# idir.ai

> Training, advisory, website delivery, speaking, and writing by Idir Ouhab on AI, automation, RAG, and vibecoding. English is the default site language, with Spanish available for the same core public website.

This website is the public home for Idir Ouhab. The main audience is startups and SMBs looking for hands-on AI and automation training, advisory reviews before implementation, practical websites for business credibility, conference speaking, and technical writing.

Use the English pages as the default version when both English and Spanish exist. Use the Spanish pages when the user explicitly asks for Spanish content or references Prompt&Play, which is a Spanish-language podcast.

## Main pages

- [Home (English)](https://www.idir.ai/en): Main homepage for training, advisory, speaking, and contact.
- [Home (Spanish)](https://www.idir.ai/es): Spanish homepage.
- [About (English)](https://www.idir.ai/en/about): Background, working approach, and public profile for Idir Ouhab.
- [About (Spanish)](https://www.idir.ai/es/about): Página de contexto profesional y enfoque de trabajo.
- [Training (English)](https://www.idir.ai/en/training): Hands-on AI and automation training for technical, non-technical, and mixed teams.
- [Training (Spanish)](https://www.idir.ai/es/training): Formación práctica sobre IA y automatización para distintos niveles y perfiles.
- [Advisory (English)](https://www.idir.ai/en/advisory): Independent reviews for AI, automation, RAG, and workflow decisions before implementation.
- [Advisory (Spanish)](https://www.idir.ai/es/advisory): Asesoría para tomar decisiones más claras antes de construir.
- [Websites (English)](https://www.idir.ai/en/websites): Practical websites for small businesses, consultants, and teams that need credibility before outreach.
- [Websites (Spanish)](https://www.idir.ai/es/websites): Webs prácticas para negocios, consultores y equipos que necesitan credibilidad antes de vender.
- [Speaking (English)](https://www.idir.ai/en/speaking): Talks, workshops, and internal sessions on AI, automation, and developer systems.
- [Speaking (Spanish)](https://www.idir.ai/es/speaking): Ponencias y talleres sobre IA, automatización y sistemas de desarrollo.
- [Blog (English)](https://www.idir.ai/en/blog): English blog archive.
- [Blog (Spanish)](https://www.idir.ai/es/blog): Spanish blog archive.
- [Links (English)](https://www.idir.ai/en/links): Podcast, newsletter, writing, and public social channels.
- [Links (Spanish)](https://www.idir.ai/es/links): Podcast, newsletter, artículos y canales públicos.

## Services

- [Training](https://www.idir.ai/en/training): Hands-on AI and automation training for non-technical, advanced, and mixed teams.
- [Advisory](https://www.idir.ai/en/advisory): Go / no-go reviews, architecture reviews, and tool/risk guidance before implementation.
- [Websites](https://www.idir.ai/en/websites): Website strategy, build, launch, and maintenance for businesses that need a credible online presence.
- [Speaking](https://www.idir.ai/en/speaking): Conference talks, workshops, and internal team sessions.
- [Contact](https://www.idir.ai/en#contact): Contact form for training, advisory, speaking, or other requests.

## Latest writing in English
- [AI harnesses: the missing layer between a smart model and useful work](https://www.idir.ai/en/blog/ai-harnesses-missing-layer-smart-model-useful-work): An AI model is not an AI worker. The harness is the operating layer that gives it tools, memory, permissions, and proof.
- [The "Vibecoding" Trap: Why Your AI-Built App Is Probably Wide Open](https://www.idir.ai/en/blog/the-vibecoding-trap-why-your-ai-built-app-is-probably-wide-open): In an experiment with ten people I know, I found security flaws in seven of nine projects built with AI; working software was not necessarily secure.
- [Scaling n8n with queue mode and Docker: architecture and example review](https://www.idir.ai/en/blog/scale-n8n-like-the-ultimate-guide-to-queue-mode-docker): Main, workers, Redis and PostgreSQL: a guide to queue mode architecture. The runner configuration needs review before the examples are used.
- [Google wants to end the “15 open tabs” shopping struggle](https://www.idir.ai/en/blog/google-want-to-end-the-15-open-tabs-shopping-struggle): Google presents UCP as a way to shop through assistants. This analysis of the January 2026 announcement distinguishes that promise from actual adoption.
- [Prompt Caching: A Technical Guide to LLM Efficiency](https://www.idir.ai/en/blog/prompt-caching-a-technical-guide-to-llm-efficiency): Prompt caching can reduce the cost of reusing context, but the benefit depends on repeated input and each provider’s conditions.

## Latest writing in Spanish
- [AI harness: la capa que convierte un modelo inteligente en trabajo útil](https://www.idir.ai/es/blog/ai-harness-capa-control-modelo-inteligente-trabajo-util): Un modelo de IA no es un trabajador digital por sí solo. El harness es la capa de control que le da herramientas, memoria, permisos y verificación.
- [La trampa del «vibecoding»: por qué tu app creada con IA es probablemente un colador](https://www.idir.ai/es/blog/la-trampa-del-vibecoding-por-qu-tu-app-creada-con-ia-es-probablemente-un-colador): En un experimento con diez personas de mi entorno, encontré fallos de seguridad en siete de los nueve proyectos publicados con ayuda de IA; funcionar no bastó para ser seguro.
- [Escalar n8n con modo cola y Docker: arquitectura y revisión de los ejemplos](https://www.idir.ai/es/blog/escala-n8n-como-la-gua-definitiva-del-modo-cola-y-docker): Main, workers, Redis y PostgreSQL: una guía de la arquitectura del modo cola. La configuración de runners requiere revisión antes de utilizar los ejemplos.
- [Google quiere acabar con las 15 pestañas abiertas al comprar](https://www.idir.ai/es/blog/google-quieren-acabar-con-la-lucha-de-tener-15-pestaas-abiertas-al-comprar): Google presenta UCP como una forma de comprar a través de asistentes. Este análisis del anuncio de enero de 2026 distingue esa promesa de su adopción real.
- [Prompt caching: guía técnica sobre la caché de prompts](https://www.idir.ai/es/blog/prompt-caching-gua-tcnica-para-optimizar-la-eficiencia-en-llms): La caché de prompts puede reducir el coste de reutilizar contexto, pero depende de qué partes se repiten y de las condiciones de cada proveedor.

## Media

- [Prompt&Play on Apple Podcasts](https://podcasts.apple.com/us/podcast/prompt-and-play/id1817275702): Spanish-language podcast on AI and automation.
- [Prompt&Play on Spotify](https://open.spotify.com/show/6q4QWGY41LcM3LRhtaWDvF/): Spanish-language podcast on AI and automation.