Mixture of Experts Google’s Gemini 3: AI agents, reasoning and search mode
27 snips
Nov 21, 2025 This discussion features Gabe Goodhart, an AI architect focused on cybersecurity, Merve Unuvar, a specialist in agent middleware, and Marina Danilevsky, a research scientist analyzing AI model behaviors. They dive into Google’s Gemini 3 model, exploring its strong performance yet concerning hallucination issues. The dialogue shifts to AI's impact on the economy through OpenAI’s GDPVal benchmark, and the panel debates the balance between specialized and generalist models. They also tackle the implications of a recent cyberattack automated by AI, stressing the need for robust enterprise defenses.
AI Snips
Chapters
Transcript
Episode notes
Agent Architectures Are Becoming Standardized
- Emergent standard architectures for agents are appearing, like generalist agents and tool management slots.
- Gabe Goodhart likened agent patterns maturing to standard REST API server designs in cloud software.
Reuse Agent Components, Don’t Reinvent
- Reuse modular components like memory and guardrails rather than forcing full migrations to a new agent framework.
- Merve Unuvar recommended ALTK components can plug into existing agents to boost performance.
Benchmarks Miss Deployment Realities
- Real-world deployment reveals issues benchmarks miss, like latency, memory design, and behavioral consistency.
- Merve Unuvar highlighted consistency and memory strategies as priorities after community feedback on Kuga.