https://guillemus.com/posts/Recent content in Writing on GuillemusHugoen-USSun, 28 Dec 2025 00:00:00 +0000https://guillemus.com/on-programmatic-tool-calling/Sun, 28 Dec 2025 00:00:00 +0000https://guillemus.com/on-programmatic-tool-calling/<p>Last month anthropic <a href="https://www.anthropic.com/engineering/advanced-tool-use">published an article</a> about having their models better use tools. They mention these three techniques:</p> <blockquote> <ul> <li>Tool Search Tool, which allows Claude to use search tools to access thousands of tools without consuming its context window</li> <li>Programmatic Tool Calling, which allows Claude to invoke tools in a code execution environment reducing the impact on the model’s context window</li> <li>Tool Use Examples, which provides a universal standard for demonstrating how to effectively use a given tool</li> </ul> </blockquote> <p>At the time the article went out, I did not really put much thought into it, until [Theo made a video calling them out for not acknowledging their failure on <a href="https://modelcontextprotocol.io/">MCP</a>](<a href="https://www.youtube.com/watch?v=hPPTrsUzLA8">https://www.youtube.com/watch?v=hPPTrsUzLA8</a>). I generally agree with him, but he mentions something particularly interesting: models are so much smarter when you let them write some code.</p>