AI Breakfast Shanghai

AI Breakfast #15

πŸ’° AI Subscriptions & Workarounds

πŸ› οΈ AI Productivity Tools

  • Coding & Development:
    • Mixed reviews on AI coding assistants – speed vs. code quality concerns
    • Cursor and Windsurf praised for understanding codebases
    • Prompt engineering frameworks (like "Constitutional AI" method) discussed for better outputs
  • Research & Analysis:
    • Perplexity favored for factual accuracy and source citations
    • ChatGPT-5 improved but still produces incorrect source links sometimes
    • One member uses AI for financial modeling (volatility forecasting) with Python
  • Everyday Uses:
    • Voice-controlled cooking assistant for real-time recipe help
    • Resume optimization debate – should employers require AI-generated applications?
    • Note-taking tools (Granola) for interview transcription

πŸ‘¨β€πŸ‘©β€πŸ‘§β€πŸ‘¦ AI & Parenting

  • Screen time concerns dominated:
    • Short-form content (TikTok) seen as attention-span killers
    • Preference for long-form content (films, structured learning)
    • Car AI systems (e.g., Li Auto's voice assistant) unexpectedly engaging for kids
  • Education approaches:
    • Alpha School model discussed: 2-hour AI-driven academics + project-based afternoons
    • Khan Academy recommended as free alternative
    • Debate over early AI exposure (5-year-olds)

πŸ† Hackathon Project Spotlight

  • One member shared a "Thank You Card" app built for a Tanstack/Convex hackathon:
    • Uses Firecrawl to scrape company branding
    • Multi-user "bump" system (like digital fist bumps)
    • Tech stack: Tanstack (React framework), Convex (real-time backend), Autumn (Stripe alternative)
    • Prize pool: $100k+ in cloud credits + cash

πŸŽ™οΈ Podcast Recommendations

Attendees shared favorites for long-form content:

πŸ€– AI Agent Experiments

  • Fun proposal: AI "Survivor" simulation
    • Agents vote each other off based on interactions
    • Token limits to force strategic communication
    • Existing "Elimination Game" benchmark study (GitHub)
← Back to Notes