LLM Native PDE

​This year, LLM Native is coming to Punta del Este for the first time as the only technical space for AI engineers at Punta Tech Week.

​Sponsored by AWS and hosted by Horizon, this is a small, invite-only gathering for founders, engineers, and practitioners shipping AI-first products, working hands-on with agents, tooling, and evals. The format is built around real experiences: what's working, what's not, and the trade-offs that don't make it into blog posts.

​If you're building in this space and want high-signal conversations with peers, we'd love to have you!

​Lightning Talks ⚡️

​Lineup:

• ​Miguel Langone (CTO @Horizon) - AI-Ready Codebases: Accelerating with Spec-Driven Development

  • ​Why AI systems need AI-ready codebases. This talk covers how Spec-Driven Development turns specifications into a shared contract between humans and LLMs, reducing ambiguity and significantly increasing development speed and reliability.

• ​Francisco Ingham (CTO @FieldData) - Towards the Ultimate Coding Workflow

  • ​The past few months have marked a turning point in AI-assisted development. The combination of Opus 4.5, new versions of Claude Code and parallel background agents has redefined what's possible. We'll explore the stack, plugins and workflows that can help us best leverage this shift to drastically accelerate development teams while improving code quality.

• ​Santiago Castillo (Head of AI and ML @Akua) - High-Code Agents at Low-Code Speed: Scaling with a Modern AI Stack

  • ​Choosing between the simplicity of n8n-like tools and the control of a high-code stack is often a trade-off between speed and capability. We will dive into how a modern ecosystem—leveraging LangChain, Agentcore, and Claude Code—eliminates this friction. This combination provides the necessary orchestration, infrastructure, and developer velocity to launch and scale high-code agents with the same ease as visual automation platforms, without sacrificing any architectural control.

• ​Javier Quincke (VP of Engineering @Sytrex) - Harnessing deep agents for retrieval

  • ​We'll demonstrate how deep agents can bypass traditional semantic search approaches when retrieving information from large document volumes, using iterative reasoning and basic tools instead of complex RAG pipelines with embeddings and vector stores. This agentic approach sidesteps common challenges like maintaining vector databases, tuning retrieval parameters, and managing chunking strategies - achieving comparable or superior results through direct reasoning, much like a skilled researcher navigating a library.

​Dinner & Chat 🚢

​As builders we like to be among builders. We will keep the talks short and sweet and have ample time to connect over a nice, informal dinner.