Trusted in Production: Monitoring, Guardrails, and Reliability for AI Agents Key Takeaways Prompt engineering is not a security control. Enforce guardrails at the execution layer. Instrument...

Agents That Recall, Experiment and Learn: Persistent Memory and Reflective Learning Key Takeaways A stateless agent is a liability for recurring work. Invest in memory proportional to the task...

Beyond the Chat Box: Extending Claude and GPT Agents with CLI, MCP, and Custom Tools Key Takeaways Tools are the agent’s hands. Without them, it’s just text generation. Validate tool inputs ...

See and Act: Building Agents with Vision and Interface Control Key Takeaways Vision turns any UI or document into an API. Stop waiting for one to exist. Use detail: "high" for text-heavy scr...

Agent Economy: Tokenization, Inference Parameters, and the Cost of Deployed Agents Key Takeaways Tokenization sets the unit of agent economy. JSON, code, and structured data tokenize at 1.5–3x...

Fast Enough to Matter: Agent Retrieval Latency in Enterprise Data Pipelines Key Takeaways The dominant latency in enterprise AI agents is the query execution roundtrip, not model inference. Op...

The Platform Is the Agent. Choosing Your AI Runtime for Models, MCP, and Tools.

Deploying environments with Bicep, scaling IaC with modularization & zone design.

Microsoft Fabric Copilot & AI for Everyone.

The next evolution of OneLake security.