What Poolside does
Poolside is a frontier-labs-style AI company focused on building open-weight, agentic foundation models for software engineering, plus the systems that help those models reliably act inside developer workflows and enterprise environments. The company’s homepage positions Poolside as “building open-weight foundation models and the systems that refine and improve them,” with its primary model family being the Laguna line for agentic coding.
More
On the product side, Poolside provides multiple ways to use its models: a desktop client (“Poolside Desktop Assistant”), an installable terminal/agent experience (“pool”), editor integrations (VS Code, Visual Studio, Zed, and JetBrains via an ACP setup flow), and an OpenAI-compatible chat API exposed via an API key on its platform. Poolside’s technical differentiation, based on its published materials, centers on (1) training its own models from scratch using its own data and infrastructure, (2) using reinforcement learning with feedback from code execution (described in its $500M fundraise post as Reinforcement Learning from Code Execution Feedback / RLCEF), and (3) building a “Model Factory” that combines data preparation, automated evaluations, and training pipelines with execution-based feedback and large-scale experimentation. Strategically, Poolside has emphasized enterprise go-to-market and deployability: it announced that it became a “first-party AWS offering” (meaning its product and models can be contracted directly from AWS) and claims a training cluster scale-up (from 1,000 GPUs to 10,000 GPUs) as part of that partnership. Poolside has also moved toward a more complete “agentic enterprise” delivery model. In a November 2025 acquisition announcement, Poolside said it acquired Fern Labs (a London-based research engineering company behind Bridge, a multi-agent orchestration layer) to combine Poolside’s foundation models and Model Factory infrastructure with on-site agent deployments.