Abhishek Veeramalla

AI Companies Don't Want You to Know This | VS Code + Continue

The video walks viewers through building a completely free, agentic integrated development environment (IDE) by pairing Visual Studio Code with the open‑source Continue.dev extension and locally hosted large language models (LLMs). Instead of relying on paid services such as GitHub Copilot, Google Gemini, or Amazon CodeWhisperer, the presenter shows how to run models on‑premise via Ollama, eliminating external API calls and associated fees. A core insight is the three‑model architecture required by Continue.dev: a chat model for conversational code queries, an autocomplete model for real‑time suggestions, and an embedding model that indexes the project’s source files for fast semantic search. The speaker demonstrates installing these models—e.g., Llama 3, a lightweight autocomplete model, and a Nomic embedding model—configuring permissions, and connecting them to VS Code, all within roughly ten minutes. The demo uses the Argo CD repository to illustrate the workflow. The chat model explains the entire project in seconds, the extension can highlight and describe individual functions, and the edit mode can automatically insert comments or suggest improvements, mirroring Copilot’s in‑IDE experience. Additional features such as multi‑session handling, MCP server integration, and customizable model selection are highlighted, underscoring the flexibility of the setup. By enabling developers to run powerful AI assistants locally, this approach democratizes access to advanced coding tools, cuts subscription costs, and safeguards proprietary code. Organizations can adopt AI‑driven development without exposing intellectual property to third‑party APIs, potentially reshaping standard IDE workflows.