AI Anxiety? Here's how I recovered.

Explore how physical infrastructure and our social nature create natural barriers to the AI superintelligence narrative that's been keeping me up at night.

What you'll learn:

• Why real-world infrastructure limitations provide a natural brake on AI advancement (insights from my VC days)

• How the social element missing from current AI systems reflects a fundamental human need

• My five most effective prompts for entrepreneurs using AI for self-reflection

The Story: From Dystopia to Deutsche Bahn

Despite having immersed myself in AI tools for the past two years, I've remained surprisingly quiet about their evolution. Then, last month, the dystopian "AI 2027" scenarios rattled me to my core (reading them during a midnight bout with a nasty stomach bug certainly didn't help matters). It took equal time to recover from both the illness and the existential dread. Fortunately, returning to solid-state reality (both in terms of digestive function and simply reconnecting with the tangible world beyond my apartment—these Kita viruses are brutal) helped me get my feet back on the ground. Finding another position paper that offered a counterbalance was the final piece I needed.

Last week, as I stood on platform 3 at Berlin Hauptbahnhof watching my train arrive 126 minutes late (an OK performance by Deutsche Bahn standards), I couldn't help but laugh at the absurdity. The contrast was almost comical: here I was, in one of the most technologically advanced countries in the world, watching a train system struggle with the same basic challenges it has faced since I moved here. The conductor handed me a paper slip to document my delay—not a QR code, not an automated notification through their app, but an actual piece of paper that someone had physically stamped with an old-fashioned rubber stamp.

Another not-so-appetizing example: the road to my office has been closed for weeks. They're installing new sewage pipes... because our excrement, at the end of the day, is still more hardware than software. This daily reminder of our physical reality makes me chuckle whenever I read breathless predictions about our imminent digital transcendence.

This moment crystallized something I've been thinking about as I read dueling narratives about AI's future. On one side, we have breathless predictions of an intelligence explosion by 2027. The paper constantly speaks of exponential gains in software development and then translates this pace directly to the physical world. That made me skeptical. On the other side, a more measured view of AI as a "normal technology" that will integrate gradually, following familiar patterns, felt more realistic.

The Insight: Infrastructure Inertia Is Real

What struck me about these competing visions is how the intelligence explosion narrative often seems to float above the physical world. New technologies don't arrive in a vacuum—they must be implemented within existing systems with massive inertia.

My Finanzamt still communicates with me exclusively through paper letters. Government offices rely on fax machines that I haven't seen anywhere else since 2005. The hospital where my daughter was born still uses paper patient files. This isn't merely stubbornness—it represents the enormous complexity of changing established systems that serve millions.

Back when I was a venture investor, I heard hundreds of founders confidently claim their tech would transform markets within months. "We'll disrupt healthcare in 18 months!" Ten years of watching these predictions fail taught me technological dissemination actually works like this:

1. Technical Innovation: Developing new capabilities (happening rapidly in AI)

2. Organizational Adoption: Integrating capabilities into workflows (slow, messy, expensive)

3. Infrastructure Adaptation: Rebuilding physical and institutional systems (measured in decades)

What "AI as normal technology" gets right is recognizing that steps 2 and 3 are where real friction lies. And they always take longer than technologists predict.

Another insight worth noting: Benchmark partner Sarah Tavel points out that consumer tech follows a pattern—deeply technical founders create powerful but utilitarian tools first, then product-focused founders build social experiences on top. This suggests AI may evolve from today's asocial tools toward applications that better serve our fundamentally social nature.

Your Quick Win: Five AI Reflection Prompts for Entrepreneurs

Over the last weeks, I tested some self-reflection prompts for AI power users. Here are the ones that delivered the most interesting results for me. Use them in ChatGPT or Claude - outside of projects, so it has access to your chats.

1. Biggest Bet: "From all my product and market discussions, what's the clearest unvalidated assumption I'm still betting on?" Ouch.

2. Ideal vs Real: "Based on how I describe myself as a leader, what behaviors do I idealize but not actually model?"

3. Ignored Insights: "What are the top three insights I've ignored or not followed through on, based on our past chats?" Still, this one was the most painful for me. 😅 

4. Hidden Narrative: "Analyze the emotional tone of my business-related prompts. What's the underlying story I seem to believe?"

5. Value Conflicts: "Review everything I've shared here about my goals and values. What decisions have I made that contradict them?"

I am super interested from all of you how you are using AI as entrepreneurs/CEOs. Not how your functional teams are using it, but you, personally, as the leader of a business, to leverage yourself. And – please – no "it helps me draft emails". What's your real, high-value use case?

Resources

• AI Futures Project: "AI 2027"

• Hard Fork Podcast interview with AI 2027 author

• Narayanan & Kapoor: "AI as Normal Technology"

• Hard Fork Podcast interview with Arvind Narayanan