on.agentops

What does it really mean to become the CEO of your own life in the age of agentic AI? Can we truly delegate thinking, decision-making, and execution to autonomous systems? And as AI agents take on more responsibility, how do we maintain focus, control, and trust? In this episode, Joerg sits down with Kevin Today to get alook at AgenC "the CEO command center for your fleet of Claudes" and unpack the philosophy and mechanics behind agentic AI, from building personal "work factories" to designing systems that scale your output without losing your edge. Kevin's journey: from Illinois to global deployments (Palantir, Singapore, Saudi Arabia, Brazil). The origin of AgenC: building a personal AI-powered work factory. What it means to be the CEO of your own life using AI. Using the Double Diamond framework for structured agent workflows. Inside AgenC's architecture: skills, missions, sub-agents, and repos. "Fix the factory, not the output," improving systems over time. Context lock-in and why your setup may matter more than the model. Agent permissioning: zero standing privilege, blast radius, and hierarchy models. Security risks and trust in agentic systems. The future of AI: from Q&A tools to autonomous work factories. Tools & frameworks mentioned • Agency (GitHub) • The Almanack of Naval Ravikant • Getting Things Done • Double Diamond framework • MIRN (Most Important Right Now) • Claude Agents Guest: Kevin Today Kevin Today is a computer scientist, former startup founder, and the creator of the open-source AgenC framework. He specializes in building agentic AI systems, scalable workflows, and autonomous architectures that enable individuals to operate at a higher level by delegating non-unique work to intelligent agents. Connect with Kevin https://github.com/mieubrisse https://x.com/kevinjtoday https://kevintoday.com/  

What does it really take to move Agent AI from theory into production? Can we trust autonomous agents to make decisions without spiraling into hallucinations? And when speed, scale, and security collide, how do developers keep control? In this episode, Sohaib Anwar reveals the hidden mechanics of building reliable agent workflows from serverless deployments to guardrails that keep AI honest. Episode Highlights POC → prod: serverless + WebSockets Loop control Hallucination hygiene Security guardrails And more… Timestamps 0:00 – Introduction and guest background 2:14 – Early journey into AI and first projects 6:26 – Challenges in learning machine learning and algorithms 10:01 – Discussion on transformer models vs. traditional ML 14:41 – Observability of AI Agents 18:27 – How wide or narrow to guide agent behaviour... 22:17 – The future of agent AI and closing thoughts   Tools & frameworks mentioned LangChain / LangGraph LangSmith Traceloop/OpenLLMetry Redis WebSockets Kafka Datadog AWS CloudWatch  Guest: Sohaib Anwar Sohaib Anwaar is an AI Engineer specializing in AgentOps, multi-agent systems, and production-grade LLM integrations. He has worked on end-to-end agentic architectures, scalable orchestration layers, and real-time context-aware automation for enterprise environments. His focus is on building reliable, adaptive, and high-performance agent workflows aligned with real-world business constraints. Reach out to him and connect on: Instagram: https://www.instagram.com/sohaibanwaar.life?igsh=bzk1bnZqa252bXJr&utm_source=qr LinkedIn: https://www.linkedin.com/in/sohaibanwaar/