kFS5$^9=9ulsH&w?cDB(^P0)RW@epU$Xky7xdm+N`KHjbEo8nk@9>6$XwChH3rU7MbK= zsl|eaNzEzQ$7CU`{(1AIn_^@At0B6gyu9`z@#P$w(TO6={*osy909}AE*KIqVTKnh z+t=g1<_&$bg3ZgH%?XDq=RO3}*GZJ)?yZ6&BrVcoa?Z>}RCT+Ngl7OD&jgvZHSEac z1by9ai(@UtXxxATR6a(tL^77@Y;nm`82i{N%AgtZx9%4w-cfHycJLRxe7Z5%Dj@-n zf7zYEMfr(d9Yr)PwZZ+Ml^357EgeM(`>J|rVa1jMOkz^@Ji_dDq&Me~oUg!Upf+b_ zNn_Kbg6X4FIgd4u2s@Bfb!l*eK$CD08O4W$?fAl#XQTT5xHo7?*y&=#|LvQhxYNvR zp``y-_lnWwrL_S3rOKZK0x;w^uKhUOAh5iDvtAcJKhu?k%P9j^Q?u-(ejP>em8krh zZ+TgllZ!WV7RQB)d^U2jjXjoB67bkLJ=kaFrx&cXs5eAG_$~n9i{M^9t;|?2{*gdI zy?X;zmpwVY4iBj@(yspdMncLijQRP_w&3@j!3^!W()w~Gu1pEzW#>O@mzh;?`)~-o zEASx!%7RrPKy#9`NMTZ(gd(;M{PDXUr*O!YwXsLryc*cwG)oHOF>IY+4p8>+oK?C% z@H6>G5jk! zB-LTVF+EhI45a3oUl2tDM}ws^>YZhjp)xXzCoFIe#aDAuZ0Rz8qF2Vc#&`7X1&6?A z-mYd}To`J!rQjpsbqIL0vr&U;1TqST$+C0+W$g+>U zP~^^`^>+m~PI+cC@%;FnKz~%pKn0C3r-%A_t*OiI?nx;|hv?Ejd=U$HywY-h`*D>{ zwd$n+2Qwg(!Of|tcKh>cRLKi(C!W#=wd1{?{D82BH4fg&Lqf&g`xE*nu>VUyrPQ>F zGGB>QnLV?H6;UOtJ!EM4dq{DCO)3(z3b$Tic?wVJxCr@3Mgn0mWFfc-ZCTVssX^(l=3~1j=O;E_tj{txA{t@d?1un5OG6t+{cl~v3Xahp9GaTl<^BT6 zeMZ3l6r_M($4QW$wrs-c+An_xQiX3