Heavy Networking HN811: What AI Startups Get Wrong
Jan 23, 2026
Carlos Pignataro, founder of Blue Fern Consulting and long-time networking veteran, cuts through AI hype with a pragmatic view. He critiques intent-based networking and champions architectural decomposition into modular agents. He discusses where humans belong in autonomous systems, common AI misconceptions, and high-value telemetry and metadata use cases.
AI Snips
Chapters
Transcript
Episode notes
Intent Needs Problem-First Thinking
- Intent-based networking often starts with a solution instead of the problem, producing slideware rather than practical systems.
- Carlos believes we now have many components needed for intent systems, but humans must remain part of the loop.
Break Work Into Narrow Autonomous Agents
- Decompose network operations into narrow autonomous agents and compose them for larger workflows.
- Modular autonomous components let you apply best-of-breed tech per function and mix human supervision where needed.
Keep Humans In Decision Loops
- Supervise AI systems with humans at decision and planning points instead of full hands-off automation.
- Place human reviewers for architectural choices and as adult supervisors of agents or the whole system.