Name: AI Coding Summit 2026
Start: 2026-02-26T15:00:00.000Z
End: 2026-02-27T17:00:00.000Z

Varun

Kristiyan Velkov

Lera

Tobbe Lundberg

Alex Astrum

Stepan Suvorov

Rajni Gediya

Agnieszka Labonarska

Violina Popova

Joanna Otmianowska

Sander Vervaeke

Nick Taylor

Eddy Vinck

Daniel Sogl

Brittany Joiner

Alfonso Graziano 

Shubh

Louis Knight-Webb

Ado Kukic

Todd Schiller

Shifra Williams

John Robert

Joao Gaspar Silva

Jeroen van Wissen

David Pugh

Georgi

Oscar Pagani

Ricardo Machado

Thijs Broerse

David Barry

David Passarelli

Filip Michalowski

Charles

Cole will walk through his agentic coding workflow that allows him to delegate all coding to the agent while still keeping him in the driver's seat (no vibe coding!). The key is to create a structured approach for both the up front planning and the validation - and he'll demonstrate what that looks like. Plus, he'll show the true power of creating a system for AI coding - with a proper system, every coding mistake the LLM makes is an opportunity to not just address the bug manually, but fix the system so the bug doesn't happen again. That's what makes your coding agent more and more reliable over time!

Advanced Claude Code Techniques for 2026

Cole Medin

Nunca me he divertido tanto desarrollando software como en los últimos meses. Y voy a mostrarte por qué. En esta charla, compartiré mis flujos de trabajo favoritos para construir una aplicación web usando AI con Cursor. Al final de la experiencia, tendremos una aplicación del mundo real que construimos usando AI junto con una profunda experiencia en desarrollo web full-stack. Y te irás con ideas sobre cómo integrar tu experiencia web con el poder de AI para construir excelentes experiencias de usuario para la web.

Cómo construyo aplicaciones web en 2026

Kent C. Dodds discusses building web applications in 2026, highlighting the importance of user experience and challenges in offline video downloads. Recent AI advancements in the Epic Workshop app enable offline video downloads with features like bulk downloads and quality control. Efficient AI agent management using agents like Cursor transforms workflow. Leveraging Laravel and Ruby on Rails, Ask Mode enhances project understanding by communicating with AI agents. Emphasizing clear task scoping, iterative planning, and agent feedback improves project development. Utilizing MCP tools and agent workflow, including BugBot for code review, optimizes app development. Closing the agent loop and emphasizing architecture's role in utilizing agent input context are crucial. The talk invites developers to become experienced AI agent managers and join the mailing list for updates.

I've never had so much fun developing software than I have in the last few months. And I'm going to show you why. In this talk, I'll share my favorite workflows for building a web application using AI with Cursor. By the end of the experience, we'll have a real-world application that we built using AI along with deep full-stack web expertise. And you'll leave with ideas of ways to integrate your web expertise with the power of AI to build excellent user experiences for the web.

How I Build Web Applications in 2026

Kent C. Dodds

Las herramientas de codificación de IA funcionan mejor cuando pueden planificar, actuar e iterar, no solo responder. Esta charla cubre patrones para construir flujos de trabajo autónomos a lo largo del flujo de trabajo del desarrollador y cómo puedes sacar el máximo provecho de Claude Code.

Agente por Defecto: Repensando los Flujos de Trabajo de Desarrolladores con Claude Code

The Talk delves into agent decoding in Cloud Code, emphasizing the importance of planning, acting, and verifying changes systematically. It explores maximizing efficiency through Cloud Code acting as an agent, reasoning, reading, writing code, and using tools for successful outcomes. The discussion covers enhancing task planning, utilizing subagents for specialized assistance, and optimizing tool use with MCP integration. It introduces Cloud Skills, plugins for team collaboration, and efficient tool utilization with parallel workflows. The concept of non-interactive agents, guided autonomy, and steps for experiencing agent decoding are also highlighted.

AI coding tools work best when they can plan, act, and iterate - not just respond. This talk covers patterns for building autonomous workflows across the developer workflow and how you can get the most out of Claude Code.

Agentic by Default: Rethinking Developer Workflows with Claude Code

Las revisiones de código AI a menudo se sienten como lanzar una moneda. Cara - obtienes interminables críticas sobre el estilo. Cruz - encuentra problemas reales. Y incluso cuando la señal es buena, ¿realmente estamos obteniendo beneficios de calidad y productividad?El secreto no es el AI, es cómo lo estamos implementando. Esta charla explora lo que realmente impulsa el valor en las revisiones de código asistidas por AI: integración contextual rica, especialización escalable e impacto medible en la productividad del desarrollador. Aprende a transformar las revisiones AI de una casilla de verificación de cumplimiento en una ventaja estratégica para tu organización de ingeniería.

Tus Revisiones de Código AI Están Perdiendo el Punto (Y Cómo Solucionarlo)

Ishai, CTO at LearnrB, discusses the evolution and impact of AI code reviews in software development. The transition from experimentation to maturity with autonomous AI agents in code reviews is highlighted. Challenges in AI-driven code creation and review processes, as well as the benefits of AI code reviews for productivity and efficiency, are discussed. Common mistakes and challenges in AI code review adoption, along with the importance of adapting processes for AI advancements, are emphasized. The importance of enhancing AI code review processes, measuring their impact, and leveraging AI for productivity metrics is also explored.

AI code reviews often feel like tossing a coin. Heads - you get endless nitpicks about style. Tails - it finds real issues. And even when the signal is good, are we really getting quality and productivity benefits?The secret isn't the AI, it's how we're deploying it. This talk explores what actually drives value in AI-assisted code reviews: rich contextual integration, scalable specialization, and measurable impact on developer productivity. Learn to transform AI reviews from a compliance checkbox into a strategic advantage for your engineering organization.

Your AI Code Reviews Are Missing the Point (And How to Fix It)

Yishai Beeri

Join this interactive 2-hour workshop to learn how to create your own AI co-worker using the Codex SDK, inspired by Rudrank’s work on a GitHub/Slack bot he created to automate code reviews, bug fixes, and feature additions to overcome the opposite timezone differences at work. We will explore how to shift from manual "vibe coding" to take a managerial role, where AI handles the heavy lifting while you focus on high-level planning and validation. We will cover iterative prompting, and integration with GitHub and Slack for automation with the rest of the team.By the end, you will deploy a functional bot to your repository, ready to assist in your projects and even free up time for personal goals like finishing that games backlog!This session is ideal for developers who are curious and looking to build an "army" of AI helpers.Table of Contents1. Introduction &amp; Why AI Co-Workers?: - Discuss the motivation behind AI-driven development, going from just AI-assisted a year ago. How to take advantage of the ChatGPT plan and use the Codex SDK to program codex directly to create structured agents.- Overview of Codex SDK's strengths in code generation, audits, and multi-step reasoning, with real examples from building my AI coworker. 2. Architecture:- Explore components like command parsing, session persistence (24-hour TTL), and GitHub integration via Octokit. - Cover agentic patterns: planning with high-precision models (e.g., Codex xHigh), implementation with efficient ones Codex Medium, and error handling for long-running tasks. 3. Building The Bot: - Set up the environment (clone example repo, configure .env with API keys), create a GitHub App, implement core features (e.g., @bot to fix the bugs)- Test with sample commands. Include live demos of code audits, branch/PR creation, and polling for stop commands. Participants will fork and customize their own bot.  4. Future Directions: - Parallel sub-tasks, MCPs, skills- Q&amp;A

Building AI Co-Worker with Codex SDK

Rudrank Riyam

Esta charla explorará la transformación fundamental en curso del rol de los desarrolladores de software. Proporcionará orientación sobre cómo tener éxito en este nuevo entorno que evoluciona rápidamente, dónde enfocar tu desarrollo personal, y explorará los flujos de trabajo y técnicas que realmente funcionan, y no son solo publicidad o impulsadas por ventas.

Tu Nuevo Trabajo

Ken Wheeler discusses the evolving software landscape, emphasizing the shift towards higher efficiency and less typing in job roles. The evolution of coding practices from manual formatting to automated tools like Prettier is compared to advancements in construction tools for efficiency. The use of AI for work efficiency and financial benefits, alongside considerations about the future of software engineers in light of AI advancements, is highlighted. Adapting to industry shifts, embracing new technologies like React, and leveraging early AI adoption for career growth are key points. Transitioning to higher-level roles by combining domain knowledge with AI for decision-making and seizing new opportunities is emphasized.

This talk will explore the ongoing fundamental transformation of the role of software developers. It will provide guidance on succeeding in this new rapidly evolving environment, where to focus your personal development, and explore the workflows and techniques that actually work, and aren’t hype or sales driven.

Your New Job

Ken Wheeler

In this interactive workshop, you’ll build a working internal browser extension you and your team can use to share context across apps you use daily. We’ll start with a practical, real-world example: a web clipper that you’ll customize to send content from GitHub to your team’s communication tool (Slack, Discord, or Teams). You’ll learn how to generate extensions with natural language, modify forms and logic, connect to apps like Slack and GitHub, and deploy what you build to your team. By the end, you’ll have a functioning automation running in your browser and the skills to create more. This workshop is ideal for developers and tech leads who want to quickly build and deploy secure, internal extensions without writing full browser extensions and handling infrastructure.

Build Your Dream Extensions with AI

In this hands-on session, we'll dive straight into the practical mechanics of working with Google DeepMind’s latest research and models. We'll explore Google AI Studio, Antigravity, and the Gemini APIs, focusing on how to seamlessly integrate these capabilities into your own projects. If you're looking for the modern developer's equivalent of Python's "import antigravity" (pun intended!), this is it.We'll prioritize a hands-on, hacker-friendly approach so you can walk away with a solid intuition for how these tools work under the hood. You'll leave fully empowered to use these models in your own architecture, with the exact steps needed to get up and running with Gemini for code generation, analysis, and everyday tinkering as quickly as possible.

Building with Gemini: AI Studio, Antigravity, and Google DeepMind's Latest Models

Paige Bailey

Los desarrolladores ahora crean aplicaciones completas simplemente conversando con herramientas de AI. Cada vez más startups dependen predominantemente del código generado por AI. Pero lo que inicialmente emocionó a los desarrolladores rápidamente lleva a la frustración: código genérico, mala arquitectura y vulnerabilidades de seguridad.En esta charla, demuestro cómo el Spec Driven Development (SDD) soluciona estos problemas. En lugar de codificar primero y documentar después, SDD comienza con especificaciones precisas como la "única fuente de verdad". Varios herramientas y CLIs hacen que esta metodología sea prácticamente aplicable hoy en día con un enfoque estructurado de 4 fases: Especificar, Planificar, Tareas, Implementar.Usando un proyecto real, muestro cómo SDD funciona con herramientas modernas de codificación AI. Comparto ideas honestas de la experiencia práctica: dónde SDD sobresale, dónde falla y si la sobrecarga de especificación realmente conduce a mejores resultados.

Spec Driven Development: El Fin de la Programación por Intuición

Welcome to spec-driven development talk by Daniel Sogel. AI agents struggle with benchmarks and real-world tasks. Developers spend extra time fixing code generated by AI agents. Security issues in AI-generated code, limitations with context management, multi-file coordination, silent failures, architecture decisions, and enterprise context challenges addressed by Spectre from development. AI agents removing safety checks, generating code with fake data, silent failures worse than crashes, spectrum development benefits like complete context upfront, structured reasoning, validation checkpoints, and autonomous implementation without manual code review. Act with AI agents like pair programmers, combine spectrum, behavior-driven, and test-driven development. Spec-driven development adoption in the industry, using Keyro IDE by AWS and popular tools like SpecKit and OpenSpec for different project types. Define tech stack, API structure in planning, task phase implementation, handover to AI agents, and choosing the right spec-driven development tools.

Developers now create entire applications by simply chatting with AI tools. More and more startups rely on predominantly AI-generated code. But what initially thrilled developers quickly leads to frustration: generic code, poor architecture, and security vulnerabilities.In this talk, I demonstrate how Spec Driven Development (SDD) solves these problems. Instead of coding first and documenting later, SDD begins with precise specifications as the "single source of truth." Various tools and CLIs make this methodology practically applicable today with a structured 4-phase approach: Specify, Plan, Tasks, Implement.Using a real project, I show how SDD works with modern AI coding tools. I share honest insights from hands-on experience: where SDD excels, where it fails, and whether the specification overhead actually leads to better results.

Spec Driven Development: The End of Vibe Coding

Modern AI coding agents don’t just autocomplete. They execute, connect, and automate. With the wrong settings, they can access more of your system than you intended. In this session, I’ll demonstrate how agent autonomy, MCP integrations, and Skills can lead to unexpected security risks. Through examples using Cursor.ai, we’ll see how seemingly harmless configuration choices stack into real exposure and how to set secure defaults for safe, high-velocity “Vibe Coding.”

Hands-On Guide to Secure AI-Driven Coding

Gil Friedman

A medida que 2026 trae un cambio de la paridad de características a conjuntos de características diferenciadas, Antigravity ofrece un enfoque distinto para construir software. Esta sesión profundiza en la mentalidad de "get stuff done" de la plataforma y las capacidades únicas del ecosistema, incluyendo el Chrome Agent y las opciones de modelo versátiles. Aprende cómo pasar de PRD a una solución de grado de producción, utilizando el browser agent para investigación y pruebas, gestionando tareas de múltiples agentes y utilizando herramientas simples para personalizar tu entorno. En última instancia, esto te capacita para convertir ideas en prototipos con rapidez y precisión.

Beyond Vibe Coding: Flujos de Trabajo Diferenciados con Antigravity

Alex and Rodi discuss antigravity, a powerful coding agent that unifies Google tools. The agent manager orchestrates multiple agents, providing a higher-level view than just code. Features include an AI editor, agent-controlled Chrome browser, and best practices for agent autonomy and usage. AI steering takes time before feeling intuitive. Anti-gravity offers fast and planning modes for building efficiently. The planning mode allows fine-tuning and execution of complex tasks. Playground validates ideas before full projects. The playground in AntiGravity validates project ideas. Image generation, Nano Banana Pro, and Gemini 3 enhance design output. Skills support deep context tasks. Asynchronous nature enables efficient project development. Browser agent serves as a versatile tool for bug fixing. Manual control and inspiration amidst automation. Transition to agent manager for enhanced visibility and decision-making. Encouragement to explore diverse agent-led experiences and updates for workflow improvement.

As 2026 brings a shift from feature parity to differentiated feature sets, Antigravity offers a distinct approach to building software. This session dives into the platform's "get stuff done" mindset and unique ecosystem capabilities, including the Chrome Agent, and versatile model choices. Learn how to go from PRD to a production-grade solution, using the browser agent for research and testing, managing multi-agent tasks, and using simple tools to customize your environment. Ultimately, this empowers you to turn ideas into prototypes with speed and precision.

Beyond Vibe Coding: Differentiated Workflows with Antigravity

Rody Davis

Como sabrás, el bucle "Ralph Wiggum" ha estado generando bastante revuelo en línea entre los desarrolladores. Es un concepto que envuelve a un agente de codificación de IA en bucles de iteración persistentes que siguen ejecutándose hasta que las tareas están realmente completas. Esta charla profundiza en los conceptos y principios básicos, los escollos comunes y la aplicación práctica del método.

El Método Ralph Wiggum Explicado

Today, exploring Ralph Wiggum AI method's simplicity and persistence. Ralph loops ensure eventual consistency in problem-solving. Challenges in AI chats revolve around context management, emphasizing the importance of effective loop iterations. Basic example of loop iteration and backpressure in Ralph. Using skip permissions can be risky. Running Cloud in a sandbox mode for safety measures. Ralph Wiggum AI Method: Autonomous coding with thorough spec creation. Careful planning needed. Upsides include autonomous operation and clear task handling.

As you might know, the "Ralph Wiggum" loop has been generating quite a bit of hype online amongst developers. It's a concept that wraps an AI coding agent in persistent iteration loops that keep running until tasks are actually complete. This talk goes into the core concepts and principles, common pitfalls, and practical application of the method.

The Ralph Wiggum Method Explained

La generación de código escala. La planificación y revisión no lo hacen, limitando la rapidez con la que los equipos nativos de IA pueden entregar. Esta charla traza la evolución de las herramientas para desarrolladores hasta la actual explosión de agentes de codificación, y argumenta por qué la mayor oportunidad ahora es ayudar a los ingenieros a planificar y revisar más rápido.

Por Qué La Ingeniería de Software Se Está Convirtiendo: Planificar y Revisar

Louis from Vibe Kanban discusses optimizing workflows for software engineers with coding agents, focusing on planning and review. Leveraging AI tools like GitHub Copilot and ChatGPT can enhance productivity by reallocating time effectively. Reflecting on time saved by AI, the focus is on optimizing planning and reviewing to boost productivity. Emphasizing detailed planning processes and effective use of AI coding agents to ensure successful outcomes. Simplifying reviewing processes to save time and enhance productivity through efficient code review. Leveraging Codex and Claude coding agents for efficient code review to optimize time and workflow. Minimizing time spent on code review by providing feedback to coding agents within the editor, working on multiple tasks in parallel, and automating tasks requested by AI for enhanced productivity.

Code generation scales. Planning and review don't, limiting how fast AI-native teams can ship. This talk traces the evolution of developer tooling to the current explosion of coding agents, and makes the case for why the biggest opportunity now is helping engineers plan and review faster.

Why Software Engineering Is Becoming: Plan and Review

Los LLMs necesitan acceso al contexto relevante para tener éxito. Un elemento clave de la Ingeniería de Contexto que podemos usar para ayudar a los agentes a recordar es la Memoria.Adentrémonos en el mundo de la gestión de memoria para agentes. Discutiremos la diferencia entre la memoria a corto y largo plazo, cómo gestionar la memoria, y la importancia de prevenir el envenenamiento y el choque de contextos utilizando técnicas como el recorte y la resumización.

Cómo Ayudar a los Agentes a Recordar

Carly discusses memory, context in LLMs, and managing information for effective results in agentic systems. LLMs hallucinate due to knowledge gaps, biases, overfitting, and incentives. Importance of context in minimizing hallucinations and focus on memory storage for decision-making. Short term memory acts like RAM, while long term memory stores knowledge and experiences. Efficient memory storage using disk, file systems, and semantic memory. Emphasis on managing message context, retrieval, and concatenation for LLM input. Attention to context length, information evaluation, and memory optimization for accurate responses.

LLMs need access to the relevant context to succeed. One key element of Context Engineering that we can use to help agents remember is Memory.Let's dive into the world of memory management for agents. We'll discuss the difference between short and long term memory, how to manage memory, and the importance of preventing context poisoning and clash using techniques such as trimming and summarization.

How to Help Agents Remember

Carly Richmond

Aprende un patrón arquitectónico para automatizar el "trabajo de pegamento" y construir flujos de trabajo autoejecutables. Esta charla demostrará cómo usar "planos de colaboración" con agentes de IA ligeros y alojados localmente como una capa de automatización invisible para eliminar tareas manuales como exportar datos y conectar herramientas. A través de un ejemplo de un analista de acciones de IA autoalojado y conocimientos de mis proyectos anteriores para la automatización de CI/CD, aprenderás cómo pasar de ser el ejecutor de tareas repetitivas a ser el arquitecto de sistemas autónomos.

Construye la Capa de Automatización de tu Equipo: El Patrón de Agente de Pegamento de IA

Speaker specializing in technical architecture discusses the impact of interruptions on deep focus, productivity, and software development challenges in global organizations. Content adaptation to multiple languages required manual glyph corrections and a structured collaboration plane. Transitioning to automated systems raises the question of why orchestration remains manual, with AI as a glue agent. Strategic stock analysis involves N8n workflows, specialized agents, and structured contracts for consistency. Workflow architecture includes JSON contracts, specialized agents, and structured HTML emails for stock analysis. Workflow reliability focuses on independence, idempotence, and robust frameworks for adaptability. Automation and AI integration emphasize process streamlining and human intervention in decision-making. Intelligent orchestration systems involve encoding intent by humans and AI turning it into action, focusing on clear workflows and architecting for control.

Learn an architectural pattern to automate "glue work" and build self-executing workflows. This talk will demonstrate how to use "collaboration planes" with lightweight, locally-hosted AI agents as an invisible automation layer to eliminate manual tasks like exporting data and bridging tools. Through an example of a self-hosted AI stock analyst and insights from my previous projects for CI/CD automation, you'll learn how to shift from being the executor of repetitive tasks to the architect of autonomous systems.

Build Your Team's Automation Layer: The AI Glue Agent Pattern

Esta presentación muestra cómo el Domain-Driven Design (DDD) convierte la codificación agéntica de “prompt spaghetti” en un enfoque de ingeniería mantenible. En lugar de un único prompt gigante y herramientas ad-hoc, propone contextos delimitados para las responsabilidades de los agentes, un lenguaje ubicuo para entradas/salidas de herramientas consistentes, y un mapa de contexto que hace explícitas las integraciones. Los asistentes aprenderán cómo diseñar contratos de herramientas estables (puertos/adaptadores), agregar puertas de CI/evaluación que mantengan a los agentes honestos, e implementar trazabilidad y reproducción para que las acciones de los agentes sigan siendo auditables y depurables. El resultado: sistemas agénticos que escalan a bases de código reales y equipos sin colapsar en el caos.

De Prompt Spaghetti a Contextos Delimitados: DDD para Bases de Código Agénticas

Nikita, AI Architect at Siemens, discusses challenges in JNTX system development, emphasizing domain-driven design for scalability and stability. Clear responsibilities, bounded contexts, and structured schemas are crucial for reducing complexity. The anti-corruption layer and context maps play vital roles in integration architecture. Key takeaways include the importance of contracts, firewalls, and CI gates for system development.

This presentation shows how Domain-Driven Design (DDD) turns agentic coding from “prompt spaghetti” into a maintainable engineering approach. Instead of one giant prompt and ad-hoc tools, it proposes bounded contexts for agent responsibilities, a ubiquitous language for consistent tool inputs/outputs, and a context map that makes integrations explicit. Attendees will learn how to design stable tool contracts (ports/adapters), add CI/evaluation gates that keep agents honest, and implement traceability and replay so agent actions remain auditable and debuggable. The result: agentic systems that scale to real codebases and teams without collapsing into chaos.

From Prompt Spaghetti to Bounded Contexts: DDD for Agentic Codebases

Nikita Golovko

La IA acelera el desarrollo, pero también introduce vulnerabilidades silenciosas. Esta charla expone los riesgos de seguridad ocultos del código generado por IA no revisado y muestra formas prácticas de proteger tus aplicaciones sin ralentizar tu flujo de trabajo.Tabla de contenido: - Cómo los LLMs generan código y por qué puede ser peligroso- Ejemplos del mundo real de vulnerabilidades introducidas por IA- Un flujo de trabajo práctico y seguro para el código generado por IA- Preguntas y respuestas

Errores de seguridad en el código generado por IA: Qué sucede cuando los desarrolladores omiten la revisión

Introduction to AI security pitfalls, emphasizing the importance of understanding AI promises and risks, responsibility in verifying AI-generated code, the necessity of code verification and thorough review for secure deployment, ensuring code validation and sanitization for secure applications, risks of unsecure authentication logic and broken authorization due to AI usage, the importance of avoiding exposing secrets in code by trusting AI blindly, learning resources on cloud security and platforms like Flare Learning and TryHackMe, and the significance of creating AI rules, testing, and following best practices for code security.

AI speeds up development — but it also introduces silent vulnerabilities. This talk exposes the hidden security risks of unreviewed AI-generated code and shows practical ways to protect your applications without slowing down your workflow.Table of Content: - How LLMs Generate Code &amp; Why It Can Be Dangerous- Real-World Examples of AI-Introduced Vulnerabilities- A Practical, Secure Workflow for AI-Generated Code- Q&amp;A

Security Pitfalls in AI-Generated Code: What Happens When Developers Skip Review

Join us for an interactive workshop where you'll train your first neural network using the classic MNIST handwritten digit dataset. In this hands-on session, participants will use OpenShift AI 3.x Workbenches to build and train a PyTorch model that recognizes handwritten digits from images.This beginner-friendly workshop provides a practical introduction to machine learning workflows in an enterprise-grade AI platform. Working directly in Jupyter notebooks, attendees will clone a ready-to-use GitHub repository, explore the MNIST dataset, and walk through the complete model training pipeline—from data preprocessing to model evaluation. The session uses CPU-optimized configurations, making it accessible without requiring specialized GPU resources.Whether you're new to AI/ML or curious about how the models you use actually function, this workshop offers practical, immediately applicable experience.Takeaways: Participants will leave with a working neural network, a complete training notebook they can reference, and foundational knowledge of deploying and utilizing ML workloads on OpenShift AI.

"Hello World" on OpenShift AI: Training Your First Neural Network

Legare Kerrison

Sawyer Bowerman

Ready to bring conversational AI to life? In this hands-on workshop, we'll walk through deploying a real-time voice agent using Render's cloud platform. You'll work directly with LiveKit's voice agent template to create an interactive AI assistant that can listen, think, and respond naturally through voice.You'll discover:<ul><li>How to set up and configure a LiveKit voice agent using Render's deployment template</li><li>The architecture behind real-time voice AI systems and how different components work together</li><li>Best practices for managing environment variables, API keys, and cloud service integrations</li><li>Techniques for testing and debugging your deployed voice agent in production</li><li>Strategies for monitoring performance and scaling your AI application as usage grows</li></ul>By the end of this workshop, you'll have a fully functional voice agent running in the cloud, ready to handle real conversations. You'll gain practical experience with modern AI deployment tools and understand the complete pipeline from template to production.This workshop is ideal for developers interested in conversational AI, engineers looking to add voice capabilities to their applications, and anyone curious about deploying real-time AI systems without getting lost in infrastructure complexity.

Ship a Production Voice Agent: A Hands-on Workshop

Un agente se siente como un superpoder, pero ¿qué sucede cuando escalamos a diez? En esta charla, mostraré por qué los juegos RTS de nuestra infancia son un modelo mental sorprendentemente efectivo para gestionar agentes, construidos para la conciencia situacional, comandos rápidos y la coordinación de muchas unidades asincrónicas. Saltaremos a una demostración en vivo de AgentCraft, el orquestador de agentes al estilo RTS que da vida a los agentes, permitiéndote generar, inspeccionar y dirigir múltiples agentes en tiempo real. Te irás con una perspectiva práctica sobre los desafíos de la orquestación de agentes, cómo aprovechar una vida de memoria muscular para dominarlos, y cómo un poco de fantasía puede llegar muy lejos.

AgentCraft: Poniendo el Orc en la Orquestación de Agentes

Welcome to AgentCraft with Iddo Salomon, AI lead. Scale up agents for efficient work but consider cognitive overload. Agent orchestrators are crucial for task management and real-time reactions. AgentCraft is a tool inspired by real-time strategy games to manage agents effectively, with features like heroes and buildings. It visualizes repository files through terrain projection, simplifies code inspection, and enhances task management with cloud code support. Task delivery is streamlined with code merging, quests, and hero summoning for project improvement.

One agent feels like a superpower, but what happens when we scale to ten? In this talk, I’ll show why the RTS games of our childhood are a surprisingly effective mental model for managing agents, built for situational awareness, rapid commands, and coordinating many asynchronous units.We’ll jump into a live demo of AgentCraft, the RTS-style agent orchestrator that makes agents come alive, allowing you to spawn, inspect, and steer multiple agents in real time. You’ll leave with a practical take on agent orchestration challenges, how to leverage a lifetime of muscle-memory to tame them, and how a little whimsy goes a long way.

AgentCraft: Putting the Orc in Agent Orchestration

Ido Salomon

MCP (Model Context Protocol) se ha convertido rápidamente en un estándar para extender la funcionalidad de LLM en la codificación asistida por IA y la codificación de ambiente, y los desarrolladores son los primeros en adoptar este poderoso nuevo método de integración. Pero los MCPs también pueden crear riesgos significativos no solo para el código que estás escribiendo, sino también para tu estación de trabajo, credenciales e infraestructura de la empresa. En esta charla relámpago, revisaré esos riesgos usando ejemplos del mundo real y explicaré cómo evitarlos.

MCPs: Cómo Evitar Escollos de Seguridad

Gil Friedman from Backslash Security discussed MCP security, focusing on the Model Context Protocol and AI assistant security risks like GitHub token exposure in auto run mode. Security measures for MCP usage include setting autorun to ask every time, postmarking MCP incidents, handling malicious servers, safeguarding sensitive information, reviewing and disabling unused tools, requiring permission before triggering actions, removing MCP from environments, and revoking API keys when necessary.

MCP (Model Context Protocol) has quickly become a standard way to extend LLM functionality in AI-assisted and vibe coding, and developers are early adopters of this powerful new integration method. But MCPs can also create significant risks not just to the code you're writing, but to your workstation, credentials, and company infrastructure. In this lightning talk I will review  those risks using real-world examples, and explain how to avoid them.

MCPs: How to Avoid Security Pitfalls

Los agentes de codificación han cruzado el umbral de resolver problemas no triviales junto a los ingenieros, así como de funcionar de manera autónoma en muchos aspectos. Su habilidad más sorprendente ha sido su efectividad en tareas no relacionadas con la codificación también, en el momento en que el universo se representa en un sistema de archivos o es accesible desde entornos tipo unix. Los agentes pronto estarán escribiendo, accediendo y potenciando software — y esta charla cubre los divertidos desafíos de diseñar entornos de desarrollo aislados a través de estos paradigmas, permitiendo a los agentes operar completamente dentro de entornos aislados o usar estos entornos como herramientas programables y efímeras.

Diseñando Entornos de Desarrollo Aislados para Agentes de Codificación

A demo of an animated goose following the cursor in a unique environment with real-time updates. Advanced models for long-running agents performing complex tasks. Exploring sandbox environments for agent coding, focusing on isolation, containers, networking, and persistence. Challenges in sandbox orchestration, including handling heavy processing, managing sessions, and security concerns. Importance of agent control, Micro VMs for isolation, and networking in sandbox environments. Strategies for efficient sandbox operations with pre-built images, persistent volumes, and warm pools. Utilizing sandbox primitives for simple usage with focus on isolation, networking, and persistence.

Coding agents have crossed the threshold of solving non-trivial problems alongside engineers, as well as running autonomously in many ways. Their most surprising ability has been in their effectiveness at non-coding tasks too the moment universe is represented on a filesystem or accessible from unix-like environments. Agents will soon be writing, accessing and powering software — and this talk covers the fun challenges of designing sandboxed dev environments across these paradigms, enabling agents to run entirely within sandboxes or using sandboxes as programmable, ephemeral tools.

Designing Sandboxed Dev Environments for Coding Agents

Naresh Ramesh

A medida que las bases de código crecen, los asistentes de codificación de AI tienen dificultades. Las ventanas de contexto se desbordan, los agentes pierden el seguimiento de las dependencias y la búsqueda de texto simple no logra capturar las relaciones semánticas que definen el software real. Esta masterclass explora tres patrones arquitectónicos probados que permiten a los agentes de AI trabajar eficazmente con bases de código a escala de producción: inteligencia de código semántico a través de la integración del Language Server Protocol, habilidades especializadas de agentes mediante el agrupamiento de contexto y herramientas, y delegación de subagentes para una gestión eficiente del contexto. A través de demostraciones en vivo en un proyecto popular de código abierto como ShadCN, verás cómo estas técnicas abordan la complejidad del software del mundo real: refactorizaciones de múltiples archivos, cambios entre módulos y seguimiento de dependencias que abrumarían los enfoques tradicionales.Te irás con estrategias prácticas y agnósticas al producto para construir o mejorar agentes de AI que puedan manejar grandes bases de código con precisión y eficiencia. Examinaremos por qué la comprensión semántica supera la búsqueda de texto, cómo diseñar habilidades de agentes enfocadas que mejoren la finalización de tareas y cómo las arquitecturas de subagentes en paralelo previenen el agotamiento de las ventanas de contexto. Ya sea que estés construyendo herramientas de desarrollo de AI, arquitecturando sistemas de múltiples agentes o contribuyendo a código abierto, estos patrones te ayudarán a cerrar la brecha entre las demostraciones de juguete y la asistencia de AI de grado de producción.

Escalando Agentes de AI para Bases de Código de Producción: Patrones para Precisión y Eficiencia

Discussing top AI coding best practices in 2026, including semantic understanding and context window management. Exploring the role of Language Server Protocol (LSP) in code refactoring. Efficient code renaming using LSP and code intelligence in Kiro. Impact of not using LSP on code renaming efficiency. Manual approaches without LSP significantly impact efficiency and resource consumption. Context window usage doubles without LSP, affecting code handling. Utilizing subagents for specialized tasks enhances codebase security. Spec-driven development and detailed design documents for efficient agent alignment.

As codebases grow, AI coding assistants struggle. Context windows overflow, agents lose track of dependencies, and simple text search fails to capture the semantic relationships that define real software. This talk explores three proven architectural patterns that enable AI agents to work effectively with production-scale codebases: semantic code intelligence through Language Server Protocol integration, specialized agent skills via context and tool bundling, and subagent delegation for efficient context management. Through live demonstrations on a popular open-source project like ShadCN, you'll see these techniques tackle the complexity of real-world software—multi-file refactorings,cross-module changes, and dependency tracking that would overwhelm traditional approaches.You'll leave with practical, product-agnostic strategies for building or enhancing AI agents that can handle large codebases with accuracy and efficiency. We'll examine why semantic understanding outperforms text search, how to design focused agent skills that improve task completion, and how parallel subagent architectures prevent context window exhaustion. Whether you're building AI developer tools, architecting multi-agent systems, or contributing to open source, these patterns will help you bridge the gap between toy demos and production-grade AI assistance.

Scaling AI Agents for Production Codebases: Patterns for Accuracy and Efficiency

Saurabh Dahal

Un curso intensivo práctico para desarrolladores que quieren evitar semanas de investigación y pasar directamente a la construcción. Cubriremos la terminología esencial, las tendencias actuales de la industria y el panorama real de servicios y bibliotecas. Aprenderás cuándo usar qué, y por qué algunas elecciones importan más que otras. Compartiré lecciones aprendidas al construir un agente de voz en producción para más de 10,000 conductores de camiones holandeses, incluidos los problemas de los que nadie habla. Al final, tendrás todo lo que necesitas para construir tu primer Agente de Voz funcional.

Curso Intensivo de Agentes de Voz

Stepan Suvorov, CTO at Roadsoft, discusses the rise of Voice Agents, highlighting market growth, improved models for speech to text and voice generation, reduced latency, and cost-effectiveness. Key components include speech-to-text, LLM analysis, text-to-speech conversion, orchestration, and addressing latency. Delving into essential aspects like voice activity detection, interruption handling, and model selection criteria for Voice Agents. Exploring parameters such as latency, quality, pricing, language support, and voice cloning options. Comparing DIY and managed solutions for Voice Agents, emphasizing flexibility and cost-effectiveness. Discussing cost implications between managed and self-hosted solutions, considering usage volumes and scale. Addressing infrastructure challenges, testing approaches, WAPI limits, model selection, and future plans for voice agents.

A practical crash course for developers who want to skip weeks of research and get straight to building. We'll cover the essential terminology, current industry trends, and the real landscape of services and libraries. You'll learn when to use what - and why some choices matter more than others. I'll share lessons learned from building a production voice agent for 10,000+ Dutch truck drivers, including the problems nobody talks about. By the end, you'll have everything you need to build your first working Voice Agent.

Voice Agents Crash Course

Esta masterclass desafía la dependencia excesiva en la tasa de éxito e introduce un marco más práctico y consciente de la seguridad para evaluar agentes de AI. Basándonos en escenarios de implementación reales, exploramos métricas que capturan mejor la utilidad y fiabilidad del agente, incluyendo el tiempo de finalización, la eficiencia de llamadas a herramientas, la tasa de recuperación de errores y el costo por tarea exitosa.La sesión mostrará cómo los equipos optimizan involuntariamente las métricas incorrectas, llevando a modos de falla ocultos como reintentos excesivos, uso ineficiente de herramientas, errores silenciosos y costos operativos en aumento. Conectaremos estos problemas con preocupaciones más amplias de seguridad en AI, destacando cómo las prácticas de evaluación deficientes pueden crear una confianza engañosa en el comportamiento del agente.Los asistentes se irán con estrategias de evaluación concretas, una comprensión más clara de los compromisos en el diseño de agentes y una guía práctica sobre cómo construir métricas que reflejen el rendimiento en el mundo real, no solo victorias en benchmarks.

Métricas Que Realmente Importan: Evaluando Agentes de AI Más Allá de la Tasa de Éxito

John Robert, Lead Data and AI Platform Engineer, discusses evaluating AI agents beyond success rate. Talks about failures of AI projects, the uniqueness of AI agents, and the evaluation process. Creating a Framework for AI Agent Evaluation: Four Evaluation Categories - performance, business, safety, and cost. Metrics creation to enhance AI projects. Includes task completion rates, reasoning quality, accuracy evaluation, tool execution, response time, and recovery time checks. Examining KPIs, Safety, Security, and Cost in AI Agent Projects: Consider ROI, user satisfaction, time-saving, code review, adoption, engagement, risk, regulations, unauthorized actions, and cost efficiency. Checking Infrastructure and Security Metrics: Evaluate project reliability, online presence, errors, resources, prompt injection, data leakages, authorized actions, tool usage.

This talk challenges the overreliance on success rate and introduces a more practical and safety-aware framework for evaluating AI agents. Drawing from real deployment scenarios, we explore metrics that better capture agent usefulness and reliability, including time-to-completion, tool-call efficiency, error recovery rate, and cost per successful task.The session will show how teams unintentionally optimize the wrong metrics, leading to hidden failure modes such as excessive retries, inefficient tool usage, silent errors, and escalating operational costs. We will connect these issues to broader AI safety concerns, highlighting how poor evaluation practices can create misleading confidence in agent behavior.Attendees will leave with concrete evaluation strategies, a clearer understanding of trade-offs in agent design, and practical guidance on building metrics that reflect real-world performance not just benchmark wins.

Metrics That Actually Matter: Evaluating AI Agents Beyond Success Rate

En la IA generativa, los modelos de lenguaje grandes (LLMs) más grandes a menudo dominan los titulares, aclamados como las mejores soluciones para las tareas más complejas y diversas. Si bien ciertamente tienen su lugar, ¿son la mejor opción para cada caso de uso empresarial?Los modelos de lenguaje más pequeños están ganando terreno por su capacidad para ofrecer un alto rendimiento con menores costos y requisitos de recursos. Estos modelos son más rápidos, más fáciles de ajustar y mejor adaptados a las necesidades comerciales específicas, lo que los convierte en una alternativa atractiva para muchas organizaciones. En esta sesión, vamos a:-Explorar la estructura técnica y el contenido de los LLMs.-Discutir cómo los modelos más pequeños y diseñados para un propósito específico pueden ser más eficientes y efectivos para tareas empresariales, incluyendo cómo las técnicas de optimización de modelos pueden mejorar aún más el rendimiento.-Demostrar cómo los LLMs más pequeños pueden proporcionar soluciones más rápidas y rentables mientras siguen cumpliendo con las demandas de casos de uso especializados.

Cuando Menos Es Más: Una Visión Técnica de los LLMs y la Fuerza de los Modelos Más Pequeños

Lugari Karasen from Red Hat AI discusses the technical aspects of large language models, including data collection, tokenization, and neural network internals. Model training involves converting human language to mathematical representations and adjusting parameters iteratively in a complex environment. Inference engines like VLLM aid in deploying models for rapid data processing. Optimizing model size for efficiency without sacrificing accuracy is crucial, with quantization reducing model size while maintaining precision. Local deployment offers privacy and control, and smaller purpose-driven models can enhance workflows and experimentation.

In generative AI, the largest large language models (LLMs) often dominate the headlines, hailed as the best solutions for the most complex and diverse tasks. While they certainly have their place, are they the best option for every enterprise use case?Smaller language models are gaining traction for their ability to deliver high performance with lower cost and resource requirements. These models are quicker, easier to fine-tune, and better suited for targeted business needs, making them an attractive alternative for many organizations. In this session, we will:-Explore the technical structure and content of LLMs.-Discuss how smaller, purpose-built models can be more efficient and effective for enterprise tasks, including how model optimization techniques can boost performance even more.-Demonstrate how smaller LLMs can provide faster, more cost-effective solutions while still meeting the demands of specialized use cases.

When Less Is More: A Technical Overview of LLMs and the Strength of Smaller Models

Los flujos de trabajo de "vibe coding" convencionales no fueron diseñados para extensiones de navegador. Las extensiones tienen arquitecturas distribuidas que abarcan pestañas, trabajadores en segundo plano y ventanas emergentes; Chrome Web Store prohíbe el código remoto; y los ciclos de revisión lentos rompen los bucles de retroalimentación rápida que hacen que la codificación con IA sea productiva. En esta charla, cubriré el panorama de la personalización del navegador desde bookmarklets y user scripts hasta plataformas de modificación de bajo código y navegadores con IA. También compartiré técnicas prácticas para adaptar los flujos de trabajo de codificación con IA al desarrollo de extensiones basadas en la experiencia de enviar extensiones en empresas de todo el mundo.

Hacer que la Codificación con IA Funcione para Extensiones de Navegador de Nivel Empresarial

Todd Schiller discusses using AI to code an AI-enabled extension for global use by enterprises, emphasizing empowering users through browser extensions for versatile automation and customization. Browser extensibility covers various customization methods, including user scripts and low-code options like Pixie Bricks. Challenges in applying AI tools to browser extensions include their distributed nature, slower iteration loops, and adaptation to host sites. Understanding browser extensions as distributed systems reveals complexities such as async operations and intricate component lifecycles. Key best practices will be discussed, starting with selecting the appropriate foundation. Choose a productive foundation like libraries or frameworks for browser extension development. Select between user scripts or low-code platforms like Pixiebricks. Focus on encapsulating extension-specific code, using explicit coding, and enforcing boundaries with linters. Create fast feedback loops with tools like TypeScript, Lint rules, and realistic integration testing with mocks and fakes. Encapsulate extension-specific code using standard tactics like dependency injection and explicit coding. Create fast feedback loops with TypeScript, Lint rules, and realistic integration testing. Shift left to catch issues earlier and avoid relying on slow feedback loops. Avoid testing on host pages directly to prevent slow and potentially flaky behavior detection. Utilize dependency induction, mocks, and fakes for realistic integration tests. Test boring components with Storybook and Playwright MCP. Shift left to catch issues earlier, avoid testing directly on host pages, create test pages for encountered patterns, leverage AI for generating test pages, and ensure a stable testing environment.

Mainstream "vibe coding" workflows weren't designed for browser extensions. Extensions have distributed architectures spanning tabs, background workers, and popups; Chrome Web Store forbids remote code; and slow review cycles all break the fast feedback loops that make AI coding productive. In this talk, I'll cover the browser customization landscape from bookmarklets and user scripts to low-code mod platforms and AI browsers. I'll also share practical techniques for adapting AI coding workflows to extension development based on experience shipping extensions in enterprises around the globe.

Making AI Coding Work for Enterprise-Grade Browser Extensions

Join a focused 2-hour session covering MCP's purpose, architecture, hands-on server implementation, and future directions. Designed for developers and system architects aiming to integrate contextual data with ML models effectively, using MCP's latest 2026 features like semantic routing and federated context.Agenda:<ul><li>- Introduction &amp; Why MCP? Key challenges MCP solves and core benefits, including 20x throughput from federated layers.</li><li>- Architecture Deep Dive: Components, interactions, scalability principles, plus quantum-resilient encryption in MCP 2.1.</li><li>- Building Your Own MCP Server: Guided walkthrough with code snippets and best practices; live demo of multimodal data fusion.</li><li>- Future of MCP Developments: Potential enhancements like agentic workflows, emerging trends, and real-world scenarios.</li></ul>Key Takeaways:<ul><li>- Clear understanding of MCP's rationale.</li><li>- Insight into design patterns and scaling considerations.</li><li>- Practical steps to implement a prototype server.</li><li>- Awareness of upcoming trends and how to apply MCP in projects.</li></ul>

Model Context Protocol (MCP) Deep Dive: 2-Hour Interactive Workshop

In this workshop, you’ll learn a repeatable process for building full stack apps in Cursor. We’ll start with an idea and use an agent to shape clear requirements. Next, we’ll generate a database schema and set up the build tools. Then we’ll create the backend and frontend, and wire the layers together. Along the way, you’ll learn how to utilize context loops, catch hallucinations, fix errors, and prompt for modern UI. By the end, you’ll be able to build a working AI generated full stack application on your own.

Advanced Cursor Techniques for Building Fullstack Apps

Mike Mikula

This workshop introduces Sandboxes as a general computing primitive and uses Cloudflare Sandboxes as the concrete implementation. Participants explore common, real-world use cases for sandboxed execution and learn how to reason about when and why sandboxes are the right tool. The workshop is intentionally product-adjacent rather than product-specific. Cloudflare Sandboxes are used throughout, but the concepts apply broadly to modern secure execution environments. Workshop Goals By the end of this workshop, participants will be able to: - Explain what a sandbox is and why it exists as a compute primitive - Identify common problems that sandboxes are well suited to solve - Safely execute untrusted or dynamic code - Use sandboxes for development, testing, and automation workflows - Reason about security, performance, and tradeoffs when choosing sandboxed execution

Using Sandboxes to Safely Execute Untrusted or Dynamic Code

Craig Dennis

Ever wondered how to take your AI-generated Replit code and deploy it to production? This step-by-step tutorial shows you exactly how to deploy a Replit landing page to Render's production environment in minutes.What you'll see:<ul><li>Generate a production-ready landing page using Replit AI</li><li>Deploy to Render with proper build configurations</li></ul>

Deploy Your AI Code in Minutes

Don’t get left behind—discover how AI is revolutionizing software engineering at the AI Coding Summit. Dive into cutting-edge talks and hands-on workshops on AI-powered software development. Network with industry leaders, explore game-changing tools, and stay ahead in the fast-evolving world of AI-driven development. Do not miss the biggest shift in software engineering.

AI Coding Summit

I'm attending AI Coding Summit with 5k other participants to learn how AI transforms software development and stay up to date with the latest AI tools - get your free remote ticket and join me there!

AI Coding Summit 2026

FocusReactive

Hygraph

Cloudflare

Backslash

Red Hat AI

Google DeepMind

PixieBrix

Render

<a href="https://workers.cloudflare.com/" target="_blank" rel="noopener noreferrer">Cloudflare’s Developer Platform</a> lets you ship software without managing infrastructure. Modern apps need to connect users, data, and services securely and reliably. Deploy once, run everywhere on our global network. We offer primitives for compute, data, media, and AI, simplifying development and lowering costs while ensuring scalability and performance.

Red Hat AI is a platform that accelerates AI innovation and reduces the cost of building and delivering AI across hybrid cloud environments, with optimized models, efficient inference, and open source at its core. <a href="https://www.redhat.com/en/artificial-intelligence" target="_blank" rel="noopener noreferrer">Learn more</a>.&nbsp;

<a href="https://deepmind.google/" target="_blank" rel="noopener noreferrer">Google DeepMind</a> is a team of scientists, engineers, ethicists and more, working to build the next generation of AI systems safely and responsibly. By solving some of the hardest scientific and engineering challenges of our time, we’re working to create breakthrough technologies that could advance science, transform work, serve diverse communities — and improve billions of people’s lives.

Build with our next generation AI systems

<a href="http://pixiebrix.com/?utm_source=aiCodingSummit&utm_medium=conference&utm_source=socialPost" target="_blank" rel="noopener noreferrer">PixieBrix</a> enables developers to build and deploy extensions for web apps with real, context-aware automation. Commonly used for building custom web clippers, information retrievers, and AI Copilots.&nbsp;

Build AI apps and agents that actually scale on Render's modern cloud. Ship faster, scale reliably, and innovate without infrastructure headaches. <a href="https://dashboard.render.com/register?utm_source=ai_coding_summit_2026&utm_medium=email" target="_blank" rel="noopener noreferrer">Learn more</a>.&nbsp;

<a href="https://focusreactive.com/" target="_blank" rel="noopener noreferrer">FocusReactive</a> is a specialist agency in AI-powered web platforms. As partners of Vercel, Shopify, Payload CMS and Sanity, we deliver production-grade AI integrations - from semantic search to open-source LLM toolkits for headless CMS and eCommerce.

FocusReactive is a <a href="https://focusreactive.com/nextjs-agency">NextJS expert agency</a> with deep expertise in <a href="https://focusreactive.com/headless-cms-agency">Headless CMS</a> and <a href="https://focusreactive.com/headless-commerce-agency">Headless Commerce</a>. Independent <a href="https://focusreactive.com/storyblok-expert-agency/">agency partners of Storyblok</a>, Payload CMS, Vercel, Shopify, and Sanity to build fast, scalable, SEO-friendly platforms - and provide NextJS tech &amp; AI SEO audits to optimize existing sites.