The Everything Feed - All Packet Pushers Pods HN811: What AI Startups Get Wrong
Jan 23, 2026
Carlos Pignataro, Founder and principal at Blue Fern Consulting with deep networking and startup experience, cuts through AI hype. He distinguishes real AI work from flashy chatbots. He argues for modular agents, better telemetry and metadata, and clear definitions before design. He warns against trend-driven funding and stresses human roles in supervising autonomous systems.
AI Snips
Chapters
Transcript
Episode notes
Compose Narrow Agents For Bigger Problems
- Decompose network automation into narrow autonomous modules or agents.
- Carlos suggests composing specialized agents (topology, fault isolation) into larger systems for more capability.
Keep Humans In Decision And Supervision Loops
- Place humans where decisions, planning, and supervision add the most value.
- Use humans for architectural choice, adult supervision, and system optimization rather than full automation.
Different AIs Behave Differently
- 'AI' bundles different technologies that behave differently and shouldn't be treated interchangeably.
- Carlos warns that LLMs and ML methods have distinct characteristics and convergence behaviors.