DN #2: Art School to AI Infra - The Future of Agents (w/ Devon Kelley)

The big 2026 thing in AI is the promise of AI that becomes a true labor replacement.
In this episode of DN, Niklas sits down with Devon Kelley, a two-time founder based in San Francisco who is building the infrastructure to make that promise a reality.

While the hype around Agentic AI is massive, the reality is that most agents fail in production. Devon explains why agents have a 60% failure rate, how "Autonomous Routing" solves this, and why the recent viral "Molt Book" experiment signals a massive paradigm shift in how software will be built.

Devon also shares her incredibly unique path—going from Art School to becoming a technical founder in deep tech—and gives actionable advice on Go-To-Market strategies for infrastructure startups.

In this episode:
• The Agent Reality Check: Why simple chatbots work but agents break
• Calibre: How to take agent failure rates from 60% to under 5%
• Defining "Agentic": It’s not just an LLM; it’s an LLM + Tools
• The "Molt Book" Moment: What happens when AI talks to AI?
• Founder Story: Why an Art School background is a superpower in tech
• Bootstrapping vs. Venture: Lessons from Devon's previous exit in MedTech
• Go-To-Market: Why you should target the smallest viable audience first
• Claude: Why Claude Code is winning for builders right now

🎧 Full episode on all podcast platforms
💬 Do you think Agents will replace labor by 2026? Let us know in the comments!
🔔 Please like and subscribe! Every subscriber helps our channel grow.

#DN #AIAgents #DevonKelley #ArtificialIntelligence #SaaS #StartupFounder #ClaudeAI #MoltBook #TechInfrastructure #AgenticAI #SiliconValley

Timestamps:
0:00 Intro: Niklas meets Devon Kelley
1:30 The Problem: Why Agents Fail in Production (60% Error Rate)
4:10 Defining "Agentic AI": LLM + Tools
6:16 Why Observability isn't enough (The case for Autonomous Routing)
10:23 Growth Curves: Viral Apps vs. Infrastructure
13:33 The "Molt Book" Viral Moment: When Agents collaborate
18:40 Go-To-Market Strategy: Narrow Targeting & Hackathons
22:35 Unique Path: From Art School to AI Founder
29:57 Bootstrapping vs. Venture Scale (Medical Device Compliance)
36:50 The #1 Advice for Founders: Find the Problem First
41:35 Claude: Why Devon likes Claude