http://programmer.ie/tags/rlm/ Recent content in RLM on Programmer.ie Hugo en-us Fri, 28 Aug 2026 11:10:00 +0100 http://programmer.ie/books/dspy-from-first-principles/15-chapter/ Fri, 28 Aug 2026 11:10:00 +0100 http://programmer.ie/books/dspy-from-first-principles/15-chapter/ <h1 id="search-memory-and-long-context">Search, Memory and Long Context</h1> <p>Chapter 14 gave the program tools. It can inspect an environment instead of receiving all evidence up front.</p> <p>Now we face the selection problem:</p> <blockquote> <p>When the necessary evidence is too large to place directly in the prompt, how should the program decide what to inspect?</p></blockquote> <p>Several mechanisms are often collapsed into one word: context. They are different.</p> <div class="highlight"><pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;"><code class="language-text" data-lang="text"><span style="display:flex;"><span>1. Put context directly in the prompt </span></span><span style="display:flex;"><span>2. Retrieve context before the LM call </span></span><span style="display:flex;"><span>3. Let an agent choose retrieval tools </span></span><span style="display:flex;"><span>4. Retrieve previous cases / memory </span></span><span style="display:flex;"><span>5. Let a model programmatically explore a large context </span></span></code></pre></div><p>Each creates a different evidence-selection boundary, with different costs, failure modes, and leakage risks.</p>