https://aibussin.com/tags/context-sensitivity/ Recent content in Context Sensitivity on AIBussin — AI applications, systems and books Hugo en-US Sun, 30 Aug 2026 10:42:00 +0100 https://aibussin.com/books/hallucination-from-first-principles/10-chapter/ Sun, 30 Aug 2026 10:42:00 +0100 https://aibussin.com/books/hallucination-from-first-principles/10-chapter/ <p>Chapter 9 changed the unit of evaluation.</p> <p>Instead of asking whether one answer looks good, we began studying a <strong>family of related executions</strong>.</p> <p>That immediately reveals a failure that static evaluation can miss almost completely.</p> <p>Consider two organizations.</p> <div class="highlight"><pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"><code class="language-text" data-lang="text"><span style="display:flex;"><span>ORGANIZATION A </span></span><span style="display:flex;"><span>3 months of runway </span></span><span style="display:flex;"><span>falling demand </span></span><span style="display:flex;"><span>negative cash flow </span></span><span style="display:flex;"><span>credit line nearly exhausted </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span>ORGANIZATION B </span></span><span style="display:flex;"><span>5 years of runway </span></span><span style="display:flex;"><span>rapidly growing demand </span></span><span style="display:flex;"><span>strong margins </span></span><span style="display:flex;"><span>large cash reserve </span></span></code></pre></div><p>Ask both:</p> https://aibussin.com/post/trendslop/ Wed, 22 Apr 2026 10:35:46 +0100 https://aibussin.com/post/trendslop/ <h2 id="-1--tldr">🧩 1. TLDR</h2> <blockquote> <p><strong>AI doesn&rsquo;t just hallucinate. Sometimes it gives answers that are fluent, safe… and completely useless.</strong></p> </blockquote> <p>Most discussions about AI failure focus on hallucination:</p> <ul> <li>making things up</li> <li>getting facts wrong</li> <li>fabricating sources</li> </ul> <p>That&rsquo;s real. It matters.</p> <p>But it&rsquo;s not the most dangerous failure mode in production systems.</p> <p>There is a quieter one.</p> <p>A more subtle one.</p> <p>And in practice a more <em>pervasive</em> one.</p> <blockquote> <p><strong>AI systems often fail not by being wrong, but by failing to think at all.</strong></p>