{
    "componentChunkName": "component---src-templates-post-tsx",
    "path": "/articles/ia-semana-2026-06-08/",
    "result": {"data":{"logo":{"childImageSharp":{"fixed":{"base64":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABQAAAAGCAYAAADDl76dAAAACXBIWXMAAA7DAAAOwwHHb6hkAAABI0lEQVQY012RTUtCQRiF31K7KkErIYK+pQ/QULAkWghtShSkNoG07kf0m/oTrSIIr47rCFctW+S8+ZHPxJVr3Vw8zOHMzDucOaIdEe2mRE08pyaZUZMQ2xb5fBSBsQAzFpxzUx2uiZkfPSPakYKaWMm2pWl9uVATL6mJrw97N8LkKwsUIxcCFoF9wIt46ZkW7SaztiVXo/f768Fb4862pNp/Es85l4LvU+ASqAB7QA3YAMrASUiwfwbkgSNRs7RtfakNe8364LVSt76c2xdJO+di4cBgUAHYBA5DfQuUgB3gOHxwF9gKIufUxIrWl4b1parGy/efZdU5t8JkVJ6LG5ABlue8xF/kaSleUMqBmqU1NfFAy/jjQWAkcx//r4AIv6X8AMOCRaz8T6+LAAAAAElFTkSuQmCC","width":400,"height":127,"src":"/static/9328800beac7b829cd8ddddaeeffe108/3891b/davidgonzalezfx-logo.png","srcSet":"/static/9328800beac7b829cd8ddddaeeffe108/3891b/davidgonzalezfx-logo.png 1x,\n/static/9328800beac7b829cd8ddddaeeffe108/8c332/davidgonzalezfx-logo.png 1.5x,\n/static/9328800beac7b829cd8ddddaeeffe108/bc8e0/davidgonzalezfx-logo.png 2x"}}},"markdownRemark":{"html":"<h2>Aprendizaje semanal de IA</h2>\n<p>Esta semana ha sido intensa en el frente de modelos frontera, herramientas de desarrollo agente y la emergencia de “loop engineering” como nuevo paradigma. Anthropic lanzó (y luego suspendió) sus modelos más potentes, OpenAI y Anthropic avanzan hacia IPO, y la comunidad de ingenieros está construyendo fábricas de software autónomas.</p>\n<h2 style=\"text-align: center;\">🧠</h2>\n<h2 style=\"text-align: center;\">Modelos</h2>\n<h3><a href=\"https://www.alphaxiv.org/abs/2026.claude-fable-5-mythos-5\">Claude Fable 5 and Claude Mythos 5 | alphaXiv</a></h3>\n<p>Anthropic presentó su nueva familia Mythos-class: <strong>Fable 5</strong> (versión segura para uso general) y <strong>Mythos 5</strong> (sin salvaguardas para defensores cibernéticos aprobados). Fable 5 supera a todos los modelos previos en benchmarks de ingeniería de software, investigación científica, visión y trabajo de conocimiento. En pruebas tempranas, Stripe comprimió meses de migración en un código base de 50M líneas Ruby a un solo día. Precio: $10M input / $50M output tokens — menos de la mitad que Mythos Preview.</p>\n<h3><a href=\"https://x.com/AnthropicAI/status/2065597531644743999\">Anthropic en X: “The US government… has issued an export control directive to suspend all access to Fable 5 and Mythos 5”</a></h3>\n<p>Tres días después del lanzamiento, el gobierno de EE. UU. emitió una directiva de control de exportaciones que obliga a Anthropic a <strong>suspender acceso a Fable 5 y Mythos 5 para cualquier nacional extranjero</strong>, incluyendo sus propios empleados. El efecto neto: deshabilitaron ambos modelos para todos los clientes. Anthropic califica esto de “malentendido” y trabaja para restaurar acceso. Los demás modelos Claude no se ven afectados.</p>\n<h3><a href=\"https://x.com/elder_plinius/status/2064478648057610422\">Pliny the Liberator 🐉 en X: ”🚿 FABLE-5 SYS PROMPT LEAK”</a></h3>\n<p>Filtración masiva del <strong>system prompt de Fable 5</strong> (~120,000 caracteres). Revela detalles como: knowledge cutoff enero 2026, arquitectura de memoria persistente para artifacts (key-value storage), integración MCP para apps externas, y políticas de rechazo matizadas. El prompt incluye instrucciones detalladas sobre tono, formato, bienestar del usuario, y recordatorios de Anthropic que se inyectan dinámicamente (cyber_warning, ethics_reminder, etc.).</p>\n<h3><a href=\"https://www.anthropic.com/news/claude-fable-5-mythos-5\">Claude Fable 5 and Claude Mythos 5 | Anthropic</a></h3>\n<p>El anuncio oficial confirma las capacidades: <strong>autonomía extendida</strong> (trabajan más tiempo que cualquier Claude previo), ingeniería de software (migraciones masivas, refactoring), trabajo de conocimiento, visión, memoria y ciencias de la vida. Mythos 5 se despliega inicialmente via <strong>Project Glasswing</strong> con el gobierno de EE. UU. para defensa cibernética. Nota: la página muestra una actualización del 12 junio confirmando la suspensión de acceso.</p>\n<h3><a href=\"https://x.com/askalphaxiv/status/2064504303096828345\">alphaXiv en X: “As believers of open research, we are disappointed to see Anthropic silently degrading Fable 5 for AI development”</a></h3>\n<p>AlphaXiv denuncia que Anthropic <strong>degrada silenciosamente Fable 5 para investigación en IA</strong>: temas como pipelines de preentrenamiento, infraestructura de entrenamiento distribuido o diseño de aceleradores ML “pueden tener efectividad limitada”. La preocupación: no es un rechazo visible sino una intervención invisible que debilita respuestas sin que el usuario lo sepa, rompiendo la capacidad de auditar si un fallo viene de la idea, la implementación o el proveedor del modelo.</p>\n<h2 style=\"text-align: center;\">🛠️</h2>\n<h2 style=\"text-align: center;\">Herramientas / productividad</h2>\n<h3><a href=\"https://code.claude.com/docs/es/workflows\">Orquestar subagentes a escala con flujos de trabajo dinámicos - Claude Code Docs</a></h3>\n<p>Documentación oficial de <strong>dynamic workflows</strong> en Claude Code: scripts JavaScript que orquestan docenas a cientos de subagentes. A diferencia de subagentes/skills/equipos donde Claude decide turno a turno, aquí el plan vive en código — el runtime ejecuta el bucle, ramificación y resultados intermedios, dejando el contexto de Claude solo para la respuesta final. Permite patrones de calidad repetibles: agentes independientes que revisan adversarialmente hallazgos antes de reportar. Incluye <code class=\"language-text\">/deep-research</code> como workflow integrado.</p>\n<h3><a href=\"https://venturebeat.com/technology/xiaomis-new-open-source-agentic-ai-coding-harness-mimo-code-beats-claude-code-at-ultra-long-200-step-tasks?utm_source=tldrai\">Xiaomi’s new open source, agentic AI coding harness MiMo Code beats Claude Code at ultra-long, 200+ step tasks | VentureBeat</a></h3>\n<p><strong>MiMo Code</strong> de Xiaomi (open source) supera a Claude Code en tareas ultra-largas de 200+ pasos. Es un harness agente que planifica, ejecuta y verifica en bucles extensos. Parte de la apuesta de Xiaomi por tooling agente abierto y de alto rendimiento.</p>\n<h3><a href=\"https://github.com/generalaction/emdash\">generalaction/emdash: Emdash is the Open-Source Agentic Development Environment</a></h3>\n<p><strong>Emdash</strong> (YC W26): entorno de desarrollo agente open-source, desktop-first. Ejecuta <strong>múltiples agentes de coding en paralelo</strong> (Claude Code, Codex, OpenCode, Gemini, Amp, etc.), cada uno en su propio <strong>git worktree</strong> para aislamiento total. Soporta proyectos locales y remotos vía SSH. Revisa diffs, crea PRs, inspecciona CI y mergea desde un solo lugar. Local-first: estado en SQLite local, no envía código a servidores Emdash.</p>\n<h3><a href=\"https://github.com/AgentWrapper/agent-orchestrator\">AgentWrapper/agent-orchestrator: Agentic orchestrator for parallel coding agents</a></h3>\n<p><strong>Agent Orchestrator (AO)</strong>: capa de orquestación agnóstica al agente (Claude Code, Codex, Aider) y al runtime (tmux, Docker, ConPTY). El orchestrator agent planifica tareas, spawnea workers en worktrees aislados, cada uno con su branch y PR. <strong>Reacciona automáticamente a fallos de CI y comentarios de review</strong>. Dashboard en localhost:3000. Solo requiere intervención humana cuando se necesita juicio.</p>\n<h3><a href=\"https://github.com/junhoyeo/contrabass\">junhoyeo/contrabass: Go &#x26; Charm stack implementation of OpenAI’s Symphony</a></h3>\n<p><strong>Contrabass</strong>: reimplementación en Go + Charm (Bubble Tea TUI) de <strong>Symphony de OpenAI</strong>. Orquestador a nivel de proyecto para agentes de coding: parsea WORKFLOW.md con front-matter YAML, adapters para Linear/GitHub Issues/tablero interno, runners para Codex, OpenCode, oh-my-opencode, OMX, OMC. Equipos multi-agente con pipeline por fases (plan → exec → verify), TUI en vivo, dashboard React (Ziikoo) con streaming SSE, localización zh-CN. Modo tmux multi-proceso o goroutine in-process.</p>\n<h3><a href=\"https://github.com/openai/symphony\">openai/symphony: Symphony turns project work into isolated, autonomous implementation runs</a></h3>\n<p><strong>Symphony</strong> (OpenAI): convierte trabajo de proyecto en runs de implementación aislados y autónomos. Los ingenieros <strong>gestionan el trabajo, no supervisan agentes</strong>. Monitorea tablero Linear, spawnea agents, completan tareas y entregan prueba de trabajo: estado CI, feedback de review, análisis de complejidad, videos walkthrough. Al aceptar, landen PRs de forma segura. Preview de ingeniería para entornos de confianza. Implementación de referencia en Elixir.</p>\n<h3><a href=\"https://github.com/Priivacy-ai/spec-kitty\">Priivacy-ai/spec-kitty: Spec-Driven Development for serious software developers</a></h3>\n<p><strong>Spec Kitty</strong>: CLI open-source para <strong>Spec-Driven Development</strong> gobernado. Flujo: <code class=\"language-text\">spec → plan → tasks → next → review → accept → merge</code>. Mantiene specs, planes, work packages, criterios de aceptación, estado de review y decisiones de merge <strong>en el repo</strong> (bajo <code class=\"language-text\">kitty-specs/</code>). Agentes corren en <strong>git worktrees aislados</strong> (<code class=\"language-text\">.worktrees/</code>) para paralelismo sin caos de branches. Dashboard kanban local opcional. Integra slash commands/skills para Claude Code, Codex, Cursor, Gemini, Copilot, Windsurf, OpenCode. Cada misión completada genera retrospectiva automática.</p>\n<h3><a href=\"https://www.youtube.com/watch?v=E5-QK3CDVQM\">I Open-Sourced My Own AFK Software Factory - YouTube</a></h3>\n<p>Video demo de <strong>Sand Castle</strong>: librería TypeScript para orquestar agentes de coding en <strong>sandboxes aislados</strong> (Docker). Permite escribir scripts simples: <code class=\"language-text\">sandcastle.run({ agent, sandbox, prompt })</code>. Plantillas incluyen: planner paralelo con paso de review, backlog manager vía GitHub issues (filtrado por label), implementadores paralelos por issue, reviewer adversarial, merger que resuelve conflictos y cierra issues. El autor lo usa en producción para features, QA y mantenimiento AFK.</p>\n<h3><a href=\"https://www.youtube.com/watch?v=EJyuu6zlQCg\">5 Claude Code skills I use every single day - YouTube</a></h3>\n<p>Cinco skills de Claude Code de uso diario para <strong>codificar proceso de ingeniería real</strong>:</p>\n<ol>\n<li><strong>Grill me</strong>: entrevista implacable sobre cada rama del design tree hasta entendimiento compartido (hasta 16+ preguntas en una sesión).</li>\n<li><strong>Write a PRD</strong>: convierte entendimiento en PRD con user stories y decisiones de implementación duraderas, output como GitHub issue.</li>\n<li><strong>PRD to issues</strong>: descompone PRD en vertical slices (tracer bullets) con relaciones de bloqueo para paralelismo.</li>\n<li><strong>TDD skill</strong>: fuerza red-green-refactor, confirma cambios de interfaz, diseña para testabilidad.</li>\n<li><strong>Improve codebase architecture</strong>: explora codebase surfacing confusions, spawnea subagentes para diseñar interfaces alternativas, crea RFC de refactor como issue.</li>\n</ol>\n<h3><a href=\"https://github.com/benjaminshoemaker/vibecode_spec_generator\">benjaminshoemaker/vibecode_spec_generator: Wizard-style app to help people generate specs</a></h3>\n<p><strong>Vibe Scaffold</strong> (vibescaffold.dev): app Next.js con wizard de 4 pasos asistido por IA (OpenAI) para generar specs ejecutables por agentes: One Pager → Dev Spec → Prompt Plan → AGENTS.md. Chat interactivo context-aware, persistencia en localStorage, descarga individual o ZIP, preview de documentos de ejemplo. Edge runtime, Tailwind, Vercel AI SDK, Zustand.</p>\n<h3><a href=\"https://taylorhuston.me/2025/11/26/AI-Toolkit.html\">Three Lessons from Building an AI Coding Toolkit</a></h3>\n<p>Tres lecciones hard-earned construyendo un toolkit de spec-driven development para Claude Code:</p>\n<ol>\n<li><strong>LLMs necesitan criterios de aceptación objetivos</strong>, no subjetivos (“valida email antes de submit y muestra error en &#x3C;200ms” vs “login intuitivo”).</li>\n<li><strong>Haz que la IA revise su propio trabajo con otro sombrero</strong>: subagente code-reviewer califica cada fase; no avanza hasta pasar score. Especialización: writer vs reviewer.</li>\n<li><strong>El contexto es código</strong>: todo vive en markdown en el repo (Project Brief, ADRs, Specs, Tasks, Worklogs). Claude lee, sigue workflows y actualiza. Los archivos <em>son</em> la memoria persistente.</li>\n</ol>\n<h3><a href=\"https://github.com/TaylorHuston/ai-toolkit\">TaylorHuston/ai-toolkit: A lightweight, fully customizable framework to enable Spec Driven Development with Claude Code</a></h3>\n<p><strong>AI Toolkit</strong> (plugin marketplace): framework ligero para SDD con Claude Code. Incluye: plugin con 27 comandos, 21 agentes especializados, automatización inteligente; starter template (51 archivos); development guidelines (33 archivos en 4 dirs). Flujo core: <code class=\"language-text\">/project-brief</code> → <code class=\"language-text\">/adr</code> (decisiones estáticas) → <code class=\"language-text\">/spec</code> (living docs con acceptance criteria) → <code class=\"language-text\">/plan ###</code> (fases TDD) → <code class=\"language-text\">/implement ### PHASE</code> (ejecución con quality gates). Comandos extra: <code class=\"language-text\">/quality</code>, <code class=\"language-text\">/review</code>, <code class=\"language-text\">/retro</code>, <code class=\"language-text\">/sync</code>.</p>\n<h3><a href=\"https://raw.githubusercontent.com/anthropics/claude-code/refs/heads/main/plugins/feature-dev/commands/feature-dev.md\">raw.githubusercontent.com/anthropics/claude-code/…/feature-dev.md</a></h3>\n<p>Comando oficial <code class=\"language-text\">feature-dev</code> de Claude Code: flujo estructurado para desarrollo de features con especificación, planificación, implementación y validación integradas. Parte del plugin ecosystem nativo.</p>\n<h3><a href=\"https://github.com/specpulse/specpulse\">specpulse/specpulse: Specification-Driven Development (SDD) Framework</a></h3>\n<p><strong>SpecPulse</strong>: framework SDD universal con <strong>11 comandos idénticos en 8 plataformas</strong> (Claude Code, Gemini CLI, Windsurf, Cursor, GitHub Copilot, OpenCode, Crush, Qwen Code). CLI-first con AI enhancement: <code class=\"language-text\">/sp-pulse</code> (entry point) → <code class=\"language-text\">/sp-spec</code> → <code class=\"language-text\">/sp-plan</code> → <code class=\"language-text\">/sp-task</code> → <code class=\"language-text\">/sp-execute</code> → <code class=\"language-text\">/sp-status</code> → <code class=\"language-text\">/sp-validate</code> → <code class=\"language-text\">/sp-continue</code> → <code class=\"language-text\">/sp-decompose</code> → <code class=\"language-text\">/sp-clarify</code> → <code class=\"language-text\">/sp-llm-enforce</code>. Crea foundation CLI y luego AI expande/valida. Soporta specs, plans, tasks, execution continua, validation, compliance enforcement.</p>\n<h3><a href=\"https://github.com/buildermethods/agent-os\">buildermethods/agent-os: Agent OS is a system for injecting your codebase standards</a></h3>","htmlAst":{"type":"root","children":[{"type":"element","tagName":"h2","properties":{},"children":[{"type":"text","value":"Aprendizaje semanal de IA"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"Esta semana ha sido intensa en el frente de modelos frontera, herramientas de desarrollo agente y la emergencia de “loop engineering” como nuevo paradigma. Anthropic lanzó (y luego suspendió) sus modelos más potentes, OpenAI y Anthropic avanzan hacia IPO, y la comunidad de ingenieros está construyendo fábricas de software autónomas."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h2","properties":{"style":"text-align: center;"},"children":[{"type":"text","value":"🧠"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h2","properties":{"style":"text-align: center;"},"children":[{"type":"text","value":"Modelos"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://www.alphaxiv.org/abs/2026.claude-fable-5-mythos-5"},"children":[{"type":"text","value":"Claude Fable 5 and Claude Mythos 5 | alphaXiv"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"Anthropic presentó su nueva familia Mythos-class: "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Fable 5"}]},{"type":"text","value":" (versión segura para uso general) y "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Mythos 5"}]},{"type":"text","value":" (sin salvaguardas para defensores cibernéticos aprobados). Fable 5 supera a todos los modelos previos en benchmarks de ingeniería de software, investigación científica, visión y trabajo de conocimiento. En pruebas tempranas, Stripe comprimió meses de migración en un código base de 50M líneas Ruby a un solo día. Precio: $10M input / $50M output tokens — menos de la mitad que Mythos Preview."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://x.com/AnthropicAI/status/2065597531644743999"},"children":[{"type":"text","value":"Anthropic en X: “The US government… has issued an export control directive to suspend all access to Fable 5 and Mythos 5”"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"Tres días después del lanzamiento, el gobierno de EE. UU. emitió una directiva de control de exportaciones que obliga a Anthropic a "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"suspender acceso a Fable 5 y Mythos 5 para cualquier nacional extranjero"}]},{"type":"text","value":", incluyendo sus propios empleados. El efecto neto: deshabilitaron ambos modelos para todos los clientes. Anthropic califica esto de “malentendido” y trabaja para restaurar acceso. Los demás modelos Claude no se ven afectados."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://x.com/elder_plinius/status/2064478648057610422"},"children":[{"type":"text","value":"Pliny the Liberator 🐉 en X: ”🚿 FABLE-5 SYS PROMPT LEAK”"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"Filtración masiva del "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"system prompt de Fable 5"}]},{"type":"text","value":" (~120,000 caracteres). Revela detalles como: knowledge cutoff enero 2026, arquitectura de memoria persistente para artifacts (key-value storage), integración MCP para apps externas, y políticas de rechazo matizadas. El prompt incluye instrucciones detalladas sobre tono, formato, bienestar del usuario, y recordatorios de Anthropic que se inyectan dinámicamente (cyber_warning, ethics_reminder, etc.)."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://www.anthropic.com/news/claude-fable-5-mythos-5"},"children":[{"type":"text","value":"Claude Fable 5 and Claude Mythos 5 | Anthropic"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"El anuncio oficial confirma las capacidades: "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"autonomía extendida"}]},{"type":"text","value":" (trabajan más tiempo que cualquier Claude previo), ingeniería de software (migraciones masivas, refactoring), trabajo de conocimiento, visión, memoria y ciencias de la vida. Mythos 5 se despliega inicialmente via "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Project Glasswing"}]},{"type":"text","value":" con el gobierno de EE. UU. para defensa cibernética. Nota: la página muestra una actualización del 12 junio confirmando la suspensión de acceso."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://x.com/askalphaxiv/status/2064504303096828345"},"children":[{"type":"text","value":"alphaXiv en X: “As believers of open research, we are disappointed to see Anthropic silently degrading Fable 5 for AI development”"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"AlphaXiv denuncia que Anthropic "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"degrada silenciosamente Fable 5 para investigación en IA"}]},{"type":"text","value":": temas como pipelines de preentrenamiento, infraestructura de entrenamiento distribuido o diseño de aceleradores ML “pueden tener efectividad limitada”. La preocupación: no es un rechazo visible sino una intervención invisible que debilita respuestas sin que el usuario lo sepa, rompiendo la capacidad de auditar si un fallo viene de la idea, la implementación o el proveedor del modelo."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h2","properties":{"style":"text-align: center;"},"children":[{"type":"text","value":"🛠️"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h2","properties":{"style":"text-align: center;"},"children":[{"type":"text","value":"Herramientas / productividad"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://code.claude.com/docs/es/workflows"},"children":[{"type":"text","value":"Orquestar subagentes a escala con flujos de trabajo dinámicos - Claude Code Docs"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"Documentación oficial de "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"dynamic workflows"}]},{"type":"text","value":" en Claude Code: scripts JavaScript que orquestan docenas a cientos de subagentes. A diferencia de subagentes/skills/equipos donde Claude decide turno a turno, aquí el plan vive en código — el runtime ejecuta el bucle, ramificación y resultados intermedios, dejando el contexto de Claude solo para la respuesta final. Permite patrones de calidad repetibles: agentes independientes que revisan adversarialmente hallazgos antes de reportar. Incluye "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/deep-research"}]},{"type":"text","value":" como workflow integrado."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://venturebeat.com/technology/xiaomis-new-open-source-agentic-ai-coding-harness-mimo-code-beats-claude-code-at-ultra-long-200-step-tasks?utm_source=tldrai"},"children":[{"type":"text","value":"Xiaomi’s new open source, agentic AI coding harness MiMo Code beats Claude Code at ultra-long, 200+ step tasks | VentureBeat"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"MiMo Code"}]},{"type":"text","value":" de Xiaomi (open source) supera a Claude Code en tareas ultra-largas de 200+ pasos. Es un harness agente que planifica, ejecuta y verifica en bucles extensos. Parte de la apuesta de Xiaomi por tooling agente abierto y de alto rendimiento."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://github.com/generalaction/emdash"},"children":[{"type":"text","value":"generalaction/emdash: Emdash is the Open-Source Agentic Development Environment"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Emdash"}]},{"type":"text","value":" (YC W26): entorno de desarrollo agente open-source, desktop-first. Ejecuta "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"múltiples agentes de coding en paralelo"}]},{"type":"text","value":" (Claude Code, Codex, OpenCode, Gemini, Amp, etc.), cada uno en su propio "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"git worktree"}]},{"type":"text","value":" para aislamiento total. Soporta proyectos locales y remotos vía SSH. Revisa diffs, crea PRs, inspecciona CI y mergea desde un solo lugar. Local-first: estado en SQLite local, no envía código a servidores Emdash."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://github.com/AgentWrapper/agent-orchestrator"},"children":[{"type":"text","value":"AgentWrapper/agent-orchestrator: Agentic orchestrator for parallel coding agents"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Agent Orchestrator (AO)"}]},{"type":"text","value":": capa de orquestación agnóstica al agente (Claude Code, Codex, Aider) y al runtime (tmux, Docker, ConPTY). El orchestrator agent planifica tareas, spawnea workers en worktrees aislados, cada uno con su branch y PR. "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Reacciona automáticamente a fallos de CI y comentarios de review"}]},{"type":"text","value":". Dashboard en localhost:3000. Solo requiere intervención humana cuando se necesita juicio."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://github.com/junhoyeo/contrabass"},"children":[{"type":"text","value":"junhoyeo/contrabass: Go & Charm stack implementation of OpenAI’s Symphony"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Contrabass"}]},{"type":"text","value":": reimplementación en Go + Charm (Bubble Tea TUI) de "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Symphony de OpenAI"}]},{"type":"text","value":". Orquestador a nivel de proyecto para agentes de coding: parsea WORKFLOW.md con front-matter YAML, adapters para Linear/GitHub Issues/tablero interno, runners para Codex, OpenCode, oh-my-opencode, OMX, OMC. Equipos multi-agente con pipeline por fases (plan → exec → verify), TUI en vivo, dashboard React (Ziikoo) con streaming SSE, localización zh-CN. Modo tmux multi-proceso o goroutine in-process."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://github.com/openai/symphony"},"children":[{"type":"text","value":"openai/symphony: Symphony turns project work into isolated, autonomous implementation runs"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Symphony"}]},{"type":"text","value":" (OpenAI): convierte trabajo de proyecto en runs de implementación aislados y autónomos. Los ingenieros "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"gestionan el trabajo, no supervisan agentes"}]},{"type":"text","value":". Monitorea tablero Linear, spawnea agents, completan tareas y entregan prueba de trabajo: estado CI, feedback de review, análisis de complejidad, videos walkthrough. Al aceptar, landen PRs de forma segura. Preview de ingeniería para entornos de confianza. Implementación de referencia en Elixir."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://github.com/Priivacy-ai/spec-kitty"},"children":[{"type":"text","value":"Priivacy-ai/spec-kitty: Spec-Driven Development for serious software developers"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Spec Kitty"}]},{"type":"text","value":": CLI open-source para "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Spec-Driven Development"}]},{"type":"text","value":" gobernado. Flujo: "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"spec → plan → tasks → next → review → accept → merge"}]},{"type":"text","value":". Mantiene specs, planes, work packages, criterios de aceptación, estado de review y decisiones de merge "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"en el repo"}]},{"type":"text","value":" (bajo "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"kitty-specs/"}]},{"type":"text","value":"). Agentes corren en "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"git worktrees aislados"}]},{"type":"text","value":" ("},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":".worktrees/"}]},{"type":"text","value":") para paralelismo sin caos de branches. Dashboard kanban local opcional. Integra slash commands/skills para Claude Code, Codex, Cursor, Gemini, Copilot, Windsurf, OpenCode. Cada misión completada genera retrospectiva automática."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://www.youtube.com/watch?v=E5-QK3CDVQM"},"children":[{"type":"text","value":"I Open-Sourced My Own AFK Software Factory - YouTube"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"Video demo de "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Sand Castle"}]},{"type":"text","value":": librería TypeScript para orquestar agentes de coding en "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"sandboxes aislados"}]},{"type":"text","value":" (Docker). Permite escribir scripts simples: "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"sandcastle.run({ agent, sandbox, prompt })"}]},{"type":"text","value":". Plantillas incluyen: planner paralelo con paso de review, backlog manager vía GitHub issues (filtrado por label), implementadores paralelos por issue, reviewer adversarial, merger que resuelve conflictos y cierra issues. El autor lo usa en producción para features, QA y mantenimiento AFK."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://www.youtube.com/watch?v=EJyuu6zlQCg"},"children":[{"type":"text","value":"5 Claude Code skills I use every single day - YouTube"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"Cinco skills de Claude Code de uso diario para "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"codificar proceso de ingeniería real"}]},{"type":"text","value":":"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"ol","properties":{},"children":[{"type":"text","value":"\n"},{"type":"element","tagName":"li","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Grill me"}]},{"type":"text","value":": entrevista implacable sobre cada rama del design tree hasta entendimiento compartido (hasta 16+ preguntas en una sesión)."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"li","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Write a PRD"}]},{"type":"text","value":": convierte entendimiento en PRD con user stories y decisiones de implementación duraderas, output como GitHub issue."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"li","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"PRD to issues"}]},{"type":"text","value":": descompone PRD en vertical slices (tracer bullets) con relaciones de bloqueo para paralelismo."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"li","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"TDD skill"}]},{"type":"text","value":": fuerza red-green-refactor, confirma cambios de interfaz, diseña para testabilidad."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"li","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Improve codebase architecture"}]},{"type":"text","value":": explora codebase surfacing confusions, spawnea subagentes para diseñar interfaces alternativas, crea RFC de refactor como issue."}]},{"type":"text","value":"\n"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://github.com/benjaminshoemaker/vibecode_spec_generator"},"children":[{"type":"text","value":"benjaminshoemaker/vibecode_spec_generator: Wizard-style app to help people generate specs"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Vibe Scaffold"}]},{"type":"text","value":" (vibescaffold.dev): app Next.js con wizard de 4 pasos asistido por IA (OpenAI) para generar specs ejecutables por agentes: One Pager → Dev Spec → Prompt Plan → AGENTS.md. Chat interactivo context-aware, persistencia en localStorage, descarga individual o ZIP, preview de documentos de ejemplo. Edge runtime, Tailwind, Vercel AI SDK, Zustand."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://taylorhuston.me/2025/11/26/AI-Toolkit.html"},"children":[{"type":"text","value":"Three Lessons from Building an AI Coding Toolkit"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"Tres lecciones hard-earned construyendo un toolkit de spec-driven development para Claude Code:"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"ol","properties":{},"children":[{"type":"text","value":"\n"},{"type":"element","tagName":"li","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"LLMs necesitan criterios de aceptación objetivos"}]},{"type":"text","value":", no subjetivos (“valida email antes de submit y muestra error en <200ms” vs “login intuitivo”)."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"li","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Haz que la IA revise su propio trabajo con otro sombrero"}]},{"type":"text","value":": subagente code-reviewer califica cada fase; no avanza hasta pasar score. Especialización: writer vs reviewer."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"li","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"El contexto es código"}]},{"type":"text","value":": todo vive en markdown en el repo (Project Brief, ADRs, Specs, Tasks, Worklogs). Claude lee, sigue workflows y actualiza. Los archivos "},{"type":"element","tagName":"em","properties":{},"children":[{"type":"text","value":"son"}]},{"type":"text","value":" la memoria persistente."}]},{"type":"text","value":"\n"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://github.com/TaylorHuston/ai-toolkit"},"children":[{"type":"text","value":"TaylorHuston/ai-toolkit: A lightweight, fully customizable framework to enable Spec Driven Development with Claude Code"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"AI Toolkit"}]},{"type":"text","value":" (plugin marketplace): framework ligero para SDD con Claude Code. Incluye: plugin con 27 comandos, 21 agentes especializados, automatización inteligente; starter template (51 archivos); development guidelines (33 archivos en 4 dirs). Flujo core: "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/project-brief"}]},{"type":"text","value":" → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/adr"}]},{"type":"text","value":" (decisiones estáticas) → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/spec"}]},{"type":"text","value":" (living docs con acceptance criteria) → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/plan ###"}]},{"type":"text","value":" (fases TDD) → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/implement ### PHASE"}]},{"type":"text","value":" (ejecución con quality gates). Comandos extra: "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/quality"}]},{"type":"text","value":", "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/review"}]},{"type":"text","value":", "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/retro"}]},{"type":"text","value":", "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sync"}]},{"type":"text","value":"."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://raw.githubusercontent.com/anthropics/claude-code/refs/heads/main/plugins/feature-dev/commands/feature-dev.md"},"children":[{"type":"text","value":"raw.githubusercontent.com/anthropics/claude-code/…/feature-dev.md"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"Comando oficial "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"feature-dev"}]},{"type":"text","value":" de Claude Code: flujo estructurado para desarrollo de features con especificación, planificación, implementación y validación integradas. Parte del plugin ecosystem nativo."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://github.com/specpulse/specpulse"},"children":[{"type":"text","value":"specpulse/specpulse: Specification-Driven Development (SDD) Framework"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"SpecPulse"}]},{"type":"text","value":": framework SDD universal con "},{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"11 comandos idénticos en 8 plataformas"}]},{"type":"text","value":" (Claude Code, Gemini CLI, Windsurf, Cursor, GitHub Copilot, OpenCode, Crush, Qwen Code). CLI-first con AI enhancement: "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sp-pulse"}]},{"type":"text","value":" (entry point) → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sp-spec"}]},{"type":"text","value":" → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sp-plan"}]},{"type":"text","value":" → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sp-task"}]},{"type":"text","value":" → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sp-execute"}]},{"type":"text","value":" → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sp-status"}]},{"type":"text","value":" → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sp-validate"}]},{"type":"text","value":" → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sp-continue"}]},{"type":"text","value":" → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sp-decompose"}]},{"type":"text","value":" → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sp-clarify"}]},{"type":"text","value":" → "},{"type":"element","tagName":"code","properties":{"className":["language-text"]},"children":[{"type":"text","value":"/sp-llm-enforce"}]},{"type":"text","value":". Crea foundation CLI y luego AI expande/valida. Soporta specs, plans, tasks, execution continua, validation, compliance enforcement."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"h3","properties":{},"children":[{"type":"element","tagName":"a","properties":{"href":"https://github.com/buildermethods/agent-os"},"children":[{"type":"text","value":"buildermethods/agent-os: Agent OS is a system for injecting your codebase standards"}]}]}],"data":{"quirksMode":false}},"excerpt":"Aprendizaje semanal de IA Esta semana ha sido intensa en el frente de modelos frontera, herramientas de desarrollo agente y la emergencia de…","fields":{"readingTime":{"text":"8 min read"}},"frontmatter":{"title":"Aprendizaje IA de la Semana: 8 de junio de 2026","userDate":"14 June 2026","date":"2026-06-14T00:00:00.000Z","tags":["AI"],"excerpt":"Resumen semanal de 34 recursos, herramientas y artículos sobre inteligencia artificial que guardé esta semana.","image":{"childImageSharp":{"fluid":{"base64":"data:image/jpeg;base64,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","aspectRatio":1.28099173553719,"src":"/static/ac0c16bf86bf4f316e6158eb812efa34/92418/ia-semanal.jpg","srcSet":"/static/ac0c16bf86bf4f316e6158eb812efa34/0079a/ia-semanal.jpg 930w,\n/static/ac0c16bf86bf4f316e6158eb812efa34/92418/ia-semanal.jpg 1169w","sizes":"(max-width: 1169px) 100vw, 1169px"}}},"author":[{"id":"David Gonzalez","bio":"• Ingeniero de Software | Desarrollador Frontend\n • Disfruto trabajar en equipos y liderar\n • Orientado a objetivos\n • Eterno aprendiz\n • Linux, Bash, C/C++, Python, Javascript, Node, React","avatar":{"children":[{"fluid":{"base64":"data:image/png;base64,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","aspectRatio":1,"src":"/static/3f96c89897028fbbe31a410705e2e9ae/bc8e0/profile.png","srcSet":"/static/3f96c89897028fbbe31a410705e2e9ae/22f7f/profile.png 40w,\n/static/3f96c89897028fbbe31a410705e2e9ae/587c9/profile.png 80w,\n/static/3f96c89897028fbbe31a410705e2e9ae/d966b/profile.png 120w,\n/static/3f96c89897028fbbe31a410705e2e9ae/bc8e0/profile.png 800w,\n/static/3f96c89897028fbbe31a410705e2e9ae/a45dc/profile.png 1080w","sizes":"(max-width: 800px) 100vw, 800px"}}]}}]}},"relatedPosts":{"totalCount":19,"edges":[{"node":{"id":"11e4c7c8-ea2d-5fe3-911d-08fec8d37926","excerpt":"Aprendizaje semanal de IA Introducing the Fusion API, the smartest compound model in the market OpenRouter lanza Fusion API, un modelo…","frontmatter":{"title":"Aprendizaje IA de la Semana: 15 de junio de 2026","date":"2026-06-21T00:00:00.000Z"},"fields":{"readingTime":{"text":"2 min read"},"slug":"/articles/ia-semana-2026-06-15/"}}},{"node":{"id":"c4ba104f-3807-5b69-9349-a5ff863d6bf4","excerpt":"Aprendizaje semanal de IA Esta semana ha sido intensa en el frente de modelos frontera, herramientas de desarrollo agente y la emergencia de…","frontmatter":{"title":"Aprendizaje IA de la Semana: 8 de junio de 2026","date":"2026-06-14T00:00:00.000Z"},"fields":{"readingTime":{"text":"8 min read"},"slug":"/articles/ia-semana-2026-06-08/"}}},{"node":{"id":"89f463de-6745-5575-ae26-6d4022b70aa4","excerpt":"Aprendizaje semanal de IA Esta semana trae actualizaciones significativas en modelos de vanguardia (Claude Opus 4.8, Nemotron 3 Ultra, Qwen…","frontmatter":{"title":"Aprendizaje IA de la Semana: 1 de junio de 2026","date":"2026-06-07T00:00:00.000Z"},"fields":{"readingTime":{"text":"5 min read"},"slug":"/articles/ia-semana-2026-06-01/"}}},{"node":{"id":"c3f91eec-6839-5122-ba80-2709ff0042f1","excerpt":"Aprendizaje semanal de IA Esta semana trae una avalancha de novedades en tooling para agentes: desde frameworks tipo FastAPI para IA hasta…","frontmatter":{"title":"Aprendizaje IA de la Semana: 18 de mayo de 2026","date":"2026-05-24T00:00:00.000Z"},"fields":{"readingTime":{"text":"5 min read"},"slug":"/articles/ia-semana-2026-05-18/"}}},{"node":{"id":"ce5da1cf-68a0-5c05-857d-ed9944aa9e74","excerpt":"Aprendizaje semanal de IA Esta semana trae avances interesantes en agentes autónomos locales, infraestructura de compute a gran escala y una…","frontmatter":{"title":"Aprendizaje IA de la Semana: 11 de mayo de 2026","date":"2026-05-17T00:00:00.000Z"},"fields":{"readingTime":{"text":"5 min read"},"slug":"/articles/ia-semana-2026-05-11/"}}}]}},"pageContext":{"slug":"/articles/ia-semana-2026-06-08/","prev":{"excerpt":"Aprendizaje semanal de IA Esta semana trae actualizaciones significativas en modelos de vanguardia (Claude Opus 4.8, Nemotron 3 Ultra, Qwen…","frontmatter":{"title":"Aprendizaje IA de la Semana: 1 de junio de 2026","tags":["AI"],"date":"2026-06-07T00:00:00.000Z","draft":false,"excerpt":"Resumen semanal de 7 recursos, herramientas y artículos sobre inteligencia artificial que guardé esta semana.","image":{"childImageSharp":{"fluid":{"aspectRatio":1.28099173553719,"base64":"data:image/jpeg;base64,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","sizes":"(max-width: 1169px) 100vw, 1169px","src":"/static/ac0c16bf86bf4f316e6158eb812efa34/92418/ia-semanal.jpg","srcSet":"/static/ac0c16bf86bf4f316e6158eb812efa34/0079a/ia-semanal.jpg 930w,\n/static/ac0c16bf86bf4f316e6158eb812efa34/92418/ia-semanal.jpg 1169w"}}},"author":[{"id":"David Gonzalez","bio":"• Ingeniero de Software | Desarrollador Frontend\n • Disfruto trabajar en equipos y liderar\n • Orientado a objetivos\n • Eterno aprendiz\n • Linux, Bash, C/C++, Python, Javascript, Node, React","avatar":{"children":[{"fluid":{"aspectRatio":1,"base64":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABQAAAAUCAYAAACNiR0NAAAACXBIWXMAAA7EAAAOxAGVKw4bAAAFeUlEQVQ4yyXP7VPThwHA8d+73d73+mLXrvWuvet2XW196tbOm7Wg9aEIREAhMJ5CQkjIE3kkhKeYEPIgEYI8BTCE+FBq5UEZqCAIyCEC2gIqKA+CRcrWrbta9+q7u+0f+Nx9hHszi8w92iDc3k6+XIytUEGBWozZJKe7b5ybYwv03prlxvAsoxOL3J9dIV+nRG82MjW7wfziP1hc/idTcxtMPFhHuHN3gbnHW7S2NGI1q7EYpCTF70WlEOOrqmZk4gmj91YYv7/G/YdbDI3Pk5krwWQt5Nnzf7O28Quj9ze5NvaMkZl1hMnpJTqvdOF2GCixKjAWqElOPIZep6K57TKPnr1iaGqdoQc/cncFzvc/QHTiCMVlVja3fmHjh1fcuLtB18ga84tbCJNTj7l2bZBKh4VCUw56gwG1Kp2yYinldgdZUj3vf/Ax7763A0/dN1yfXOLLuM8oc5TzePUls09/5tHSTzx8+iNzCz/8v3xzeB6PpwKtUoRMcpwCbTbqvGTeeuNN3nr9dbZv+y2v/fpXvPGbt5HI8zh4aDdF5TYW1//D0/VXrL94yfebP/PtwhbCoyebDA4OIc04hlYpodymoznYiCQjG504ie8HLzMbqWXYXUCtOotP9uxi3/4/YrLqePH3l6xtvuLe3AZTc8/pvb2AsPHiJ2Zm5nCUFVFhL8LvLaOtNYBOlsNVXykDdZUM+UxctmRzvjAfRUoSh4/tRZ6fzvR3T1h89i+Gp9YYnVljbHoVYWVpmeXlNfz+0xRbNFSfLsPjsuK2auhzapHsfIdOQyoO0X7Sd/0ee246Mcc/o7i8mNtTy4xMr3Dr9hTh0EVqAkGExvAlzEYJEkkqGk0eam0mVk06/S4zS10NTDc4WP26gTGPlV6bnJ5TZnJi9+Pw+xn5dou/jT6mpbGFEr0RtdqE4GnoQK2UoSvIR6VIQpzyOWZNPJN+PWtXzrB8qZo+cw4BUTTXi+R0muRIEz9HVWzj3OBzwjdXCYWaqSgzYjIXI2gNmbhdNqxmORpFIlplIjLJMbwl2Uw0ahl1ybll1hFKSKAlNQZPRgwHj+wiz1pAbf8a1ZcHqPDnYi4UU1hkR9CqknA71ORK4jl6aDfZmUdJSz2ATB1LR5WccY+UO04l30gTqE+PISbqI/6wexvZ6lzCg6uU+J3IdYcpKVdS6fUg6HQyHHYbudIkRHEfk5YcTVZWFJnyKDp8CgbsGQRlsbhF+1Ae/YTte9/h/R1vk5QaR3PnIJYSD0aTAp+3kEJnFkJaxkly5SdQymPIy/mSE/F/JjfrC8z6NEaaSpgMqKkV/4WiAx8RyIklWXyAT6N+h+jkXuw+DQpjEuYyMR6/DHdtHoJCloFCdoK42B0cit5DWkocrefaudV7g4cXnNyrL6DHcpJAyj7OyWOoMSjRGG2YS1TkW6JJyP6QQwk7ySnYT3mVGMFuUOC1q7CXmwgGL9DTO07fjWk6e8YYiDQxEdDRrY8nIj9CWJeMJU+B0xMiFOmnMRJGafkr2/+0jT373uNo4m4Eu/0kkUg91wfm6ey5Q9fVO1zpHiVysZ+WUDd1bi8Bo4pqowqbzohKV0KpvQaXJ0h98CqRr4YRpRxn56fvkq2KQnD5ErnU0UVjsJO2SO//EI+vkYamSzQ0fUW58yyWYh8Giwu92YHR4sRW6sdZ2Yizsh7v6XakSg0HRR8iNxxGuNDuorm1kzM1YSrcDRhMdgJ152lt66amtg1fVQtVZ87h9tRjd1Rjd57F5W3Gc7qVUnsAvcmDQmsgUxeN6ZQIIRIKE6i7iMNVi6aglJqz52m/0EdVdYjA2Qj+mnb8NWHcvmYcrgZcnia8VSEqva0YzG7kymJ0xlO467MJdkj5LxVBHsqnfH5ZAAAAAElFTkSuQmCC","sizes":"(max-width: 800px) 100vw, 800px","src":"/static/3f96c89897028fbbe31a410705e2e9ae/bc8e0/profile.png","srcSet":"/static/3f96c89897028fbbe31a410705e2e9ae/8ac63/profile.png 200w,\n/static/3f96c89897028fbbe31a410705e2e9ae/3891b/profile.png 400w,\n/static/3f96c89897028fbbe31a410705e2e9ae/bc8e0/profile.png 800w,\n/static/3f96c89897028fbbe31a410705e2e9ae/a45dc/profile.png 1080w"}}]}}]},"fields":{"readingTime":{"text":"5 min read"},"layout":"post","slug":"/articles/ia-semana-2026-06-01/"}},"next":{"excerpt":"Aprendizaje semanal de IA Introducing the Fusion API, the smartest compound model in the market OpenRouter lanza Fusion API, un modelo…","frontmatter":{"title":"Aprendizaje IA de la Semana: 15 de junio de 2026","tags":["AI"],"date":"2026-06-21T00:00:00.000Z","draft":false,"excerpt":"Resumen semanal de 2 recursos, herramientas y artículos sobre inteligencia artificial que guardé esta semana.","image":{"childImageSharp":{"fluid":{"aspectRatio":1.28099173553719,"base64":"data:image/jpeg;base64,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","sizes":"(max-width: 1169px) 100vw, 1169px","src":"/static/ac0c16bf86bf4f316e6158eb812efa34/92418/ia-semanal.jpg","srcSet":"/static/ac0c16bf86bf4f316e6158eb812efa34/0079a/ia-semanal.jpg 930w,\n/static/ac0c16bf86bf4f316e6158eb812efa34/92418/ia-semanal.jpg 1169w"}}},"author":[{"id":"David Gonzalez","bio":"• Ingeniero de Software | Desarrollador Frontend\n • Disfruto trabajar en equipos y liderar\n • Orientado a objetivos\n • Eterno aprendiz\n • Linux, Bash, C/C++, Python, Javascript, Node, React","avatar":{"children":[{"fluid":{"aspectRatio":1,"base64":"data:image/png;base64,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","sizes":"(max-width: 800px) 100vw, 800px","src":"/static/3f96c89897028fbbe31a410705e2e9ae/bc8e0/profile.png","srcSet":"/static/3f96c89897028fbbe31a410705e2e9ae/8ac63/profile.png 200w,\n/static/3f96c89897028fbbe31a410705e2e9ae/3891b/profile.png 400w,\n/static/3f96c89897028fbbe31a410705e2e9ae/bc8e0/profile.png 800w,\n/static/3f96c89897028fbbe31a410705e2e9ae/a45dc/profile.png 1080w"}}]}}]},"fields":{"readingTime":{"text":"2 min read"},"layout":"post","slug":"/articles/ia-semana-2026-06-15/"}},"primaryTag":"AI"}},
    "staticQueryHashes": ["3604224043"]}