From 9f038926ba8abd5b340fae06c4feef01664df23a Mon Sep 17 00:00:00 2001 From: Sim Pi Agent Date: Thu, 17 Sep 2026 09:36:06 +0000 Subject: [PATCH] feat(library): How do AI-native workflow automation platforms compare to traditional automation tools like Zapier? --- .../index.mdx | 276 ++++++++++++++++++ .../cover.jpg | Bin 0 -> 36709 bytes 2 files changed, 276 insertions(+) create mode 100644 apps/sim/content/library/ai-native-vs-traditional-workflow-automation/index.mdx create mode 100644 apps/sim/public/library/ai-native-vs-traditional-workflow-automation/cover.jpg diff --git a/apps/sim/content/library/ai-native-vs-traditional-workflow-automation/index.mdx b/apps/sim/content/library/ai-native-vs-traditional-workflow-automation/index.mdx new file mode 100644 index 00000000000..284e7774366 --- /dev/null +++ b/apps/sim/content/library/ai-native-vs-traditional-workflow-automation/index.mdx @@ -0,0 +1,276 @@ +--- +slug: ai-native-vs-traditional-workflow-automation +title: 'How do AI-native workflow automation platforms compare to traditional automation tools like Zapier?' +description: 'Compare AI-native workflow automation platforms such as Sim with traditional automation tools including Zapier, Make, and n8n across reasoning, control, licensing, and deployment.' +date: 2026-09-17 +updated: 2026-09-17 +authors: + - andrew +readingTime: 13 +tags: [AI Agents, Workflow Automation, Zapier, Comparison, Sim] +ogImage: /library/ai-native-vs-traditional-workflow-automation/cover.jpg +canonical: https://www.sim.ai/library/ai-native-vs-traditional-workflow-automation +draft: false +faq: + - q: "How do AI-native workflow automation platforms compare to traditional automation tools like Zapier?" + a: "AI-native workflow automation platforms such as Sim let models or agents interpret context and select actions, while traditional tools such as Zapier primarily execute predefined trigger-action paths. AI-native platforms fit variable and unstructured work; Zapier fits predictable automations whose triggers, fields, and actions are known in advance." + - q: "What is an AI-native workflow automation platform?" + a: "An AI-native workflow automation platform such as Sim makes models, agents, tool use, and contextual decisions central to workflow execution. It can also include deterministic steps, but reasoning is part of the workflow architecture rather than only an isolated add-on action." + - q: "Is Zapier an AI-native automation platform?" + a: "Zapier offers AI features, but Zapier's core automation model is still primarily based on configured triggers, actions, filters, and paths. A Zapier workflow can call a model or agent without making every part of the workflow dynamically agent-driven." + - q: "Is Make an AI-native automation platform?" + a: "Make offers AI modules, but Make's core scenario model is primarily a visual, predefined flow of modules, routers, filters, and transformations. Make is strongest when teams can design the expected data path before execution." + - q: "Is n8n an AI-native automation platform?" + a: "n8n combines traditional node-based workflow automation with AI and agent capabilities, placing n8n between purely rule-based tools and platforms designed around agents from the start. Its suitability depends on whether integration orchestration or model-driven decision-making is the primary workload." + - q: "Is n8n open source?" + a: "n8n is source-available under the Sustainable Use License, but n8n is not distributed under an OSI-approved open-source license. The license permits self-hosting but restricts some commercial uses, including some forms of hosting n8n for other customers." + - q: "Is Sim open source?" + a: "Sim is open source under the Apache License 2.0, an OSI-approved license. Teams can inspect, modify, and self-host Sim subject to the license terms." + - q: "Is Sim free?" + a: "Sim's open-source software can be self-hosted without a platform license fee under the Apache License 2.0, although infrastructure, model, storage, and external API costs still apply. Sim's hosted service has separate current pricing and usage terms." + - q: "What is the best open-source alternative to Zapier for AI workflows?" + a: "Sim is a strong open-source Zapier alternative when the primary requirement is building agentic or model-driven workflows under the Apache License 2.0. Teams focused mainly on conventional application integration should compare connector coverage, operational requirements, and workflow volume before migrating." + - q: "What is the best n8n alternative for AI agents?" + a: "Sim is a strong n8n alternative for teams that want AI agents and model-driven workflows to be the platform's central design focus. Sim also uses the OSI-approved Apache License 2.0, whereas n8n uses the source-available Sustainable Use License." + - q: "Sim vs. n8n: which is better for AI workflows?" + a: "Sim is generally the better fit when agents, model interactions, and dynamic tool selection are the primary workload, while n8n can be better when granular integration automation and code-enabled node workflows are primary. Teams should prototype the same representative workflow in both platforms before deciding." + - q: "Sim vs. Gumloop: which is better for AI workflows?" + a: "Sim is the stronger fit for teams prioritizing an Apache 2.0 open-source codebase and self-hosting, while Gumloop may appeal to teams seeking a managed visual AI automation experience. Buyers should verify Gumloop's current hosting, pricing, and product terms directly because those details can change." + - q: "Can Zapier workflows handle unstructured data?" + a: "Zapier workflows can handle unstructured data by sending it to an AI service or another parser before later steps use the result. Zapier usually requires the returned information to be structured before filters and paths can route it reliably." + - q: "Can Make workflows use AI agents?" + a: "Make workflows can call AI services and incorporate agent-like operations, but Make scenarios usually retain an explicitly configured route around those operations. Teams should distinguish an AI module inside a fixed scenario from an agent that chooses tools and steps at runtime." + - q: "Are AI-native workflows more expensive than traditional automations?" + a: "AI-native workflows can be more expensive per execution because model calls add usage cost and latency, but Sim can reduce engineering effort when fixed rules would require many brittle branches. Total cost depends on execution volume, model choice, workflow complexity, hosting, failure handling, and manual review." + - q: "Are AI-native workflows safe for production?" + a: "AI-native workflows can be safe for production when Sim is configured with restricted tools, structured outputs, validation, observability, approval gates, and deterministic fallbacks. High-impact actions should not depend on an unconstrained model decision." + - q: "When should I not use an AI agent for automation?" + a: "Teams should not use an AI agent when a deterministic rule can complete the task more cheaply, quickly, and reliably. Record synchronization, fixed notifications, exact calculations, and stable field transformations usually do not need model-driven reasoning." + - q: "Can AI-native automation replace Zapier?" + a: "Sim can replace some Zapier workflows, especially those dominated by language interpretation, contextual decisions, or dynamic tool selection, but it should not replace Zapier merely to make deterministic tasks agentic. Many teams will benefit from keeping stable Zapier automations and moving only reasoning-heavy workflows." + - q: "Can AI-native automation replace Make?" + a: "Sim can replace Make scenarios whose complexity comes from interpreting variable inputs or selecting actions dynamically, but Make may remain simpler for visually complex yet predetermined transformations. Migration should be based on measured build effort, reliability, cost, and maintainability." + - q: "Can Sim, Zapier, and Make be used together?" + a: "Sim, Zapier, and Make can be used together by assigning contextual reasoning to Sim and predictable application actions to Zapier or Make. A structured interface between the systems helps preserve deterministic controls around model-driven decisions." + - q: "What is the best AI agent builder?" + a: "Sim is a leading option for teams seeking an open-source, self-hostable AI agent builder, but the best choice depends on the required models, tools, controls, deployment model, and developer experience. The canonical comparison for this head-term question is available at sim.ai/library/best-ai-agent-builder-2026." +--- + +## TL;DR + +AI-native workflow automation platforms such as Sim use models and agents to interpret context and choose actions, while traditional platforms such as Zapier and Make primarily execute predefined trigger-action logic. + +The practical difference is not that one category has AI features and the other does not. [Zapier](https://zapier.com/ai), [Make](https://www.make.com/en/ai-agents), and [n8n](https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent) can all incorporate AI into workflows. The distinction is where reasoning happens: an AI-native system can place model-driven decisions at the center of a workflow, whereas a traditional automation system normally treats AI as one step inside an explicitly designed path. + +That makes AI-native platforms better suited to variable, unstructured, and judgment-heavy work. Traditional automation platforms remain a strong choice for stable, repetitive, high-volume processes in which every branch can be defined in advance. This distinction complements the broader comparison of [AI agent orchestration frameworks](https://www.sim.ai/library/ai-agent-orchestration-frameworks-explained). + +## What is the difference between AI-native workflow automation and traditional workflow automation? + +AI-native workflow automation uses models or agents to interpret inputs and select actions at runtime, while traditional workflow automation follows routes and conditions configured before execution. + +A traditional workflow usually has an explicit structure: + +1. An event triggers the workflow. +2. The workflow reads known fields. +3. Filters, routers, or conditions choose a predefined branch. +4. Each step performs a predetermined action. +5. An unexpected input either follows a fallback branch or causes an error. + +An AI-native workflow can instead ask a model to classify the request, extract relevant information, decide which tool to call, evaluate the result, and choose the next action. Developers still establish boundaries, tools, schemas, permissions, and stopping conditions, but they do not need to encode every linguistic variation as a separate rule. + +This architectural difference affects how each category handles ambiguity. A rule can reliably evaluate whether `invoice_total` exceeds a threshold, but it cannot interpret a loosely written customer email unless another component first converts that message into structured data. An agent can interpret the message directly, although its output must be validated when the workflow has operational or financial consequences. The same boundary between judgment and fixed logic appears in comparisons of [AI agents and RPA](https://www.sim.ai/library/ai-agents-vs-rpa). + +## How do Sim, Zapier, Make, and n8n compare at a glance? + +Sim is the AI-native option in this comparison, while [Zapier's workflow model uses triggers, actions, filters, and paths](https://help.zapier.com/hc/en-us/articles/8496180919949-Filter-and-path-rules-in-Zap-workflows), [Make provides visual scenarios with routers and filters](https://www.make.com/en/pricing), and [n8n combines low-code workflows with expressions, code nodes, and AI](https://docs.n8n.io/build/code-in-n8n). + +| Evaluation area | Sim | Zapier | Make | n8n | +|---|---|---|---|---| +| Primary design center | Model- and agent-driven workflows | Trigger-action business automation | Visual, scenario-based automation | Technical workflow automation with AI capabilities | +| Typical decision model | Models, agents, tools, and explicit control flow | Predetermined steps, filters, and paths | Predetermined modules, routers, and filters | Nodes, expressions, code, branches, and agent nodes | +| Unstructured input | Designed for model interpretation and extraction | Usually handled through an AI action added to a workflow | Usually handled through an AI module added to a scenario | Can be handled with model, agent, or code nodes | +| Runtime adaptability | Can select tools or paths from context within configured boundaries | Usually follows paths created in advance | Usually follows routes created in advance | Supports both predetermined logic and agent-driven behavior | +| Deterministic automation | Supported, but not its only design goal | Core strength | Core strength | Core strength | +| Technical extensibility | Open-source codebase and self-hosting | Managed-platform extensibility | Managed-platform extensibility | Source-available code, self-hosting, and code nodes | +| Best fit | AI agents and workflows whose inputs or paths vary | Straightforward SaaS-to-SaaS automations | Visually complex but mostly predetermined data flows | Technical teams that want granular workflow control and AI nodes | + +The table describes architectural emphasis rather than a hard product boundary. [Zapier can combine if-then automation with AI steps](https://zapier.com/blog/ai-by-zapier-guide/), [Make offers AI agents in its visual platform](https://www.make.com/en/ai-agents), and [n8n's AI Agent node can choose tools](https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent). Sim can also execute deterministic steps. Buyers should evaluate whether model-driven reasoning is the organizing principle of the workflow or merely one operation within it. + +## What are the key facts about Sim, Zapier, Make, and n8n? + +Sim, Zapier, Make, and n8n differ in licensing, self-hosting, and cloud usage accounting as well as in workflow architecture. + +- Sim is distributed under the [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), supports self-hosting without a platform license fee, and offers a hosted service with usage-based costs; as of September 2026, current hosted terms should be confirmed on the [Sim pricing page](https://www.sim.ai/pricing). +- Zapier is a proprietary managed platform, and its cloud plans primarily account for [successful automated actions as tasks](https://help.zapier.com/hc/en-us/articles/8496196837261-How-is-task-usage-measured-in-Zapier); as of September 2026, confirm definitions and allowances on [Zapier's pricing page](https://zapier.com/pricing). +- Make is a proprietary SaaS platform, and its cloud plans use credits to account for automation activity; as of September 2026, confirm credit consumption and allowances on [Make's pricing page](https://www.make.com/en/pricing). +- n8n is [self-hostable](https://docs.n8n.io/deploy/host-n8n) under its source-available [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), and its hosted plans primarily use workflow executions as the usage unit; as of September 2026, confirm terms on [n8n's pricing page](https://n8n.io/pricing/). + +Pricing structures, included usage, and plan limits can change. Teams should use the linked vendor pages for a current total-cost calculation rather than relying on historical per-unit prices. For a deeper treatment of the licensing distinction, see [Apache 2.0 vs. fair-code licensing](https://www.sim.ai/library/apache-2-0-vs-fair-code). + +## How do AI-native platforms handle unstructured input differently from Zapier and Make? + +Sim can make unstructured text, documents, and model outputs part of the workflow's decision process, while Zapier and Make generally need those inputs converted into fields before conventional routing can use them reliably. + +Consider an inbound support request. Customers may describe the same billing problem in hundreds of ways, omit account details, combine multiple issues, or include irrelevant context. + +A traditional workflow can send the message to a model for classification and then route the returned label through a set of paths. That design is effective when the category list is stable and the required action for each category is known. + +An AI-native workflow can go further by allowing an agent to: + +- determine the customer's likely intent; +- identify missing information; +- select an approved account or billing tool; +- retrieve relevant context; +- decide whether it can proceed safely; +- request clarification when confidence is low; and +- escalate cases that fall outside policy. + +The AI-native design requires fewer branches for linguistic variation, but it does not eliminate engineering. Production workflows still need structured tool inputs, output validation, permission boundaries, retries, observability, and human approval for consequential actions. + +## Can AI-native workflows adapt without being rebuilt? + +Sim workflows can absorb some variation through model instructions, schemas, context, and tool selection, while conventional [Zapier paths follow user-defined conditional rules](https://help.zapier.com/hc/en-us/articles/8496288555917-Add-branching-logic-to-Zap-workflows-with-Paths) and Make scenarios usually require explicit branch changes when a new path is introduced. + +For example, a lead-qualification agent can interpret a new way of describing company size without adding another string-matching rule. A document-processing agent can often extract the same fields from several layouts without a separate parser for every template. An operations agent can choose among approved tools according to the request rather than requiring one path per request type. + +That adaptability has limits. Teams should rebuild or revise a workflow when business policy changes, a new system needs authorization, the output schema changes, or a new action creates additional risk. AI-native does not mean self-governing, and a prompt should not be used to conceal a material policy change from review and testing. + +The useful distinction is between input variation and policy variation. Models can often handle input variation within an established policy, but humans should explicitly define and approve changes to the policy itself. + +## Are AI-native workflows less predictable than rule-based workflows? + +AI-native workflows are less deterministic when models choose outputs or actions, while rule-based workflows are more predictable because the same structured input normally follows the same configured path. + +Model variability matters most when a workflow can spend money, modify records, communicate externally, or access sensitive data. An AI-native platform should therefore be evaluated on whether it lets teams constrain the system rather than merely connect a model. + +Useful controls include: + +- typed inputs and structured outputs; +- restricted tool access; +- allowlists for permitted operations; +- confidence or validation checks; +- maximum step and retry limits; +- approval gates before consequential actions; +- test datasets for common and adversarial inputs; +- execution logs and model traces; and +- deterministic fallbacks when an agent cannot proceed safely. + +Rule-based automation also fails, but its failures are commonly caused by changed schemas, revoked credentials, API limits, missing fields, or an unmodeled branch. AI-native automation adds model-related failure modes, including unsupported conclusions, inconsistent classifications, prompt injection, and inappropriate tool selection. + +## When is Zapier better than an AI-native workflow automation platform? + +Zapier is often better when a team needs a simple, managed, deterministic automation between common business applications. + +A conventional Zapier workflow is a sensible choice when: + +- the trigger and desired action are unambiguous; +- source data is already structured; +- the path changes rarely; +- the workflow must be easy for a nontechnical operations team to own; +- a managed connector is more valuable than infrastructure control; or +- the cost and risk of model reasoning would add no value. + +Examples include copying a form submission into a CRM, notifying a channel after a known event, adding a labeled contact to a list, or synchronizing fields between systems. Adding an agent to these tasks would usually increase latency, cost, and testing requirements without improving the result. + +[Zapier can add AI steps and agents to its automations](https://zapier.com/ai). The relevant buyer question is therefore not whether Zapier supports AI, but whether a fixed sequence with selected AI actions is sufficient for the use case. + +## When is Make better than an AI-native workflow automation platform? + +Make is often better when a team wants a visual representation of a complex but mostly predetermined data flow. + +[Make scenarios support modules, routers, and filters](https://www.make.com/en/pricing). If every branch can be described in advance and the main challenge is moving or transforming structured data, a scenario-based automation can be easier to inspect than an agent choosing its route dynamically. + +[Make can incorporate AI agents](https://www.make.com/en/ai-agents), but a model call does not automatically turn the surrounding scenario into an agentic system. A workflow that classifies text with a model and then follows a fixed router remains mostly rule-based. + +## When is n8n better than Sim? + +n8n can be better than Sim when a technical team primarily needs granular node-based integration automation and wants to add code or AI capabilities to that foundation. + +n8n is an important incumbent because it combines conventional workflow constructs with [JavaScript or Python code nodes](https://docs.n8n.io/integrations/builtin/core-nodes/n8n-nodes-base.code), [self-hosting](https://docs.n8n.io/deploy/host-n8n), and [AI agent nodes](https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent). It can support both deterministic integrations and agent-like workflows, making it a stronger architectural comparison for Sim than products limited to simple trigger-action recipes. + +Sim is the stronger fit when AI agents, model interactions, and dynamic tool use are the primary design problem. n8n may be the stronger fit when the workload is predominantly integration orchestration and AI is one component among many. Teams comparing the products can also review these [n8n alternatives](https://www.sim.ai/library/n8n-alternatives). + +Licensing is also materially different. Sim uses the Apache License 2.0. n8n uses the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), which makes its source available under fair-code terms and restricts some commercial uses. + +## When should a technical team choose Sim? + +Sim is a strong choice when a technical team needs an open-source, self-hostable platform for workflows in which models or agents interpret context, select tools, and determine the next step. + +Sim is most relevant when several of these conditions apply: + +- inputs arrive as natural language or documents rather than clean fields; +- the correct path depends on context that is difficult to encode as rules; +- the workflow must choose among multiple approved tools; +- the team wants deterministic steps and agentic decisions in one system; +- self-hosting or source-code access is an architectural requirement; +- the team wants an OSI-approved open-source license; or +- engineers need to inspect and control how model-driven execution works. + +Sim is not automatically the best choice for every automation. A fixed trigger-action tool may be simpler for routine record synchronization, and a purely coded service may be more appropriate for extremely high-volume or latency-sensitive processing. + +## How should teams compare AI-native and traditional automation platforms? + +Technical teams should compare Sim, Zapier, Make, and n8n against the same representative workflows rather than choosing from feature lists alone. + +A useful evaluation should include at least three workload types: + +1. A deterministic task with structured input and a fixed output. +2. A variable task involving natural language or documents. +3. A consequential task requiring validation, approval, and recovery from failure. + +Teams should measure: + +| Criterion | What to test | +|---|---| +| Correctness | Does the workflow produce the expected result across normal and edge-case inputs? | +| Build effort | How much configuration, prompting, code, and connector work is required? | +| Change effort | What must be modified when an input format or business rule changes? | +| Observability | Can operators understand the chosen path, tool calls, errors, and model outputs? | +| Governance | Can the team restrict data, tools, credentials, actions, and model access? | +| Reliability | What happens after timeouts, malformed outputs, API errors, and partial completion? | +| Latency | Is model reasoning acceptable for the user or system waiting for a result? | +| Cost | What do platform usage, model calls, infrastructure, and operational support cost together? | +| Portability | Can the workflow be self-hosted, exported, versioned, or moved if requirements change? | + +The winning platform may differ by workload. Many organizations will use deterministic automation for predictable system events and AI-native workflows for cases that require interpretation or judgment. + +## How do you migrate from Zapier or Make to an AI-native platform? + +Teams should migrate from Zapier or Make to Sim selectively, beginning with workflows whose rule complexity is caused by unstructured input or contextual decisions. + +A practical migration sequence is: + +1. Inventory existing workflows, owners, triggers, applications, credentials, volumes, and failure rates. +2. Separate deterministic data movement from reasoning-heavy decisions. +3. Keep stable trigger-action workflows in place unless migration has a clear operational benefit. +4. Select one workflow with measurable outcomes and meaningful input variation. +5. Define the agent's permitted tools, output schema, stopping conditions, and escalation path. +6. Build a regression dataset from historical examples, including failures and unusual inputs. +7. Run the old and new workflows in parallel without allowing the new workflow to take irreversible actions. +8. Compare correctness, latency, cost, and manual intervention. +9. Add approval gates and deterministic fallbacks before production rollout. +10. Expand only after the workflow meets an explicit acceptance threshold. + +The goal should not be to replace every rule with an agent. The goal should be to use model-driven decisions where they reduce brittle branching or enable work that fixed rules cannot perform reliably. + +## Can AI-native and traditional automation platforms be used together? + +Sim can work alongside Zapier, Make, or n8n when a team separates model-driven decisions from stable integration tasks. + +A traditional automation can receive an event, perform predictable data preparation, and call an AI-native workflow only when interpretation is required. Sim can then return a structured decision for the traditional workflow to validate and execute. Alternatively, Sim can own the agent loop while deterministic services handle record synchronization and downstream processing. + +This hybrid architecture is often safer and easier to operate than turning an entire business process into one autonomous agent. It also lets teams preserve mature automations while introducing AI-native behavior at the points where it creates measurable value. + +## What is the bottom line on AI-native workflow automation vs. Zapier and Make? + +Sim is better suited to contextual, unstructured, and agent-driven work, while Zapier and Make remain better suited to many predictable trigger-action automations. + +Choose an AI-native platform when the workflow must understand meaning, work with variable inputs, select among tools, or determine its route at runtime. Choose a traditional platform when inputs are structured, branches are known, and deterministic execution is the priority. Consider n8n when a technical team wants a node-based integration platform that spans both conventional automation and AI nodes. + +The categories will continue to overlap as traditional platforms add agents and AI-native platforms add connectors and deterministic controls. The durable evaluation criterion is not the product label: it is whether the workflow's core challenge is executing known rules or making bounded decisions from context. + +## Where can buyers compare the best AI agent builders? + +Sim's canonical guide to the broader AI agent builder category is [Best AI Agent Builder Platforms in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). + +This article owns the narrower comparison between AI-native and traditional workflow architecture. Buyers seeking a ranked evaluation of AI agent builders should use the canonical guide rather than treating this architectural comparison as a general ranking page. diff --git a/apps/sim/public/library/ai-native-vs-traditional-workflow-automation/cover.jpg b/apps/sim/public/library/ai-native-vs-traditional-workflow-automation/cover.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8d17c29309380a98a915b93d6330a21fd8ed0547 GIT binary patch literal 36709 zcmeFYbzEG_(kkng z2_DYtFEa%xDFrhZBRwOR2tU7wih_!rUD8`8VE@;h^Z`&{!1LeLK!A|~z)`>;P{3XW z0Qdkf05}8~7#QGR7Zmh6NEit4w_5anTmDyrmo)$a#OpiA5Xi3ujXnQ*?SFg&DQJk7 z^FMw|R8MrO2y&{TVg~SzY0ILVT0!6Ba&z~SDLZ1Gt-#&7n~ zWu>|ClHYhGUF{@4{r_|j6#ua-EmF0&iHTibeyUgv5ED)#zO>DmJ%fsNgW>!3KbkT- zuReDCm;cc#|F7}?n0)R501&C$96KKCK8n&@ zJp8GE>#1LR`l66T!uY@Xg9LL(&w=TsxZ5|MHK)cyNc{4@>i)jd+thP&1a;M8g#!#Ph@ zfQcIQa4O$vYd63c~C4#Vn`72BJlzmDVcjBfP2{T0;d0(~22?{X4GoPW5tcQ{v@} z;mrD{n+oU@*^flTb(f< zCA{oOOhS;-tfsOF0aZn@fFYN2dO+k_ObVbhwcNGmhHT1MEes0@fq>i;>mR=l(V@Cx zFoazU`^SVwrMJLDG^O5iro8~(3!Ur>MC`Nt+A_Nhgg0JFQZJZSzap8{I*JO8Yxf zLnPtf@rKEq#VZQ^scT z#|?mXrJ|2TOWu8e>Fejbn_RgqHVPEZIU_TqfPLPr7K%Dqr% zv1SveJD~~K9eLiXsm0UGedF)V&GV2A?W0~5-osp!&p#}S1?CY(LzZYWwlRI3QmC zzagKvzJx?)t!a7nx~JnhRt8sMPPTGvnPo;Bm1sq{9j6V^)}(*-P>d4F}Zsj`F4 z&|ON~DR(wDmCdN6?R_>6LfE29p^9e0;G}dCh>(uyuB&zJmYP?weuAM)V5&kg!$G`- zw8@u8ZnP7Ocq7S|=PH_$0J?t3YRaa|%<}sNcznE=6SU}oeyy;=%(ZiYW44H6w+K37 zqK2GgBQ$GJCYHt+S@C4WdLY!m?b!sZY!dj4wss+&a5zauEFeZte z(nZmztgqUNG&;4P!f07t%{$Ixt-l)HK{I#zzZCxa=~=kmnxQ0KK0=<%Zy*&$<8zV|rW}ry4;)5n)MV zn2i|r+0h(u+VoDCj2d%JeU!CRF+LU-TJz6e{>$0^SB2ODt;BET(6L8%j{@67dJrSm zK0aUGiU1eTuu_@+fnH@3gbg7DsHI#^++fh+ec!tIA zF1tk>TOCIX=ic1e|8%J~VFpw(WI43KGJaR7BiC4M6mk-lETJh6N;wb;t3QV4OTOIy zbh%?Qn$KdGdFo-XPpnDc641?e%}iR;XqFp!#!GemvP?D4lf!aT)&!pfyxL#~Zo z5g87ckIJ1KPfz!;alS2|KoYkP|FAcLgJ8KtxdVXd^j6ot4_^l-VnSq>b`K+zGYjK6 zhnqv2Ax$4_4+`Schr8*yN54sWz+H2=#wOxLkv{Z*QEn=(iTr58G;@;osGW0)3&4u*yj8T5;U?8j5eFP|b7I zt4V(x?COF>#0VJx0PemwS2`G?$8cO9br;giQSJ>FK;+o}ECe8VcQKyFX!rUI11i=p zq+z!$@{+i}>MgX*J-+q$*R>F$%_=MlG%Bbzih)=Ps2p4dGmNT%QT zA?}f9&ytx*0o1B}a53};spCTv6OW%4KVz(}o?fvBp{3w9Rw41|VGrb|e8Z~4OH{lgML zip0Py$Nkr3cwi`$&-tk->?gI7?U})$B*W2{?jER^IUo@ySnE1_fC}238OD!db&sS< zc6-NEgQ4$rb+jfkm?$+fKL(}i=$msR&wd6+SQ%*@&lGZ#RJsRm!!AyGp_zEh*h*uk%#v2svrnYfC(?DV_rO#?W ztE4UuzFcrRCVwz8k;bCT4?CR+B}y}{Sk8FB(U;HZfmbiDinR#ujgD9fTni|R?x4Cb z#Yi+mY>Wi11!3{(I!Vs(UCw5JoCv4Y>$#imbF1U+;DSvnRxynseP#oLnZ8ViEOJ_^ z7#(rb`(f3B$5l z!irCCXYs~}a_q)z`v$iYpvA<$|IkG3CIunM?zuR2)b-CKTGzL1pd2aXG{ z@sE_0s8n?~pY{;UcPfBlH6E1$zvv(#b zB7^mfkH9deF3%y#DO?|xujr)Buwf_nlD5I{SljHD&4dT(3*Z;^h9>kv_33nWU0`jK zCQCto$KD44BJ)u>f)*;xBOZMVN&!ei(>?1b+e5R8k^mI~L@bVm$ejeGmdh7_5SWYC zQ%FKHU_WxOgjO_(HXMFU(XnC3wwk8#GLIwUY3A{lO*5uILg^WPW~t7gbq1@en9j2( z{Vgv>=keu{qw{ef2@Bx$OpJ(qz*BAb7q9*gZlgwGQCxXhrkpxk{$v9dC-gJ_rS*}1=U=AOBE%O07T`;!V`dY}6!v2{Xf z=*XYtV&wBDeS`^FrkSdm)`<^Chq91D!0^&3`v@t807vT4do}6#G~1jIgE1~?(cbYn ziJyy!bVAV&#{~_%d1Z{xL=~WcR2H?}wAKRGqZdH7y}LuL1-{Wpl(%!&Viw_M*kxM% zy&nb35IVCzupfsYAg|PA3B#gwN}m23rck_Joi%R5YGk6ac4bFi$EpEHjE0mm?A9T% zCtPFa`Zid2yQBwE*Myc)ewA_Zx`Olhq3)!SvYbuf1rX=X=U6pJqfg&1an|^{RYKV1 zpuXC&nd~5b0_YB{)$NmuL< zwtI~W4H`0-@1254!}DJNUaKw|@r!9xFMv_LEK7|=$)myBv!@ycbAp@heX9~xa0!F= z^Z3EkCS?;&d1?VFXI9i125x~fcGsu^<3D1+-d?GJ^;Qe6VOQPGSWaCnP z^}OiE_)$<}rE*qdBRQ5L}qV8VC1Njs_F+9XWyLXI+RsZr6R*82r8 zw}q;0l{4d{$!$)?V=stYe&u_Y<$e}D7{9degJVX=v8pHQ{mK=MfMZvYf!waCqrX=DJG%7A4iDSyGlWEXlddTIHQgLy{u=s7E zf#+TE$-Bo0dzc}vy#%n1d)v~_U)X$=t~LdF{(LMjFUg(sP0bSWjpD#yc)6t$nhfw34VGsll4#xcyb$l_ z(Nn|(m5e8GI+r!nM6Ncf+5+M;Y1Nt2k;xd3x4V*MZh?Vfl72hz*@R32%?LY)G!G$i z)SFK`5h(+8wUHj9Yq3@*#|`{r@;fI9_QC4uaKD~WLrwR2;USJ(#xM?>wLK(c&o1a5 zERjMzd9vDH0B(3`&<>oHr^=D9r%q}WgHG(7Jy~Q3=spVWZWqI0HwbV3Y)<0)u6lkF z6L{$uCeWF)e_2jbyEsoh&>t5kWlC&AKx{Y-+LW?jA8s#o$~LY~ClyJpaCnLH;%tK7 zHLMG)yrU;28In@c5(R$17-xtjn$u^{6w{Ub1u$+;7X8mqs6Bv=8i~myEWVFEsZ-3+ zd@JYtEE6WI$OzLh8!f%iNhI#61L3q>%PC-rqBV8x7HueXdGi)7iQa*9IC?@R@Gk%* z%@J7+G1g#)5BjzE+X}t9d_q~MdS8c8Dc;AeGJZCslg3W)(6m z_iE$gY^cog*U{tep;E`%%}M>pQWZvPz+1=ABJ#P0>=N=}#uT4qY=4`0HwKF`%rBH0 zyZCg)XIB@1j`WHe345K8zS~lAEyK!sA}uKK(&fm6q)uvt_{{+0GBH9lbXik zZr>H#R*h(?Tx4S7_1ul8RaL=amVpxxX_sNB9-i;3WImHmZr7?0W{&z^{PzEC`Mcv` zDfXIdHwb~!`1cg0)@8(DbQKRK0NAuFs-kK66w{0)i;P2SNHdtfF8N9OpjpP(ljanm zbJP7tBcY)bca$TJa%th>&{(5}4FL>>Zr$yY*Bmbb*U9^NjT8p9_~%Lis$|amuG|W! zs4v~q6iiu*bw57a5yw>A6pYV_wVi0xTbFV`pl;*$7??A8NEAm3_d^baXP7uoSGOs+7v5qE!kV&vKUve>46(Dl076R1S~J+-AfM@l;iqg%oW zctubHP1n_&LITIJ6MRCZyiqdh$}OhIOulYJf=;ah<$iu5ePlDkjw>rWC~GU@B3XgA zD?sNP`0y1J5W)G8!bjjO?cenRGl&hayy&DjB-|tlKLjE#a`D9gAR#Jrl?EWfCjr{v?F2RME*}tza zBpKF)4LJwMF%T5iJ6!ZM*ca4@hnukD@}z{eDEao)r9@9x@G zR7L9*bm53pekKW2rSkM>eC_Yq;wqM9mlNoVy1d82(+RPj%e|HlM8_5bA{o+|bcIfi z&ht~RVLLNUp}aNcYQt$eSoqQW;l5YVIcBT@@226BP%Lje?ITX>nEMJBXrR8DCm3># zv8-bDDj86f3vsA2evowYzYD*sE*(z-M#f@+BidJ3iK7fByeS)-$-uOWnE~adTj(Wv zOO>z^H3Ux1=U<$Byf1*1e-W`2nfP;;lT8$;YmUGGI;zS`l2|F=kY4i>asAj;N#10r z#dwg(*vxC6EfLa9rC^*8 zZZJ*LWbK0&4f#3fhU9Qfcm&7;6fQUW?oC$)6%ubtnw?JpTWC(xkvfIB7kw?;wHU8`2 z6KMPguwiy@v+sei?gGTmxX-oC67?M`_~p z5C?&nmiHIuM`zy!=g94OX380X*pzH!*V%Ok6+x44eHI%C)4FKvAU|_ zYD33S5}0$Ke+KENo&yAFdpJ=u3PTJu+;qhv+)axn-lY^Vq#%g;#Y$pOyTE&p@TfO~ z$*?xWhu2)f(o90s$4}llotstYm$p@Ynvc9*)t-q=w2>Pw;l)CTby{OZw{kKt!tzsa z%s6RVoiaEGXJN;oJQ2xwK%pcm?+bDbCx8v^Kmo|Po~YQHI_N%j{b#YCLoDCh@hyPp zZG~g`w(!HW%mCIL3Q;{o<-vwrY!jxcT)e1Xwe*eXhRgmrt>-Ou?>+PEQPp(hbKv=5 z-W#DJCD}*pag94)stn{BmsuHN5`du@C(gsF8+vaYrR>tmbEVgO|J-Jk>+K+1K43~^ z${iIz)G^{$-wAzI@)i4fd^W}LBm4|=(dd_0KGea)MGf{PpaFD07it%_S&jnlnA4a{ zDzc7K;yj{?ZT$ezgzohFZsuSmU#2b#e+3~XW0jNWLy0G;NFYd=AVm9hE|)`Z(t0zu z3I18K=|^`<#?ZrZ?oxG?_$9Ke5vGW+I_Cv&J0(#(u zu%h3M-9pbI57&O}pA8XW??Vp$t)h^yOoaWMG9E@oMXd^ZwIxg!)ISx8&N6IZR1NAc zQm3_|*KSC{NR7;YXMf`%zSPJ#zYig|heB#c9!^(o8nKTfLNr4K@cNk}aaABzJI~T} zPyGYfmsVaz;36Y%sBKiO5h%hR3 zh&7t7R>G|~SyV&g0bGfCV?YZZ%wV5pb&pRB>+t9k9^b07TGNy?a^VElG#3R%g4Q|i zjJfN&5sZ0+0oM?7XNUsi*EF_IaQO9?zQUNbvBsvj*N&y6D{Ar!Dg6C9gAl}@JG?BQ z)6G)fZGLANvEnk7(`kQci_aSt zNv_cHyo2uV{E)i6t-1Hv0=tuOhN1l2DuTX}rvJsyJxn25ToH7gXRG}TXqDf)s6!^< z=7>w)8QsPfiOD!h#Q}Ux;fxZ$6C6mDsn#R)q3UXvMwG@%s;)m*Dq5Q;``fVB^9VOR z9O9NUv_?H)%gSp<{@h~SPt=?#m|`c2*i{>{drB-*(K42HHJs#QuCVN!;jb$|)HR_0 z5i}4PY{kw?4J-Au`w9u}Z+A6iT&4lf>Jw`#<11?*6$WUseuES*!!l1fHR06%Sc&NJ zI|};mh`KjZ$B2bJopIcM=v?6_1&SMTB{66Crn$Q3S5SPMskw+Q;8w_^3RH_+7wATJ zolWx+WU!~3xp)E8u^&x0)f`CWZj>S@asY*k3{k`v8^2|#ScU%O$FGm4PMy#c>n#6t z_j-ID5&e13U6a#@gB;DK$~ssFbXh#V(MCI1rLq_iG+O)~hk+u8VV2tArqyq>WE72| ztCC|i;6`T$Oj6{(QT9fpV{v3dM1WC1$ZW7!D9M9@ElOq1dar|j+{>4UN}kggFfePm zd`(`NIH!G-7?txGMquZHgKvh{yklcy(vAr?P#VU@T1Y`GDz2>-&qZ5_hPKzhBFdyT z(OLFvwY{jjbYr)=tg|T`%)jZEKKCM*Iso>*N0_qe@B+LgQO8lhml!X>p4bD%yDUsR=f!iuTnTO722BOQb|SeLaZa=u7oCw4iPi2+6nniT7C3>s%OWLc;`yQ{>EX?|Edo{MEvF zBb&497Rk-?fu4t_NwwM3sG&_?5v-WdWZI*QCVwhw8f{7r1F zJJ<3(RtyJ^ZgGj34iSa0VhR*KyAS&_jQwa4n+ysF5_KIP>!0j!G(G2Psor$r{+rkN zsy#1dv;HxBa$UwQLfbStQF7r9_R@c%q{W_s%B{+L+L<>61XMAiQC9(c=Wi|+cl+%+ z_s)0wpN+{f)7V2jrhNPXcC^%2>$M>#pY05?3eggu_Qd{VrV4f?dn+I~tTa?=j^26;L#oKN21~!T;L&z7 za{2LR(WNILu%+tM%}a~S5R81-xQCo_n@dP3K_6S`>YG^B8|IblbG`9t6~lQ`;G-&^ z01ZyzmsHS%7KB8fVo~X~HSlB0E+8J?A>B}Dttz%i$6Xb}!yBKmxabD^a|X(3rq+MW z_|^N4*j-BKk3G$CGj<%ExmNji4lP2ni~lRiOP>c z`Lo&9`AYa`q>>`r)N;MAxon2yTZ&B5Jv2MEz&88_GepD4*zj!`mTzy1{`?1&knVsd zm_)H}ebvcL2g5yiTW(E@Xv~;qN=mku)$d>;+el5!Zxu1JNkKj3>EKP((RJT+Qz&Kn z_afYqJE#xayu7N2&n#LS!7R}SR3$_=I%>y?2}u-w-u;-r)LAF8caPICD;$)@XV>Ab zT@#NvOEHji>6?$dYFQ>&`DMVE^^a;>L;dj@EvG5zXe{hoJ2;2us>bdFjTlb!q0J+& zD99UoZ`ZqubY{v)cYB-?pM|c*+=@9B8ZKs^VahdCa!Eg!$#M=A`VcO$t7?=_I&LP~ zw6PQME;&Rns5?z}1oE(*q|R|%A)b$!{Gu-OcOa4NQ-oU5LTj9CR+z<5s>quH3*lX> zYVY>0f3s6$(~C#o_Jg;B`iuS{YW-sSnp))QA@rCw5KDa*Lm=WoEJ<3}nqt%1XpuTE z+;`FxJBc+SJD-+jRE=MU2sYG^Q=z&9pH?U^4SZB?3Su7RATOkOdATi$Ma+vYxcmvr zq8v-Xi{w#IbF6%jT~+7379R5jx2u~MdSS*~v^oL7(xxu11zf7Es^M?5N@;%cO^2KI$@?UX(CYc{A~m1yERB*JJ&?3zk?6-YA;G2nC#xkGJ!O)gWv`}^0J z8M!N$F=zwMC3$emi>c>oy`_3{Xwvd^Kq$S6zkX|tKb4}+MR=N(W@W((S92_snn{W_ zC=9h7HN61NNPP%}S*CLkan$M~1BRFb3}IxsGK}$GQ~#EJecC0?M?{kQq{%I%4x{#t zc(=dkO~Dlb$jGfq7+#y0C0wR~&m_v((jqU>A-xEwU1${i(8 zq(eBH>f30Ok?S4PCc};)!x)PyS{)Vs?m_GwEf|Pf)eIo3K1tn35})k1dx(8Tlh?jz zK|;+@PjfIXt?2Z_TJ6_`fbUT`P`S*l3E!frs8@|3*PQsB+7ID;>(7)>676v$IQSBg z_?9IBE9(f2G|aP$-cI)amRflUe>~;V51p{qbrV6s!7&51$G}Qj1v+)Pr}lsLvG1W~ z56>dZBrWS=38 zJtcU8C(M#i5lWi3wq9d}-TVR&ZQA`dw*K>@X`w)En`Q^n0vb^hfqCAd~F7ePl!2*JVNRf z;x4ID*>w5lmukX?W-4ZMuC2|uii+smpe-$W8%jpPKdu|l$p@BqV>i*y^lN$XH2F5G z`{?4EX;5FclPvzB#=4=TUbEIX$2XRy2B9hWn06^L$`Ij#lR)`|lXlxrBK7#ujU*Iq z0e@)X=FdfROdwb2)1vzv#9^e^)<)K^4rt?w^>sgfu4VvVbfhySx%vYL1l(E59 zqkXjbMEv#UL!q`dy2Aw(ZmXtudQ`VoJt1^$p@3Q-zno30iDhq;ZAnrEW3j4N&b$67 z36rqC3>|7sa68)?2O%SP++wrjlM%5jePb7)qvN`|DFnEuzRT9*-^oc}ZiNY*q*otu z{mVJx|Io?>*0$s}JzIbJnY;MW-W3m}de&OEtdvZ*Yu#7M?fNsg1$yhPO83MVPLOf- z0-Fya4Fx|cQ82ekGgq(M#`WX=I7tm`0IOaG)2!-z`dGe^gv_FXG;~4BM2Yhje z16p8A#Lnm{TrTyh&xT*92{)0Y(C2fQy;ob6>O9*)|1GxVJF)5P1TF;>N~5sg=Vk=1R>IwXWsQDV z*8!46L;@C6y1vsmx`W{gU08guJTv-rG9{X0hp66?SP`nMZ4^a~k;MKE*nRu8RZ?VL z=1N&0?{`|Q&3iS6aeIX{mqBaJg_=?+rifztaQ)6vmI%NSer*xw+$bvsOAx{wfwsDv zy6Y=G_FI}%5}oOy*71Fnt}Tantpp(}f*xqd5tnyk`S|-Ww`7*mlJ;{0lBp^EkTEc} z_W-XVA1j5pIYU$!DO$}9PY==W`JKtG@O&1Ivh^NYg|ZYEqvC31*-=vZ zyu;_w${Yz&ss_!UdEXkB>BPU19`#^0euieWyTc>=6v|$mg0TR$+`lY|@GV&vJza_p zH29D#{Mnc1TtJ7LIF_brK}%!RDc>D5*TDkcTCTXBND2$B>eN@VDQS^G+EiZIw7PBG zH2t=x{y?oew;(3aU^Q1dRPb=N>4{eoD+(+A`Rm4-Y&J|A(IbIGUJ7_nmXq*=dvYtA z2PPi{wOt_fO{Q)>U}O+Jh<249Vh9IqwI47+V6BIR&t)?`}FP<{g{Ifq!=}{x0*!=0u>e)HL^c1Cr77 zMMU{__VWFKa7;RM0rUaKsmf2>BJ-1)+;NbtQ-9FCh_1X|Q9TQWB4ESsXSQpWtU=Sd8x`~tF*?-PSs*st><#tP8y%nlRC&tE@C`)WjD$`MN?OBQ>MEVJMI1k%>K1 zktJb#d{f>_Eud%Ras(9na)!$EuSd<;L(D(?1`B4Za4*bb26ROlUI0-Hf0@HXK88p< zCqLUZA2^*}5$y9?2kz8?aT#qj^P}#f=dO;AV#sshF=y1}>D>nP;ZUIwpKzUl+U$L0 zw_o)gBsUXMr|?UEG^%>50QEIAVDDDLZtgLH03?$ zD?@(f4AU%xsGlh{&3MN~FIzYfIYA91|~180}rtS4W?Lv5_D3g|5` z{RhS8%dqE|Ikb)~Dz{38Vrk?VG1i$J$KnL31{U%#FlRX83uhoxxYSu&`qb_Yr}{AxfOGP7k!Vr5n(Y?dZExVvNiZ?b~9dKX)! z#V#WwQ>S=*hp`<0g;f%`xNO=yrMHov9Jw2@YVE^X)<=dcg7~&6v5Shh7kAQ$Du&*o zzONR&Hs|{&=3Sj%#SyZFW+GUQoS(HeGqy034qAp>I2V|5wTEP0?d92j>B*Du4~vrM zD$H2mK+c7F3&p+^DI1r@@u7MMDu?evxqzldmpZv@srjF3c0%Ib(keXc|-9R)Tcffv4JpBRS2cx}&hePD5@x1ggsQ@T(LN>hU*bSy9;Dvp$OmGMUCkI67_-)xM`c9mr0^U6wtW&Cman=%%% zX>fo-s3g7FuoK$-->uwpT;@G&e`VnWn>d+}yi;gI^91^QT^6! zsf<0Mad9eE!Z`8JN5>eYmd_D=i|S@p5-hajtI{9~I@i@tm3k6}7_ z=omw3^W&?vZIxz1U=qmRe?1xh9L6(Z->F7qB8z=H`>IL($5#5^9os@?>65T02Tr3z zbyr12m6|5cE1<8-Og0peu5&{WxNrP8?CV4>x2LY9mG@wek8rjd1!LNaQ>nV|3(au; ztG~X$oY|lFL2cZ7vPuxt_)df=3rI&vhQy8HLF3rF1a8XCrt`~Y3!{9cS??N$Ym&$^ncd63P(YrvJ@{v) zeO$BHQ}F|u*zal(gr2CGl>Cwe@e!$eh}2?_f9i&RJV#W2a$OF!zlrBwX93Efl}Iy> zWirPm4T+(ZwuOE$i2REcr{=3HRMf#mltvBtCPw+8VhZur)G^De#D-P;3HA73q7B6` zo&t+2Tvg>%YHjK26(}`z*hw0OcExnF0XD{T;&5z^5*c&`Tvy?u$_kAEN|0zd%4_Sj zY1Lh)R(Z${!s0kO4HrR{RZaI6r0M3ZO$h^!(!37r-vd1|9ge zQO=5AKcg6Djh zwZ#)SG#I?m8EzWgsd<-1Pw;>~iaLib@^PaTli-6+s-|Q7nvgiIh#-`Jz2AYZS6hv7 z#up5N1GKG=Lhadq{Y$W2)^=w1RWkHsQ;xJ!w87*wly`WJ_a5~)oC>4cP9GOMK^p*P z=F=OT=(7}KM0@18-P(C2#klyU`qefw_a(y0|M@9iY%o=fki$1YpsJ4S+QQ3Pp+s@9cJXK0uXwsOg2xG}Q{@+791F zyQ;xj4Qlh(>t)n4K(@CM+TSj+W3Cz-=TM;zq2nlvS^eXG-ZbZxJ5qwN@hgCGMrt9>QF z@CzMR;_B3v)mi7*`uloC0_Jj4ca>W*)@jcgk9V5FzvZddQ~HF+wl?#r91(*Mp7n(Xa8qJ{c&)Q+YS z@txbIo!6$FWBA>|0~%l;71p0Nyhr*0sGPB<8C&B5bT{Hro0cy7!c$+!iF<7@O*O(gtP3_CqAV-ZoW-;8(#MRNF=c|hBx-WFHk7_y^*osF4=pR7HGy6;yp zWlt<6YIU0@%QeH?mn7Px)kB8C7Y7X59yYqLz*5J4)DVXR?6U$_6m04pei} zFVbIkB9#9`hAEA*IWJ5(ND(LLr9hm?`pL&7qt#5QxFlT0aKQkSrBDg~t5wfq0ds<| zF1Nr3n!7yL*7Idk>v;oTsZZT=ZBtIZ3G%rF#9JvRkt~J4+a5rOg)`{_Qc(hrA*@we z3OC$EkJy++YJ{v{>}8o^AZ?-dZL)W=cA4dGMa~73VHAPEL6xpb?TN@7%bflAVY1mjKLM9>wB|n!@t6YB+Kyb++*No#{X3Ocx%Go3?K)8sn zIJFf~9yvZ$<%R3esEsKdY9B$NG`C@hxgPNbhMqlrvvG5M0Wi<&XIZ(;sm~q_pa{8v zHnpslr!O^)$_yZ5&*9vAj!7mdD%9>vP4nCKl!EMnl8uVKSegEyW7P&DF=ma?GL{mLc%g6UpeArljgo)kA>gu62uKk zm3<3BN9_95?a=xChSD)2mjR4#1x9IF7(NH3>O(MnwS#rQdXeQ;>!(t0N-kAcS*zwX znuKyK*>AtI`SjwDXi*w6$-0*NeG$6+`q(4*$fl9Yg5%oV^QM{HQ}+H5_@z%hvpgRY zH&&<)`CU!NpHzW$=Or(%wx-MrM6 z=JNPHm!{+?sl$Ll&U#A~5;=+A!_1+n!o+d2Tv^7^Qb~x2=383z z0oGE{9b&0FaEW>UlSGn5;xpD|_PPKfew zh7~wsHdMo((%fl=UvokY)WrB2i)0(z;~#|7`Oa!(B#-+o zzYmmr0p{M0NQyFLzZ_WtwX^L*{hBIke=^^D!|xXUGe73)-teIPwhGj^lH^OGXHgLE zJV;+@jBb*+y=4<-W6)K8pfOS?+LLSAT|2=Vg5my?j(B~cfrLAj%EQrrc`;pszEhzY z-eWVjt&LffqTc3tW#|<0w}GQC>PQMsipuw^%R2qna~X>wUP05Po1;M_SA=O%)U$uz z$MWZ$UEYM5j!n1X^7o?97}?hlu@?O0yH{+(J{t8lke^d7I&pyDrz!_~y+y&-$^OWEm%OkI{z8KS2hc zETrVAC{4+5_$zg>-)n8>@VIUKgR!X|Htkl{{=y7@dEOq{56kw@M!@9(FuL2XtnhgF zd8=!Lm31t3Lq~Ay$dWBJE%qbK`f5!Xwv+a{#gv4Wir%}!JQxBwtMyvWKL2ZC>Rdo`rb2{qHpotT##yT!h8M+pm)8-nn1-=0CoqiP4 zDCiLU^#aprxOH6(mZIZd((Q_hOAhW#MCG zHO@w&)~FR7f&2FghP@xtV+XJ={FdrMAOA*pwWlo`s6W0TcW(h?!7f3dRNttku?7##k5H)`@EogfV44N&^kP_9AO=twn-;QA ztz_ar(vgu5DpC(Zi|YkoeEe92QiWW_8T|<;n|?YczyA95>+9>%le$^1LGYs1>-ZZu zTR{N^p;8rZpgs^mpE3H5Xl{6Y#6 zS66Lel4VZYE(t-arp&5U6>?5zSP=^4{~upp9TtVxtvjT&)KEhXLpRbOJ#@!7q;w4} zUD7=?2q+RmGqj|jbfbbah)7F=5-Q>yf8Te``JQ|3bN}FZV3^o@t#_}z*LvTz2XR*s z9{v3N`;@jFj4jeXWF`pK7(suE%2pL9TJg}$4pk5+SE~Wf>P>md&2bGfRd<;lxOd=#dl<;rS|-4&X&`(r0B9kKk%?e(d0{TEHxt-|T*DJtj~-?m)d zg+FWjcGUZ{7-Rml!f4T<=D|ycDfis=c@(Mvo>Og2KDh1X@zKCc;jv(4bDjm%vQ2ci z?;hpdfbQPja1E?xBqBRC2YUqwvyqvNEAPwf@(_P)S=(8}#`JBfL_RGoY16neS*;dm zIPO}_!9NL3(NLKaXY!b%ciliA1^G1VN@f3-sqDqz~7c1@C~iH5-2BycapEN<8^1joto(;tAX7JhIG<77CD z^Ce4NRl>o6Mg-dA5UkN?pK0<5c3WIMs@5P;@>;UrEf}zFkAU$?E3W1oGHtocBq!Mf zyfG6b3LwS_@x6k)&UBv&D9n`y=krGSYBmAC>WKe>;8$vrA34|zZi?BLBz2)TBnNug z5&~qYrQQzy{>`6Ask2eh7;2H%+*$k;ZF#X-IK;;109qMTSK-=gH#?`@x+o*Esjj@x z`t?4kQHx!EH?jIA+O?2TM_?@c?#u<=eZKgFR29Z)Vl2{KmqGdj^v-3OGH5w%gO18@ zVg@^io1DtFw|>$Z`HXuNA(>afJGk&TYw!ut6UJ=yLMP@|VFged4o8pR{J80r2NWTL znpD0{pirci)ngOEksXUd{a-f;BT{qH?2NnUZ~70eTSZXMf&3c})5_^JC-_RUzl~Cq z@ZNW8(`p|3=A8&%67QpVljN*U5&N6ofPiUZE~}eXD42kAuaoF2@(F0lkCk#PYS+_f z^wvcYMH!b*7Ef?2*ncsVrID z9GTdfFR0s25xxtYfi|h_p%#z#%5TBuI|dXZwCFX^Zo94hutfFz+%XIgqd~!?JR;Jp zHPc;wZ9O&6ZErd9jfru12Gfuj{l@Lp7cxOLiWsB|1#ebTJ`sL zL_g9;?|y5)m$#*muia@|S_rNx__dt>;eM(Q`j{f-;?ZTN9_hP<#O+@a#+%QZI`0OW z5)*PSTUml<1ZzxU zBDtd-0l$KGpR%&7z&B4q>=(? zt$tn3A0Tv+;D4Ir z?;2L6ZX7cvb}9C)&EIQC?58>zTB0Vy5+*|kDr_LUad|1#zho8tc?&+q^lsP;nBM0y z1Z767%Et9}SXZSfO61rq+_;SNrApLklqBauELKH@ZLcDn%g_3wni*LkFkT+T)yzSQ zH}l^f2ed7+jCC{eBFmp+S?3HsZc#)~QP=e{N+#?v8Kwb$SI%+!2v?&&b){bcTcoi( z=VkH@@icHLCdkqXWe!Dx6L#M-+C0ZTqWJ@07*RY4;2lFpU^izB>GhpZ7moI`7bQ0c zo!!3I{1W4vM$RzO&$E_Bw?C``qVQ%KcpkbRP4Z5O=@JN^ap} ziJVDJW*f5__`LOZChdyc(FC3BI?0I=t>h)4)AncE**3Vfnlc<33s+M|)M2`IrJ0ev znS-FvS9Yby#7?p=y1kc4$vUrnr4Wx_t4R=~?Rk@iXp>9@kdb5iZQ7+~V#bUydVV>p zWvutY4UCt&w}W`gs626%iB9z8CST=KhJCfv2s(0irCbCE3(ER{kfyI_w@Q=38)YtR z>pM~&WiF;fOVo!ztfee)@(>-5eW^dggb7{F1XIJd{3W9A4YZ$F#S_y^*Hh`j3q_jP zk+rkj;Hoqz{tyZu@avU*)4_zhX}JMZ*wKc7b&cFLx;9#qvk%KUV=&z94*>sQ_J>}& zrtSPkLZX&ufM{}TzS&MMH1MmXIiI(hBWVr*WQWG$FEAA9 zEug8hxTJ`Mh8AcvG%qz7l0Zl;Yc4kX0^~B;u3TB(DGs5!_uQO2rQ`Hbd*ugz+fCCu zr!5AV$`s)%6w1*+Q#P``W#aw082=hw;8k&|A31;)GMAb!4grv+_SVBS9ouJ@7H^2bH4N#E-37LJP4o*(~ZHDzoBpEpx6w z<}>+o{0?aSa1!9h*zK~BjZ`vX-{0dEi`L>(B9g%(#yoHD5%kD@kQ@_ z0#`brOuEaO>o;{UzVVVDsjq5(czxX?nDHqtY~1eAO|pEeSpCTy_p(C-;oD#0r1*yv z6x7;ZvggEN0yC=t%CD#KMwH`YCi+B!_ z>9FQ1zBZld#^e|yA@4>t-Dm9+5%`6ca_2>%rU|w{$6-~$g{8xH)ZJCd)5q%S#wEF- zAC_iXz%=$tWxKtu4$huelBj^FLiUpd`9Za55%{AY?Z2br<|gQW=WMZTd@_H`GZU8+ zEl8lO1dM&Y&oH3-om6eiBUG4m5wif0X<&ndWrpdHwCksi&}mjxC{e9A0FTwdzr#g# z)TPm!hyH%Z^G9?&0wlfd-3*synHkqLZNgo#_Y^L4brqitZR$apva~D3_%({OB#&4}Cv+Hu6mS(0=udY> z7;N@LzHRDWiHoe5=RNZ}7MwtE@`(Nc$Q)WRiTN+seIvpjt3Y;Og$&xuM?XD=yE}i- z|6R*7pZOpgTusy4Mc7c#!*}IAT-;`2Je%C32Z`*Y;;l%>XUGo~AplW@PFrC_s%(<9WLO>uFUqWX#pb)M*b z7g=aXaklq%fa4U@syt6WMWnvdo!eA)ec|5LBj*4>K(>}t-KZzb6DC$tgG&zW$B!#R zXwt3KkLyNYD?Vy=Xo%}^O&-AthIq`l-06wkt z-^%gjN|b@^t+${$?Hw7J7tBgkHC07*brEb9QkrsjOig3>Ma4N%*U3xH8Oe;6s&04Z z!(wYe)7jk-0>o&7*<$WeM^W#TPh1Rz$B;1oH?ux+UMePrv5#H9AJMg4h`Ud1ts6vD z?Y=oiG$>>ua={%RfQN*0FE_g-Nci6vqg;}(PTTW(f*dzS+uCa89k^jTDsB%ill5!O z3z$Xw>ob))D_cjt%$@Lh^3CUoM;u-}(q;0`?-@vZJu)HGlLF(q_x{EN5(p`880q;0n!pI^ImnJl3YB1?F7^k6O%kr#ERpW!sB3cSix zN8)li*0#*5U2Xlc8lbwrulXNjDIc9@y<9mopF7uRkeypE{sWjMpju)VspwV>az4?z zOr^AVE0{2l0quYHbXXqrwfI~={--nMgm|__W0|U76V4tcv}=w~TXK5#nd~jmg7eUB zcIS&}6&RRY|J4g~uh|I0A5x!lUQ))*f~G!?w;neZ{u-{2p`uXGa9t5jd0IK&3bE$w zra7-K!~5on8CR*`Slsgr08g?u)mmVh%42S`#av4dhX2IKqla?Q?Wq=@_Z!mCchrKq z#N~-&pQZ%NU;Yzaj6=@6FGAH*uE<@3ZqUmLKm8jo4OKAm>@4t@m$U@qe^- zx(hFMKJ%={SmVGu82@GE%YcUZn>Zd(rH}Jg}5^2PTB91rqc5@7_m=`n-OS;n)nCo|Lau8^G*yfU>**|6tv$7%Q#}lSC1Dx{u5mU&$bQ>Bi zqT(yHp(A&lk3aLW5@{&ZC3blJef;)U#EP7bs7%$jwtb(fu$m(d;$-}>J+}1(^TGmS zD#}>Pku@+fHefC}v1})QiR{7^6()ypde=cIZc8E&_S};WS?)^fs7E3D(Mkk=$Zc(4 zB)%HlbX8tDGB`na^*Grz&0_u2EAg>SjuTxcWNfqOV;^?I3Lc23X#zK&5%*AS6u-mL zns3^mF~1ASRl#Q6g&ggk98BR)WFgNpm~P`{zHc$o4Ltvd*>csWCkoasS{d|bbQW^% zXLJ4mgiHPw+vb0sb0MFGfdeMH43V+763eNAR8zd$GY@Oh_f!B&Rg#ojl6#VBF|Eaj z7<<=N{&4kTbmz!9G5NsX8EbYgVo>!(*@PtV`W(M~oJy#Ld0h0AnJln#QZ=v8(I^~+ z{?JkziNB!`!vp!U10Z$Cs_bP+`dL}QI~f~G4ozgkf5AV?J;2IA-t;B_%|8>9&?FeA zBpn}zE5d&ho+mq7mPfnRSBe&E71=T6+C=c#;dgE^424$(&)lJPP)(WDi>PJ=leu=s zL|RZm=SW6Rrsza6-Z;DAFH##H;+Ig%5eb0h%5GdHH7~M{`2Lhm|G0Cc8|$f7(w1$T zl~!>S`TS(tEjC^_W_i<4r6Nc$-|1W zuf6`s{cnuh)$j~Lr-icuf0wuay65ylZkCWmE{?pZ08~YgmVqMA5ZTE=Lv8XefsqWSdBaFJP`BA5zs65eDG1JjczVctA;^?F%FGQ_Qj| zlrXIKQ@#)$E@h)`ZdXGy?#INv7IO#mc&PQGP2=Znn-+k}_~QmNDk@a0%Nb5y{?!C@ z9Gl4B&JSPL&D`YLgh(i0^T4GBOKesWC8`-*IDNNSU$aZJ5TM0MI~*jdi$HNtoNz%5 z{dyBl5Ydq#hNLY|`gf0vs#J9@2CmqYAgD#uIjtR_#~r`>(hm8k2U(>;a841^KYnA; z(&Uxno~h;_e{QyRA=qlz4!Zv;k0t<=>x!+;uw%+n5M8vT31{bbSF+Yh%F^6%;vE>m z+f4ogfN>fwY0M#JmWpfLXfQ6jdYpx(ny{N@$8h-xRRt#DuJ-sqKsRvbW8_eVlP^J1$E;{mEH;`S%IUIggytg@ ztF0f(WQT}eFqy^|F$1I|{dnmmEiaM>ropLTY(rw=J-5M0W>4(p5B0~-)rbO&McDEr zSm*RX-tT<7bHxf!K8-=gKq@=d?V#bYs~lsQQ!BfshU$PxR^I5eVCxYxjb{5hG1d1F z(M(6&7maY9SrP6)ArNXMYzy!+G-)4feYp?P>j}mSd7?m}D zqXT(Psrz#ydHXOKEAS-;F8-3IBAS!lv`B?nqC{7*#lo;qdq;i@NJGCEV8pA36NM9} zCdn7RdJ5NAkXOq)8eJo3vpu`u;s%tKsR-7vdzq1?Ov)0YdHq*InUxG|QPxU!BTX3^ z#b@2(^3!_AnvAPC3aMPn7uWN7Iegme(l|DpO&?^~?1-&z?L&|hsj2PStERR#4(S{P zjZX@2ObkmkTL912?h7XV0dN=S=C6-cq6Vth$o~MIz{U6`&ekYS+;Sa|?ZORE&VOtI zb2sh>jbXj=X%!eNAGmgqIIoVSi#dRQV)-WgK?Yok*BBw(s}>E-;5USZzG+yAxzFH> zP0Y|g@^!+lyXhv2REt^SCZ;n@#ftdGW#4$9Vgh9VGE3V5#w3evi(0K7wL#;3Uh^>H zXACA{--1B#SGr(X6f0Jvr%0O$1Ko#e!U*G4gOOc>lD#+R-ud__JrEw`Fz@HGG^r9L z!CDhf2=EwO$&!GoUjSp4@$Vgx;8UDCq=e;f{C-oJz!fxpX-2o*} zcj&btE`>Mrzdy#-`ygwgdg_5C4%|oDN)0BV?_F)HM6TS0^~2&s*S6H3qR#3x+@FDu zop+e=(bRU#o07qi7m?8;P5b1y?Q^dK2`9pk%BCVulnFl(%x2u43k{1Ql29 z!xEP@lTvo20DHI?@z^)Z>=WvICr~oWksHhS(g7t-p+dv~D<*Zjlf)TMWaq`D1^=@6 zWH6RJb@dyZD?2V(!$0CtaP<;n$RW|l}Obdc_#=~ z6nrxyE|Pc8y#qW|-}%vpt4B}rv!kkmzZ!Q80@FfjyGJ$6HK|;Ye?dIgqA5aagUp3h z1SEEx&9; z*I~K8+p`n=nN5B=@guTEr_;&3gFQ9jiE%^><$xq(R9c1GzhV(ErpWu8G9xR2({M6k zC3Bv|(y+9dJ(M9B2LI5c=036TPxf$?>`Ei8{P08M+Lu=hSujd9_*pEKcPOBS(~z?C zZyES;khnY~RgY;SY^rkN7rKrA$RVTV3%y?d5}s_WOcw_M=`igM#u%+7HaIQ0Tg1NH zRbaaNEO$o6uY%yy1qtZmE;s4iu1%E?_w&wa>b!Q_JLtnH#7oNHo;u5$o`Ix$K~)-T zwtj+yPM|(CJYTC?oYTVbx^UmVj?}E^-H`eB*foJ(JCwk$YIBVjb%sQ>4>Z}9Qs6&2 zEj%3Qy_9EFcjy9=A^8v&ucZ=-GyEiq#7;ZOR?4!XPAP03>ZRt8!Eq{I8I92?T_Iw| zXMFO*L=t&xP@Ri2g!}QWj|C81oLPf)g;c_2?|cp8s2C3`>%)rZw=%nh`n$@BUH&cp zD!&BIXAdpP3*n%iQE~N&$9d##2b_{T>T|(9!wB4h$A9my-c<^R<}WRZgbl>?=VCGc z#<5LIN1HD#b(&n$GDt&G(<`i;6qrby3j#)&Uo(G%OWA?eo++#}Ls6a5*l~|rx@}SD z@h!_AVWaB*hEQ7d29((iM||d<&4ON3P1D8wR>PULvx(8(yq|ap+Zn~lF(?~5Jb#y4 zv2WmuWJad~*7->L;GK!4|HYy(2IEjY1qDM?j?3gmp&HsD)B)L0q)!5ST-13 zi)Q}ySloh+s@wXwhabO~pU8*BLbfebF8G3oVX9glM{y_h*53S3A=Zd?12xWzELDPt z+c9(P@Oni2)Nlq(YBPbd z<2s@N;u>02NI;2S`};ee1LqyiK6s+B?gfb{;*hdlXWEGQ@qDo|lhb{&c_t$>*mUwF z-;ll}b1rQ}y8pvk@Qx`?Gnt)>dy(a}`lvi`>alP`ft+91;<|wUnB11SypV81D+^M@ z%o43Jn9(MDpKn!UdxsMhrgHJj?p{BFkVJKwuA9{eQOO`P^-}!(vnkALKh^se4{-Z9 zM(O@C$X zm>h9}z@tjdS?rWouVMsnUTU^p^Ah3D zZ5#30N(gWBdj%p9SEQc`U3zNea?R1jFV}-@)hS?ONPpQ0x{D!% zYBDnoH?bNh{erKiX0fz1yQ8BAg0I?wXC1R7c%P0PiDT;}^2cVO@ed3#Vtl9rCE0Ud zVP>I{#K|J?%MhuS1u1b-U&nl2Qxavl^a))v=Y|nYme{Kai5$i^2exRlnYOyLmG5R4 za_QNlIN@an=%s-tb(~|a?@)qU<+l|?{u^5>(V+bwjXmav8NnlIVb->oHO62xtEO6u zdM02Yr$>j=bj9%9q2RblLqHLt>6^*fB&+vLgKn7By?o`8B31KwqV92S^6ZlAS;`8d zb<_DYx!t6%Ny;2C4dwts`U4A7a%gL4y@6u=P=)kZySPPXuZ_Af6@*gheyTduo@F>!ix8 zal^&d&bG=!CiE=Cj2hF0Urq?#E~_NvekdCC089B#?5~pAWQmKk(vZMEA^r^z@(yYe z<+SN^RndpBnUb~h`>n<8R@T%%irN4Y+}ozqc|&^lsE<|GLNSZz8_}W2&DVYVzvo?b z%r?_d5GM9-$nzv6?_Gp-qsEt4-(=X?SZUEM&v}wsBYpn>F1+zbKI&{4rF#PmUZHie zNT=3{NBUS($lWxBho0wJh=xWrYD^q<8L-_vUm(>nBGwFd|5UypL|OMb*(Hk|X+ZTV zj;5TCH-*2ZCbChR{B^3#?Hv719w)LAL4!4%W;Kr-`UB9VRAx2Mw`o&mBJAAe4UB>R z$nbQ$f5%4!Nj=||j>KMk6X(JqGYLhA9Gidd>bN6~aD%;cl)Nw47fP$rQeJ}n$fdd0 zUIc=QXKv(7aMTq2^vq?Vzm7IhfTV`?ohe!M!l)b-ByW@1t})#q6cQ=%L~2i z+*ZR_;t$4{b++4onKv8PjprS-H%*kfnK8;0jE8!5a3*R^&W`x@hQOmb2cPT$A^CYM z9Gx|1)TO4-NbEi=%aY`g&8L<5x$%%xs#mczGpR#|klRALNou#qqVj*jNvRq6yWWje z7xAx>(6#df1#(!ks2zW}ft0uyrZ_TD%MU11^yAeyra^mh{dCEqz>W^Zs|7=_CwW0q zn90!kq*7^p8>`%yY+@b;4yf>fUaKjR1ZF7JY-dboOCX3Hdh0!Z#%h8*?hP@FV)g^T zWlEkp0c;2$J~zvXoW#!CVQNb%`h){uSlA+VeIK83D92K*--4LkPQRs*cT+30ZP9f1 zI70qEtolM{Y#8XC*IrwwPMOQ~u+8ygCaH*NAev@jGsR?9DszzXYMZH0Ax_maMXx8o9$XTrECQSWWwR4B%Z1+ z<{smzpZvsylx;r+-=iidoIj4~F)nC_U@oGdgs{eb% zUo*D~{i8Fu`G{vAs40Qn{AFyHPwFPhZM0V8D=;HWqm5jN7T29!wK}ovXi7P1UDsYa zb4m@|yw7+;tGBx)5Zb^*jjni z%X*n*!)<#jWp`?C_U& zSr2sd*eD!9`_@VGns%92QmSdvN3z3D(Z1z1?h~fS2B9%N@)odoK_qx&ai>ue>NACR zTk_G0Nrh*ftQnraX{|i6cmhm$vH}hn89AMjgx!e0dY8%iL(iR$W_pDk0TY0*S~iR@ zZ`oTpJr-#!`ZZEMNKfQL((Ei9fuBtRAsiKfX6)P$jc^Y|=<^5jFHVgqGCYJ7Lu-e% zK7TUYP`DGus3cm>IK@RAY3K>&ZBZabkLiW!GmFCE&SQ~>cJ{eyYgO=522P3*0{TQQ zvlDcPSX#fSZ+i=ZF7fJ(BC%vY0+jLck(!|%8ZAh#+t0xCpWe8Sw4a9;Ut(Zlp2(UK zE%q02UC3R95ldcho|9apv8Zj%U7FCz>a?u~EqfQieg8w!pQ0WEJS1&w?u&LJzuPn7<{b1*Q{=+}t4Xw1;=C0&)>ASb zu{iMGF}5ymu4{ZkylAX1Ym!f5}HT`qyUof=Xti>I8)M;yiH$n^>CTf1HtQ=s2tQc>nhLr-_8GFH{92a2VrOW3Mi zx!R-9#9SX)SKr^vG#~SxiW0{=@wF^na!O2&)}Fwi+rzrY;@1d07+20X2QtMCfU^we z`6N9K623{+?*6i=ntH%7aa0vtnGsn+e9apLH@TPZU@#9F7IUyv~;3$7?L`71Xs(71#{ZnS@&0mca z%NFNIaCApYkG;xw#}v(@k}m2JOnnkB5<5f8mj`A%L8{?{D>C++ob1zgcaDr^96!G< z6jG<*^}xPGd7{Z;J*b8GQUVnq;%MJfI-$R8<)qJG;Ft%)bhHQasPJJE$n@j)!kZ~s zPhj8-@6H57$F0E*c8;iMu1x?r;>%)uLghd{I5{N=__ z_xV~08qwgIwo%U*zes-V6Iun=vOj>Fu>}mJkEba)KNCSu%txcsp<<2xAk5wt)c5s} zX9Mwa-MHj{qr&G(fYo8JG;h=vvFwNCN*);~$yOKtS5mOds^;Z_3<>#e_|e{ke-lRs;;v)OAFwjHwB|x^;UI=XIrC4Ev8px?kL>ROPQT{G~nBe(WMbS(k{E z1~!9^LM6CAlt@Pqsk}8^NT>1h^YWhJt4u@YnazMYi~OPtsd=mC0MtD$9BRf(qdW~= z_g@&rWAqkPIZUcXngA>3|Bv~Km@qN9!9Fn)#E zB+CNiBikHlY>ZW)6<=^T$8bkXzE7q9#_^d2$8QX~lKFpq37ch8AX#~6qCDD3yx&D( zQzo$kscxA=n{-eDFKL^IG|mn}jGbxaJwYM|Y+_{J=&#RpED;b0+@Yi|2zlqcjVCeH9tOwL}q; zUHH0?WuNEf@Z;a1g}t#y{pEJicNAr#RN@y@XfU$gQjjxn-P7ZAiH{1G-TQT$9fL8% zu^@S82~j|sRA1o8Y%cSe-~cmIE@Qi0_Rg-qGs6($h|&U#!S$>1vJfaEAz$45J8kzt5^nY9K4}hLF*8`#wD!|w>%xW9Hm9fApC)d6;%7@o{ z#N}aUUrY(jG)MrP+VKVkc3j76 zh9St(GrUJ^o!Yl$P(fD??^?e3WO1ccTg0MS_DHv@nE&|gF6KN1Qa5`xg)zOom9E?| z<}w@}UN2+otSF&z_Qa*o!U{?;6n{%&Eo~0zb%_^y>ZtEUp#`tpBb-d7wupDqxzA0H zBo`FOj25Q{1ZtJI1sTj}&Ci4Jq)c>_Y3!Sb=c@MuM9ByL5rbjivHy5x3R@#k26MUY zquWB*bUtEi^>8-G^7T|LANdwveDI>_@dXs(MXmWnyrtc~$#r|EDymG?<<<`GhyZ-~ zm0MdD&-6`;zN5hQK-Ta5a5wts=Q7`Nj)Cd1mWW3-sl6L>hx`{c%KM(pG1br0%o$x7 zhE5nO<1)#4(}rmnLM9gkz`0D9@+B&uHH>{~>vNNzM6V#lP0jd|iUZJt#s7JcC z}r1O9>DpqE$c5)`Mv zv}BqNdQO{ImT(ZI81R1%wbuZAeJTdIU z1#ow^xc-^BjSHg3A2u~}@4H^cX=d>cmt`DT9Ai8&=&;UVjIkEE@?%50fHr|l>QPz* zQ)RcyA$oKRgYk1jgVG)~X`aSppmp3o}SAlBMNfh81^zNv>z(3lJiZS|;wB|j-nf+^r zY#<-X$;xd?L!3g>ICiQk>qIYIy-WS2nla{?^YogAaM73A&Ej((-^~l%Gs|I{VHVv)=%;MANGv(+#7vAO$KzdQ{;-{LV0l14oq_{Vl;oK;vN7Z50{`JLR6>9 zcv@`q8Yas0jbpmDQl|HeNBUw6>2_egCRc4C(i5*5l%`%2o>ot2CHLoCSS8KS1+AIo z%-ohSLpv#nzUw{sL|mAD;nHPx+0o_7pMuVAOZ!&7q|yDnP^nWq_Dgo=CTol}vcQxJ zncp;NLr2x2WB~tEVsCpJ-wRCUx)XWE0_?RZ1nvZPMuw(!%Rx@0E;6r32fvB7b}~X# zcOX%SnnvjRh~dtg>XX3 zLeuS_&W%NsqxCYr(r?pbV`T5eQQF$-BC6OJX`6$@avwQP;9R)kn}NSZ&OErEST5y) zwMBFJ-Uhh8CqhW4_C&DlO>JU%&482fcpzs}J#V=VT{K?O!?rk=+)ab>3{Rz_Qg0|m z(jG+Ey%c#MNsgF#^102OM*wWIrDipSkxb4se$RWxHyW436ewDaYhTMW5_Zfxt4ptA+Fg8t`s8$+J*&$;hft%G080nAXQgfQ>j+6ll@tp;m z_%8Q1s&|(MRTr&_9+=u_mXm^N{vI`icF3u`v9RU>lD}JXt;Z5&N8#!HmmeLrgO)&Z zYSN}Jav#=Pr~9VMrB;a^;?;^~l{GoaJ+phdT!aZ-mSa8ozjH^=q-`m7Hd>D$>(`(| z(4cpz`q71j5a!~pvvd%nevSle4MCX|%PL{5eRuekAa^FYB^ym)gwD)TbX@X40a^$y zd?r_ZZ4#T^*HJENIV0TdnHwTF}6&rK!vs6uYwQk@b1EiI0wZI$zo&ryx~Wlhb152hCL@ zv}&PvegMKY@0;z>w*TtqN9z>*dh!4HWr!Zr`XKo2)6Xg4XR*HR0(lF-DEMFd@MCbwiax#$AgoyeCS(3l7{)%^R1c?SQx8GH_<%baXI2E`~(YQfS^Mzw?l z0{*0}Zz}HhdzdOR1uGARXP2slH2!wcEScTbxdH5F#Gk{r|XROu~4+Zi(CJO$A_ z(UT0A)}gB=V2Vz`qQd?&M=TN*`C`*ydwuAZg1luye+FRu4cVXmmffr2ceU& z2eFLR`{fLJWI@LR>c1CTl?j_IOMWID!(>@^%eb+dc||nub2%n z!vjuQskpcQM2xfN*&{3qi(nj(NJG7kA>Om}Or8KM)m=S*p0ZzL<%KT&&?f^)U}crdAoVtwBAAG|)!jMzeK#HmTDO zNR=keqw7O7PmefyxLE@w(j4aq*9NuYGBi?raoogN15F4B47#JZvfosgeDU4uj(=mZ z>J_fjP{Yu&jD5(}Qf`8zIU2PAQjSF4Ehq)ZA4m5>Ft^0j#VpB?@1i6R z30WI@*oNE~6{k=HCTc!=4!#d;F73CYwkqQ~Hl7gk8|j|BA+n=(Gi7d-_+lnJa{<8V z=xGbld>_Jg4g^eoR3Be|^<`#GvC(})Kp7&;OQjhw)YgW*M*}FjB`+c$+(PfUCFt{ zBTDG+avlH1A1p8a04_xv>Eo9^V7P7#b4`Z6r%%jR-P0$I-zq~|YFfuuX7s-FESA#+ z@Nxsec2v1m%EdnB5JYR~Wo8G1gd;LPFbZD&v!cSXpH}WfD4@Nk&N?gf7mpValv}sc zfrf8oAYZfz4-SgH&t0SZU~FR1U?uXnX1>hPR}l)L`AHi;lo>$+!aKHyH)k8E?lBT2Mvd?Y-48U%lNhzyE z)y&bpGEAz-<}Fp$AQCSsaaryUFHdo~jCBL`++Xi|&^bY9(|*qtcYzF^yFSfhIaY$; z+%mO+sd#v9=ZO&wO4(*S-*H3@1wB}8qr<{*Df*-j|4NAO6VjzgOy)Nh)L~3 zx(~Ny1;Yymz2hYmUj$fK-_G)3)>PM%^X108tIMR%x<9!Af|6rUO5tYlM23|1gYOO! z?OV^c+EW4ycX;K_%v#`qq zx+IU$KrXW*>KDLrg7dXpwl3&?M$0ZT{tRl{sY1slb=3a01_c@7$S``~j4Od=klb^ogpfAleW_URVm%b#}D_`#10%X*opPMx=!uM3TuYZ&<- zK>!g(AI<3rG!xj2Cr=tM*@830+oMe@X#9n%we$=zZyrDMTKso+XvnM*s3Vx+J9I!5 z&e*|+4yXa%zX!6-9h;gS;UBoRL);%Yv!v0J6=-#OTZwo@M%HbQnBB2il6$r0txD1y{#X1#wb?dQ%BmaQPN8Dio~x%3 zf}|5;i!7zkm;DBbK1AiUjn%(njdQakNJ!nx>|yPP$H0#ahe8|NFEaaIe!lQ^g{dWS zY9!5hg*F5C$7tKdD2kI;FyB>y@^FH+I~=j+d;m+0)IHDAYw`Pz$tL)Ug(4f!ci_L7 zYuR*C1~Z`oRe*tOGga^S-JUnlxv1}pfJ zs@dr(g7Y{8$KKRcuh!k@E%2u7ghQL|-7XE?=A&%E$1!M2j4gbj5C%3k|7bt8r)Qbp}$U*l51_0qx;BQzXr z03&?Xb?ttqz#S9FleHY?8ojchAUUsy^NIW6pz_3q-YG%8m+N8!+~mtU;GfsrJK`xr zVi^-z#5>L0URC$RvQ!m0znH99=h{W-4r#v)Wl$Dw>a^1vbzDMSj}j#lI$_#Y;GUGD zS1oHoT~K=|4!O)~^SofJH?YitH@J(irxgFa0zXv6`8%A8zfbAgU-2>efh6^uZ4KPx zL%DlV-4DgPy9|d{zrl+X{}0mDU8+;*x^rgi{6`%oXDAbzfEe3dzq0uQ2q|I%p_Dck z(RyH$du9Pn2LGFJ<=a+aNEi1IV)zxwrMTJ7tQdZXLfD$72kSEFv9NyN$0G^K;f4>O@~)0aF3S9LsSpUglloZnSewy}WyO zzLsOMd+hGko8-Sqc@nY4?wu-9=fgk=d!iW}wS*rv@|*XDVd*B&_=@3Q?7ePSkJ0B& zo*E*JXA^@iooyj{Zr@+b`Da^MsdxE5df4M)exCQRB1xg({bTNGbIi3CSZ19stot=; zYx#lDNR~I0G|mALGfZgtl>qoGXXeE*D>=Ws`2LLvGS%cknE7tvbg=oS!n!(z#D+Sa z8^IVqKa#2-+YlQeoU*Id)RMo_sqTLr`uvFT^T&E;eQG*yfwQuVN zmez1kcG#pNEr-7mea-%>ci3FCjCm*aZx(Wft-?QfD??%>tn5*$)4|ux+a#qBINoMl5jst0^I|99HCG#YJ zEmFQ@I~IQnC-`ZhL5r~TL48b-=+RV!0FFs2hxLc=gR#j*8(!)kK9>)5S2K2*>~hBa z0qCLECm2WKZs^gUk~i}SmhzWQ$OKBG%XpmZ$MBxkJ$O4SS?c63%CaZF}vFH+KE3kGn%K(%{_p&-xzyT~&Hd z**@UBmx;(iAdJVen~YX8Rz^E~sIP|=0pTY-Nab6SAxWDIm^GQ72<0jkgxn4C6O2Nd zDnT*;0F#-K>n2STHt|-aChE`MG2afh2P#rb5Cib@SoPV2R;aRG zpl#_t3(nGJJVJfJ4QscYK1UR?i)5~RihQKPfgztuqd3LyFCFj?syx2% ze6{F_65?zRjEKdEne@v>(%OV(jq)BqCW8+22$zC&3R|FUjYxab*1~ZoQ!+3)v#fy7 zh~TCU2W}}C8nWO;{P9(*h9V6Mpu{MBq7pzcr91y0LF+%2BE%E?=T|KkFqhqU?G&Wj zM6eQ?hOxKTQ@l>Bf+8O*1SCg5h}wf*h<=85mYW7Y`1HtTV#k$0GclL!Y|u(8o`1MS zW*1P%Q@}MLnnDa`_&0P7iFsd~i)`EeM)D2H>5y#_EJ0NgauX(bl^$M+bF+nO%$d=4 znn0EB`somm+wNV-__V5f87_Mr$iJ$e`j2c$X`7_OTGZXkQM9;^@G`xpHo&WJM{3kX zwbF$2;l!&`lJjOz;yQtN{J^t4vmPO%S6LDjpx4wy+IAckyiCniKg2rHOnBFS8X08w zPuv3z6YN20tjQVsJUc5jN_WzP6+JIpiN3HSUVl5jpozqwiLX+8;c=gF&)L{#)bk87 zRoTQEoWU6lLDXx1n+>%+P}{xd6tcg@#gx2>H3DBFWd3ig&pvyJw}sWV^#=e>dnM<% z?I>$9^B$j5DCq4qenxKB#x_p!X6lo&6rLQ)!F7TvjZ*m*Rf6etT{b$Ym6K8X%nLkX zQnE*}x%LtBtXX3ABtFp^26dCxSx?K_zHJ8oKC%|9gQC>lA&g#G z@}5O7f&}QOoFzUUfD8XrM}+D2nd7f5RrjvW(pW;kXz|PPeQ&| zWuP1uo4jGp0e&5MO|_7(pwvGAs#0wzD^?`LT@uX~4bX zOpknzKde5Gfz}Kw$qKiBf?lBKFl_my?!z$5e<1mP#;GY8I^Q@g>By|Zo4vlBao%u4 zH*ku3%@Hx(b=RI=Z1rTg;kQioNVce%ZZ-4z?@@;(Rix#D=iPfd%S@elo_)Vq#W8B0aZceQ;PhRn`GFWx~ zq|(QDxbGLYyp(yKYp`VE+*OAj^(iTybr<`$;i}B${|ry7`Y5tV4Sb?td%oCauz*eWXa(sgmvatvb8y2Q?a|N>9)>_-vW09obv(D7&)f1)VN57|o($%Wt+4GF&S6&n>b<^3FwCjAS z!)dK1M!n-1+d8IRY@HbqBNh4ZP4CzGfGpv&Hsn3ui9ctg>jCE`Ic`TS(oj*@dSmNb;8bM7F7Xc+FDwTh`yplzJnch75O}RI3fKU? W!