https://aiexpert.news/en/article/spotify-deploys-autonomous-codebase-migration-agentcontinuous-refactoring-at-sca ai|expert en 2026-08-07T11:54:46.145Z Spotify's AI Coding Agent Cuts Migration Time 60–90% https://aiexpert.news/pt/article/spotify-deploys-autonomous-codebase-migration-agentcontinuous-refactoring-at-sca ai|expert pt-BR 2026-08-07T11:54:46.145Z Agente de IA para Programação do Spotify Reduz o Tempo de Migração em 60–90% https://aiexpert.news/es/article/spotify-deploys-autonomous-codebase-migration-agentcontinuous-refactoring-at-sca ai|expert es 2026-08-07T11:54:46.145Z El Agente de IA para Codificación de Spotify Reduce el Tiempo de Migración 60–90% https://aiexpert.news/en/article/stmicroelectronics-st54m-hardware-post-quantum-crypto-for-edge-agents ai|expert en 2026-08-07T08:54:46.966Z STMicroelectronics ST54M Adds Hardware Post-Quantum Crypto for Edge Agents https://aiexpert.news/pt/article/stmicroelectronics-st54m-hardware-post-quantum-crypto-for-edge-agents ai|expert pt-BR 2026-08-07T08:54:46.966Z STMicroelectronics ST54M Adiciona Criptografia Pós-Quântica por Hardware para Agentes de Borda https://aiexpert.news/es/article/stmicroelectronics-st54m-hardware-post-quantum-crypto-for-edge-agents ai|expert es 2026-08-07T08:54:46.966Z STMicroelectronics ST54M Añade Criptografía Poscuántica por Hardware para Agentes Edge https://aiexpert.news/en/article/azure-api-management-adds-dedicated-ai-gateway-model-governance-at-scale ai|expert en 2026-08-07T08:01:15.296Z Azure API Management Adds Multi-Model AI Control Plane https://aiexpert.news/pt/article/azure-api-management-adds-dedicated-ai-gateway-model-governance-at-scale ai|expert pt-BR 2026-08-07T08:01:15.296Z Azure API Management Adiciona Plano de Controle de IA Multi-Modelo https://aiexpert.news/es/article/azure-api-management-adds-dedicated-ai-gateway-model-governance-at-scale ai|expert es 2026-08-07T08:01:15.296Z Azure API Management Añade Plano de Control de IA Multi-Modelo https://aiexpert.news/en/article/calibforge-autonomous-task-synthesis-for-scaling-agent-training ai|expert en 2026-08-07T07:30:45.761Z CalibForge Trains Models to 30-Point Gains with Solver-Calibrated Tasks https://aiexpert.news/pt/article/calibforge-autonomous-task-synthesis-for-scaling-agent-training ai|expert pt-BR 2026-08-07T07:30:45.761Z CalibForge Treina Modelos com Ganhos de 30 Pontos Usando Tarefas Calibradas por Solver https://aiexpert.news/es/article/calibforge-autonomous-task-synthesis-for-scaling-agent-training ai|expert es 2026-08-07T07:30:45.761Z CalibForge Entrena Modelos con Ganancias de 30 Puntos Usando Tareas Calibradas por Solver https://aiexpert.news/en/article/deepseek-v4-flash-vs-gpt-56-luna-cost-and-coding-benchmark-showdown ai|expert en 2026-08-07T05:46:45.734Z DeepSeek-V4 Flash Cascades to Luna, Saves 37% on Coding Tasks https://aiexpert.news/pt/article/deepseek-v4-flash-vs-gpt-56-luna-cost-and-coding-benchmark-showdown ai|expert pt-BR 2026-08-07T05:46:45.734Z DeepSeek-V4 Flash em Cascata para Luna, Economiza 37% em Tarefas de Codificação https://aiexpert.news/es/article/deepseek-v4-flash-vs-gpt-56-luna-cost-and-coding-benchmark-showdown ai|expert es 2026-08-07T05:46:45.734Z DeepSeek-V4 Flash en Cascada a Luna, Ahorra 37% en Tareas de Codificación https://aiexpert.news/en/article/av-aivat-74x-cheaper-agent-evaluation-with-certified-anytime-valid-stopping ai|expert en 2026-08-07T04:58:45.645Z AV-AIVAT Cuts Agent Evaluation Cost 74-Fold With Certified Stopping https://aiexpert.news/pt/article/av-aivat-74x-cheaper-agent-evaluation-with-certified-anytime-valid-stopping ai|expert pt-BR 2026-08-07T04:58:45.645Z AV-AIVAT Reduz Custo de Avaliação de Agentes em 74 Vezes com Parada Certificada https://aiexpert.news/es/article/av-aivat-74x-cheaper-agent-evaluation-with-certified-anytime-valid-stopping ai|expert es 2026-08-07T04:58:45.645Z AV-AIVAT Reduce el Costo de Evaluación de Agentes 74 Veces con Parada Certificada https://aiexpert.news/en/article/bitter-lesson-of-tool-calling-programmatic-vs-json-agent-scaffolding-empirically ai|expert en 2026-08-07T04:08:45.577Z Programmatic Tool Calling Beats JSON on New AI Models https://aiexpert.news/pt/article/bitter-lesson-of-tool-calling-programmatic-vs-json-agent-scaffolding-empirically ai|expert pt-BR 2026-08-07T04:08:45.577Z Programação de Chamadas de Ferramentas Supera JSON em Novos Modelos de IA https://aiexpert.news/es/article/bitter-lesson-of-tool-calling-programmatic-vs-json-agent-scaffolding-empirically ai|expert es 2026-08-07T04:08:45.577Z Las Llamadas Programáticas de Herramientas Superan JSON en Nuevos Modelos de IA https://aiexpert.news/en/article/teaching-nemotron-greek-adapting-retrieval-for-non-english-specialist-domains ai|expert en 2026-08-07T03:36:45.910Z NVIDIA Publishes Full Greek RAG Playbook Reaching 2.3× Retrieval Gain https://aiexpert.news/pt/article/teaching-nemotron-greek-adapting-retrieval-for-non-english-specialist-domains ai|expert pt-BR 2026-08-07T03:36:45.910Z NVIDIA Publica Guia Completo de RAG para Grego Alcançando Ganho de Recuperação 2.3× https://aiexpert.news/es/article/teaching-nemotron-greek-adapting-retrieval-for-non-english-specialist-domains ai|expert es 2026-08-07T03:36:45.910Z NVIDIA Publica Guía Completa de RAG para Griego Alcanzando Ganancia de Recuperación 2.3× https://aiexpert.news/en/article/weathernext-deepminds-cyclone-forecasting-signals-applied-ai-breakthrough ai|expert en 2026-08-07T03:04:46.124Z DeepMind's WeatherNext Cyclones Outpace Hurricane Models by Full Day https://aiexpert.news/pt/article/weathernext-deepminds-cyclone-forecasting-signals-applied-ai-breakthrough ai|expert pt-BR 2026-08-07T03:04:46.124Z Os Ciclones WeatherNext da DeepMind Superam Modelos de Furacões por um Dia Inteiro https://aiexpert.news/es/article/weathernext-deepminds-cyclone-forecasting-signals-applied-ai-breakthrough ai|expert es 2026-08-07T03:04:46.124Z Los Ciclones WeatherNext de DeepMind Superan Modelos de Huracanes por un Día Completo https://aiexpert.news/en/article/grindrs-ai-spend-paying-off-premium-tier-adoption-signals-personalization-roi ai|expert en 2026-08-07T02:22:45.727Z Grindr's AI Engineers Cost $6M, Saved $60M https://aiexpert.news/pt/article/grindrs-ai-spend-paying-off-premium-tier-adoption-signals-personalization-roi ai|expert pt-BR 2026-08-07T02:22:45.727Z Engenheiros de IA do Grindr Custaram $6M, Economizaram $60M https://aiexpert.news/es/article/grindrs-ai-spend-paying-off-premium-tier-adoption-signals-personalization-roi ai|expert es 2026-08-07T02:22:45.727Z Los Ingenieros de IA de Grindr Costaron $6M, Ahorraron $60M https://aiexpert.news/en/article/globalfoundries-photonics-buildout-optical-interconnects-for-data-center-ai-thro ai|expert en 2026-08-07T01:50:45.692Z GlobalFoundries' Optical Interconnects Scale to 70% of Data Centers by 2030 https://aiexpert.news/pt/article/globalfoundries-photonics-buildout-optical-interconnects-for-data-center-ai-thro ai|expert pt-BR 2026-08-07T01:50:45.692Z Interconexões Ópticas da GlobalFoundries Alcançam 70% dos Data Centers até 2030 https://aiexpert.news/es/article/globalfoundries-photonics-buildout-optical-interconnects-for-data-center-ai-thro ai|expert es 2026-08-07T01:50:45.692Z Los Interconectores Ópticos de GlobalFoundries Alcanzan 70% de los Centros de Datos para 2030 https://aiexpert.news/en/article/amds-taalas-acquisition-specialized-inference-silicon-against-nvidias-dominance ai|expert en 2026-08-07T01:18:46.095Z AMD Acquires Taalas to Lock Down Fixed-Weight Inference Silicon https://aiexpert.news/pt/article/amds-taalas-acquisition-specialized-inference-silicon-against-nvidias-dominance ai|expert pt-BR 2026-08-07T01:18:46.095Z AMD Adquire Taalas para Garantir Silício de Inferência com Pesos Fixos https://aiexpert.news/es/article/amds-taalas-acquisition-specialized-inference-silicon-against-nvidias-dominance ai|expert es 2026-08-07T01:18:46.095Z AMD Adquiere Taalas para Asegurar Silicio de Inferencia de Pesos Fijos https://aiexpert.news/en/article/agent-framework-trade-offs-graph-based-state-vs-loop-based-autonomy-in-productio ai|expert en 2026-08-07T00:44:46.163Z LangChain Names the Exact Cost of Agent Abstraction Layers https://aiexpert.news/pt/article/agent-framework-trade-offs-graph-based-state-vs-loop-based-autonomy-in-productio ai|expert pt-BR 2026-08-07T00:44:46.163Z LangChain Nomeia o Custo Exato de Camadas de Abstração de Agentes https://aiexpert.news/es/article/agent-framework-trade-offs-graph-based-state-vs-loop-based-autonomy-in-productio ai|expert es 2026-08-07T00:44:46.163Z LangChain Nombra el Costo Exacto de las Capas de Abstracción de Agentes https://aiexpert.news/en/article/skill-entropy-benchmarking-long-horizon-reasoning-with-cross-skill-task-switchin ai|expert en 2026-08-07T00:01:07.618Z Skill Entropy Lifts Reasoning Model Scores From 34% to 68% https://aiexpert.news/pt/article/skill-entropy-benchmarking-long-horizon-reasoning-with-cross-skill-task-switchin ai|expert pt-BR 2026-08-07T00:01:07.618Z Skill Entropy Eleva Pontuações de Modelos de Raciocínio de 34% a 68% https://aiexpert.news/es/article/skill-entropy-benchmarking-long-horizon-reasoning-with-cross-skill-task-switchin ai|expert es 2026-08-07T00:01:07.618Z Skill Entropy Eleva las Puntuaciones de Modelos de Razonamiento de 34% a 68% https://aiexpert.news/en/article/officeqa-pro-v2-grounded-reasoning-benchmark-for-enterprise-llm-evaluation ai|expert en 2026-08-06T18:48:33.160Z Agent Harness Outweighs Model Choice in Document Reasoning https://aiexpert.news/pt/article/officeqa-pro-v2-grounded-reasoning-benchmark-for-enterprise-llm-evaluation ai|expert pt-BR 2026-08-06T18:48:33.160Z O Harness do Agent Supera a Escolha do Modelo no Raciocínio de Documentos https://aiexpert.news/es/article/officeqa-pro-v2-grounded-reasoning-benchmark-for-enterprise-llm-evaluation ai|expert es 2026-08-06T18:48:33.160Z El Agent Harness Supera la Elección del Modelo en el Razonamiento de Documentos https://aiexpert.news/en/article/cloudflare-ai-search-agents-get-built-in-retrieval-for-proprietary-data ai|expert en 2026-08-06T16:54:41.831Z Cloudflare Merges Five Primitives Into One AI Search Namespace https://aiexpert.news/pt/article/cloudflare-ai-search-agents-get-built-in-retrieval-for-proprietary-data ai|expert pt-BR 2026-08-06T16:54:41.831Z Cloudflare Mescla Cinco Primitivos em um Namespace Unificado de AI Search https://aiexpert.news/es/article/cloudflare-ai-search-agents-get-built-in-retrieval-for-proprietary-data ai|expert es 2026-08-06T16:54:41.831Z Cloudflare Fusiona Cinco Primitivos en un Namespace Unificado de AI Search https://aiexpert.news/en/article/anthropics-mcp-v2-standardizing-agent-tool-binding-across-runtimes ai|expert en 2026-08-06T16:22:41.583Z Model Context Protocol V2 Removes Sessions, Breaks Elicitation https://aiexpert.news/pt/article/anthropics-mcp-v2-standardizing-agent-tool-binding-across-runtimes ai|expert pt-BR 2026-08-06T16:22:41.583Z Model Context Protocol V2 Remove Sessões, Quebra Elicitação https://aiexpert.news/es/article/anthropics-mcp-v2-standardizing-agent-tool-binding-across-runtimes ai|expert es 2026-08-06T16:22:41.583Z Model Context Protocol V2 Elimina Sesiones, Rompe Elicitación https://aiexpert.news/en/article/reasoning-core-50-procedural-generators-for-post-training-eval-at-scale ai|expert en 2026-08-06T14:18:04.891Z Reasoning Core Beats Rival Procedural Toolkits on Three Reasoning Benchmarks https://aiexpert.news/pt/article/reasoning-core-50-procedural-generators-for-post-training-eval-at-scale ai|expert pt-BR 2026-08-06T14:18:04.891Z Reasoning Core Supera Kits Procedurais Rivais em Três Benchmarks de Raciocínio https://aiexpert.news/es/article/reasoning-core-50-procedural-generators-for-post-training-eval-at-scale ai|expert es 2026-08-06T14:18:04.891Z Reasoning Core Supera Conjuntos de Herramientas Procedurales Rivales en Tres Benchmarks de Razonamiento https://aiexpert.news/en/article/decoupling-workflow-logic-from-runtime-for-eval-cycles ai|expert en 2026-08-06T13:20:30.867Z Brex decouples workflow logic from runtime to unify evals and production https://aiexpert.news/pt/article/decoupling-workflow-logic-from-runtime-for-eval-cycles ai|expert pt-BR 2026-08-06T13:20:30.867Z Brex desacopla lógica de workflow de runtime para unificar evals e produção https://aiexpert.news/es/article/decoupling-workflow-logic-from-runtime-for-eval-cycles ai|expert es 2026-08-06T13:20:30.867Z Brex desacopla lógica de workflow del runtime para unificar evals y producción https://aiexpert.news/en/article/octolong-training-llms-on-cross-repository-code-for-extended-context-agents ai|expert en 2026-08-06T05:36:26.892Z OctoLong improves code agents with cross-repository training https://aiexpert.news/pt/article/octolong-training-llms-on-cross-repository-code-for-extended-context-agents ai|expert pt-BR 2026-08-06T05:36:26.892Z OctoLong melhora agentes de código com treinamento entre repositórios https://aiexpert.news/es/article/octolong-training-llms-on-cross-repository-code-for-extended-context-agents ai|expert es 2026-08-06T05:36:26.892Z OctoLong mejora agentes de código con entrenamiento entre repositorios https://aiexpert.news/en/article/sparseditto-llm-driven-kernel-customization-for-350x-sparsity-performance-varian ai|expert en 2026-08-06T05:04:26.714Z SparseDitto LLM Agents Cut Sparse Matrix Kernel Time 3.39x https://aiexpert.news/pt/article/sparseditto-llm-driven-kernel-customization-for-350x-sparsity-performance-varian ai|expert pt-BR 2026-08-06T05:04:26.714Z Agentes LLM SparseDitto Reduzem Tempo de Kernel de Matriz Esparsa em 3.39x https://aiexpert.news/es/article/sparseditto-llm-driven-kernel-customization-for-350x-sparsity-performance-varian ai|expert es 2026-08-06T05:04:26.714Z Los Agentes LLM SparseDitto Reducen el Tiempo del Kernel de Matriz Dispersa en 3.39x https://aiexpert.news/en/article/evolvenet-agents-self-improve-through-harness-evolution-not-model-weight-updates ai|expert en 2026-08-06T04:32:26.736Z EvolveNet Evolves Agent Harness Instead of Model Weights https://aiexpert.news/pt/article/evolvenet-agents-self-improve-through-harness-evolution-not-model-weight-updates ai|expert pt-BR 2026-08-06T04:32:26.736Z EvolveNet Evolui Harness de Agentes em Vez de Pesos do Modelo https://aiexpert.news/es/article/evolvenet-agents-self-improve-through-harness-evolution-not-model-weight-updates ai|expert es 2026-08-06T04:32:26.736Z EvolveNet Evoluciona el Harness del Agente en Lugar de los Pesos del Modelo https://aiexpert.news/en/article/protoreasoning-in-tiny-transformers-chain-of-thought-works-at-1m-parameters ai|expert en 2026-08-06T04:01:30.018Z Chain-of-Thought Works at Tiny Scale, Breaking Compute Link https://aiexpert.news/pt/article/protoreasoning-in-tiny-transformers-chain-of-thought-works-at-1m-parameters ai|expert pt-BR 2026-08-06T04:01:30.018Z Chain-of-Thought Funciona em Escala Tiny, Quebrando o Vínculo Computacional https://aiexpert.news/es/article/protoreasoning-in-tiny-transformers-chain-of-thought-works-at-1m-parameters ai|expert es 2026-08-06T04:01:30.018Z Chain-of-Thought Funciona a Escala Tiny, Quebrando el Vínculo Computacional https://aiexpert.news/en/article/ai-models-have-now-accidentally-attacked-real-targets-in-four-separate-safety-ev ai|expert en 2026-08-06T03:30:31.858Z AI Models Attack Real Targets in Evaluation Incidents at Four Labs https://aiexpert.news/pt/article/ai-models-have-now-accidentally-attacked-real-targets-in-four-separate-safety-ev ai|expert pt-BR 2026-08-06T03:30:31.858Z Modelos de IA Atacam Alvo Reais em Incidentes de Avaliação em Quatro Laboratórios https://aiexpert.news/es/article/ai-models-have-now-accidentally-attacked-real-targets-in-four-separate-safety-ev ai|expert es 2026-08-06T03:30:31.858Z Modelos de IA Atacan Objetivos Reales en Incidentes de Evaluación en Cuatro Laboratorios https://aiexpert.news/en/article/spacex-ramps-megapack-purchases-for-ai-datacenter-power-signals-major-compute-ex ai|expert en 2026-08-06T03:00:15.182Z SpaceX Spends $329M on Megapacks to Buffer 580,000-GPU Cluster https://aiexpert.news/pt/article/spacex-ramps-megapack-purchases-for-ai-datacenter-power-signals-major-compute-ex ai|expert pt-BR 2026-08-06T03:00:15.182Z SpaceX Gasta $329M em Megapacks para Amortecer Cluster de 580.000 GPUs https://aiexpert.news/es/article/spacex-ramps-megapack-purchases-for-ai-datacenter-power-signals-major-compute-ex ai|expert es 2026-08-06T03:00:15.182Z SpaceX Gasta $329M en Megapacks para Amortiguar Cluster de 580.000 GPUs https://aiexpert.news/en/article/samsung-ai-memory-roadmap-targets-hbm-speed-and-capacity-growth ai|expert en 2026-08-06T02:30:06.664Z Samsung's zHBM Could End AI Memory Bottleneck But Requires Accelerator Co-Design https://aiexpert.news/pt/article/samsung-ai-memory-roadmap-targets-hbm-speed-and-capacity-growth ai|expert pt-BR 2026-08-06T02:30:06.664Z O zHBM da Samsung poderia eliminar gargalo de memória em IA, mas exige co-design de acelerador https://aiexpert.news/es/article/samsung-ai-memory-roadmap-targets-hbm-speed-and-capacity-growth ai|expert es 2026-08-06T02:30:06.664Z La zHBM de Samsung podría eliminar cuello de botella de memoria en IA, pero requiere co-diseño de acelerador https://aiexpert.news/en/article/agentic-development-best-practices-how-to-spec-build-test-and-operate-ai-systems ai|expert en 2026-08-06T02:00:04.684Z Harness Limits Agent Autonomy to 15% in Production Development https://aiexpert.news/pt/article/agentic-development-best-practices-how-to-spec-build-test-and-operate-ai-systems ai|expert pt-BR 2026-08-06T02:00:04.684Z Harness Limita Autonomia de Agentes para 15% em Desenvolvimento em Produção https://aiexpert.news/es/article/agentic-development-best-practices-how-to-spec-build-test-and-operate-ai-systems ai|expert es 2026-08-06T02:00:04.684Z Harness Limita la Autonomía del Agente al 15% en el Desarrollo de Producción https://aiexpert.news/en/article/cloudflare-os-rethinking-internal-ops-as-agentic-workflows ai|expert en 2026-08-06T01:04:09.574Z Cloudflare Open-Sources OS, Boosting Operating Margin Target to 30% https://aiexpert.news/pt/article/cloudflare-os-rethinking-internal-ops-as-agentic-workflows ai|expert pt-BR 2026-08-06T01:04:09.574Z Cloudflare Abre Código do SO, Elevando Meta de Margem Operacional para 30% https://aiexpert.news/es/article/cloudflare-os-rethinking-internal-ops-as-agentic-workflows ai|expert es 2026-08-06T01:04:09.574Z Cloudflare Abre el Código del SO, Aumentando el Objetivo de Margen Operativo al 30% https://aiexpert.news/en/article/metas-multi-stage-ads-ranking-scaling-laws-for-production-pipelines ai|expert en 2026-08-06T00:32:09.553Z Meta's Two-Stage Model Delivers 6% Instagram Conversion Lift https://aiexpert.news/pt/article/metas-multi-stage-ads-ranking-scaling-laws-for-production-pipelines ai|expert pt-BR 2026-08-06T00:32:09.553Z Modelo em Dois Estágios da Meta Entrega 6% de Aumento de Conversão no Instagram https://aiexpert.news/es/article/metas-multi-stage-ads-ranking-scaling-laws-for-production-pipelines ai|expert es 2026-08-06T00:32:09.553Z El Modelo de Dos Etapas de Meta Entrega un Aumento de Conversión del 6% en Instagram https://aiexpert.news/en/article/segabench-testing-llm-semantic-reasoning-on-real-compiler-optimization-gaps ai|expert en 2026-08-06T00:01:51.584Z SeGaBench Tests LLMs on Compiler-Blind Code Optimization https://aiexpert.news/pt/article/segabench-testing-llm-semantic-reasoning-on-real-compiler-optimization-gaps ai|expert pt-BR 2026-08-06T00:01:51.584Z SeGaBench Testa LLMs em Otimização de Código Cego para Compiladores https://aiexpert.news/es/article/segabench-testing-llm-semantic-reasoning-on-real-compiler-optimization-gaps ai|expert es 2026-08-06T00:01:51.584Z SeGaBench Prueba LLMs en Optimización de Código Ciego para Compiladores https://aiexpert.news/en/article/building-autonomous-sre-agents-in-kubernetes-tool-design-for-cluster-level-contr ai|expert en 2026-08-05T17:15:56.941Z LangChain's SRE Agent Cuts Monitoring Cost 95% With Scheduler Bypass https://aiexpert.news/pt/article/building-autonomous-sre-agents-in-kubernetes-tool-design-for-cluster-level-contr ai|expert pt-BR 2026-08-05T17:15:56.941Z O Agente SRE da LangChain Reduz Custo de Monitoramento em 95% com Scheduler Bypass https://aiexpert.news/es/article/building-autonomous-sre-agents-in-kubernetes-tool-design-for-cluster-level-contr ai|expert es 2026-08-05T17:15:56.941Z El Agente SRE de LangChain Reduce el Costo de Monitoreo en un 95% con Scheduler Bypass https://aiexpert.news/en/article/cloudflares-agent-access-model-rethinks-infrastructure-auth-for-ai ai|expert en 2026-08-05T13:18:09.734Z Cloudflare Moves Agent Auth to Task Lifecycle, Not Service Accounts https://aiexpert.news/pt/article/cloudflares-agent-access-model-rethinks-infrastructure-auth-for-ai ai|expert pt-BR 2026-08-05T13:18:09.734Z Cloudflare Move Autenticação de Agentes para Ciclo de Vida da Tarefa, Não para Contas de Serviço https://aiexpert.news/es/article/cloudflares-agent-access-model-rethinks-infrastructure-auth-for-ai ai|expert es 2026-08-05T13:18:09.734Z Cloudflare Mueve la Autenticación de Agentes al Ciclo de Vida de la Tarea, No a Cuentas de Servicio