Problem

Software teams—especially large enterprises with thousands of developers, complex legacy codebases, and strict privacy and security requirements—need to increase engineering productivity without surrendering control of their code and data. Conventional generative-AI tools can create additional concerns around third-party API access, token-based costs, governance, and deployment boundaries. Poolside’s central opportunity is to make software development more productive by giving AI enough context and autonomy to work through real coding tasks rather than merely autocomplete snippets.

The killer use case is long-horizon engineering work: an agent can plan, implement, execute, inspect the result, and iterate on tasks such as legacy-code refactoring, incident analysis, and security reviews. Poolside says its models learn from execution feedback across real-world codebases, which targets the economics of expensive developer time and difficult maintenance work more directly than simple code completion.

Product / Service

Poolside is an AI lab and software-engineering foundation-model company offering agentic coding models, self-managed inference, APIs, and agent workflows. Its Laguna models are available as open weights and can be used through Poolside’s agent workflows, an OpenAI-compatible API, a terminal CLI, desktop and editor integrations, or hosted inference. The models are designed to plan, implement, and iterate on complex coding tasks, while Poolside’s reinforcement-learning approach uses code-execution feedback to improve performance.

The enterprise delivery model is the main differentiator: customers can run the system inside their own VPC, on-premises infrastructure, or air-gapped environment, keeping code, prompts, and data within their security boundary. The 2026 Poolside Platform adds containerized agent sessions, secret and network-policy controls, searchable audit trails, integrations with common developer and business tools, control of model weights, and a predictable subscription plan. The result is a more controllable alternative to sending sensitive engineering work to an external AI API, with the option to fine-tune models on a company’s own code and data.

Market

Poolside competes in the enterprise AI coding-assistant, agentic software-engineering, and foundation-model markets. Its reference competitors include GitHub Copilot, Cursor, Devin, and Claude Code, while open-weight model providers and cloud-hosted coding-model platforms are also relevant alternatives. Poolside’s positioning is less about a lightweight autocomplete feature and more about secure, deployable software-engineering agents that can handle longer-running tasks inside an organization’s controlled environment.

The company is not merely pre-revenue in the available evidence. It announced a $500 million fundraise in October 2024, was described by AWS as valued at $3 billion in that round, and said early enterprise customers were pressure-testing the product in demanding environments. Its AWS agreement made the Assistant and foundation models available through EC2 and planned Amazon Bedrock deployment, while its 2026 platform and Laguna releases show a broader commercial and open-model distribution strategy. A third-party profile estimated 2025 ARR at $50 million, but that figure is not company-reported in the cited evidence, so the strongest verified traction signals are the financing, enterprise testing, AWS relationship, and expanding product availability.