The Winchester Mystery House Problem in AI Development

<p>AI models are starting to act like appliances, locked into one narrow way of working, instead of the flexible infrastructure they used to be. <em>Drew Breunig</em>, an AI and data strategist working with the Overture Maps Foundation, joins us to explain why, and what it means for anyone building something that doesn&#39;t look like Claude Code.</p><p><br></p><p>Drew walks through his &quot;Winchester Mystery House&quot; idea: what happens once code gets so cheap to write that the only real bottleneck left is feedback. From there we dig into DSPy: signatures, the GEPA optimizer, and the brand-new Flex optimizer, which rewrites your code instead of just your prompt, complete with a real before-and-after on cost and accuracy. We also get into why so many AI-built apps and websites end up looking identical, the actual difference between an agent and a workflow, what Drew learned a year after shipping a code library with no code in it, and why he thinks the most valuable thing you can do right now is close the laptop and go talk to people.</p><p><br></p><p>CMPND: <a href="https://www.cmpnd.ai" target="_blank" rel="ugc noopener noreferrer">https://www.cmpnd.ai</a></p><p><br></p><p>Drew Breunig: <a href="https://www.linkedin.com/in/drewbreunig/" target="_blank" rel="ugc noopener noreferrer">https://www.linkedin.com/in/drewbreunig/</a></p><p>Demetrios: <a href="https://www.linkedin.com/in/dpbrinkm" target="_blank" rel="ugc noopener noreferrer">https://www.linkedin.com/in/dpbrinkm</a></p><p><br></p><p>Timestamps:</p><p>[0:00] Cold open: when Claude Code tries to call itself</p><p>[1:19] Biggest AI news: labs trading diversity for reliability</p><p>[2:35] How harnesses get trained into models over time</p><p>[5:41] The problem: your harness starts fighting the model</p><p>[9:13] When do you need your own harness?</p><p>[10:02] The Winchester Mystery House warning</p><p>[16:13] The blank page problem: why everything looks the same</p><p>[20:50] Infrastructure vs appliances: the thesis lands</p><p>[22:40] Current tool loadout: GLM, Kimi, Claude Code, Pi</p><p>[27:04] The Raspberry Pi personal agent running on Slack</p><p>[31:00] Crystallizing tasks: when to replace AI with pure code</p><p>[33:10] DSPy explained: separating what from how</p><p>[35:23] How prompt optimizers actually work</p><p>[39:31] DSPy pre-dates ChatGPT: model-agnostic programs</p><p>[44:00] Why you still need to ship the code, not just the spec</p><p>[50:00] Don&#39;t plan more than a month ahead anymore</p><p>[54:00] Coaching agents all day feels productive — it isn&#39;t</p><p>[57:58] The dopamine of building with agents vs. why you still need human feedback</p>