Curiosity Builds! Ep 12: Airplane Coding + Chatter Board (LLM Debugging + Human vs AI Labels)

Welcome back to Curiosity Builds! (and apologies on the audio—my podcast mic is back in Seattle). This episode looks a little different because I quietly did a big chunk of work mid-flight (yes, I paid $8 for airplane Wi-Fi and coded for hours in the corner of my seat 😅). We’re continuing the Conundrum Desk—our tribute build honoring the Slate Political Gabfest’s 20th anniversary—and tightening the experience right at the finish line. What we do in this episode: - Major UI refresh: the board now feels like a desk/pinboard (it’s a “chatter board,” not a leaderboard) - Add pinning + “chatterworthy” stamps for the top conundrums - Clean up toss/revive flow and bring back an option to write your own conundrum - Expand the drawers (topic/vibe/stakes/people + a new “constraint” twist button) - LLM + data layer reality: we debug why the conundrum generation wasn’t landing in the UI (OpenRouter + a prompt we wrote to match the Gabfest vibe) - Fix a parsing issue where the model was wrapping JSON in markdown/code fences - Start improving readability/interaction: a bigger “notepad” view + cards that expand on click - Work on labeling conundrums as Human / AI / Human+AI (and discover the Supabase/SQL update we needed) We end with the core loop working again—and a clear next step: tighten the prompt (length/bleed), refine the drawers + reset behavior, and keep improving how we show Human + AI together. Watch the Curiosity Builds! playlist: https://rebrand.ly/CB-Playlist Learn more about Curiosity Builds!: https://hypandra.com/curiosity-builds Hypandra: https://hypandra.com #ailiteracy #chatgpt #curiositybuilds #claudecode --- Welcome to the Curiosity Builds! video series. This is not a polished tutorial. It’s public learning. We’re vibe coding in the real world—using modern AI + creative tools (Cursor, Vercel, GitHub, Claude, Codex, and whatever else ends up in the stack)—to show how these tools can help you create, iterate, and chase ideas fast WITHOUT outsourcing your critical thinking. You’ll see: - how features actually get added (the messy way) - what breaks (and how we debug it) - how to use AI intentionally instead of letting it do your thinking for you - the human part: curiosity, judgment calls, tradeoffs, and experimentation -AND MISTAKES! (both Humand and AI)