Metacognition in AI Agents | Agentic Software Development & SmartDB
summary
Claims that agentic development adds true metacognition: agents run internal checks of assumptions, keep persistent memories of failures and lessons in Agentic SmartDB, and require audit history, access control, escalation rules and quality gates. A demo MVP generates an enterprise workorder and ties developer CVs to tasks in about one to two minutes, exposes governance and evaluation workflows, and integrates an agent studio/agent factory; the system runs live in the demo and the session registers ~11,000 viewers.
AI-written summary · about
chapters (10)
- 0:00Metacognition in AI agents
- 3:38AI generation limits
- 6:44Next-token prediction limits
- 9:10Metacognition in AI development
- 12:04Enterprise needs beyond generation
- 15:27SmartDB and AI memory
- 19:27Demo of metacognitive agent
- 27:27AI generation to agents
- 30:00Path to AGI
- 36:41Q&A
related talks
29mPairing with the machine: How AI Agents are rewriting the rules of software development
29mAgentic AI: What It Is and Why It Matters
29mBuilding Trust in AI Agents
58mAdvanced Autonomous AI Agents: Memory, Collaboration and Evolution in Action
27mAgentic Engineering: Working With AI, Not Just Using It
34mAccountability in the Age of AI Agents
51mSentience Agility: Harmonizing AI Agile Intelligence
27mAgentic Systems: Leadership, Risk, and Accountability