Mihai Farcas·1d agoWhy Row-Level Security Is the Safest Place for Authorization in AI-Built AppsHow many times has an AI coding agent written a feature that works, and left the tenant filter out of the query?
Mihai Farcas·Oct 29How to Implement LLM Tool Calling in Automation Workflows With Error HandlingLLM tool calling means the model does not run anything itself. It reads a list of tool definitions, picks one, fills in the arguments, and…A response icon1A response icon1
Mihai Farcas·Oct 26Instant GraphQL and REST from Any Postgres Database: Introducing graphpgIf you like GraphQL but found tools like Hasura too heavy, here is what a database-first Rust engine looks like.
Mihai Farcas·Oct 15When “Try Again” Is the Wrong Fallback: How to Build Fallback Logic into AI-Powered AutomationHow do you build fallback logic into AI-powered automation without risking incorrect actions, duplicate transactions, or broken workflows…
Mihai Farcas·Oct 7How to Autoscale Workflow Automation on Kubernetes: Queue Depth, Concurrency, and Safe Scale-DownA production playbook for autoscaling n8n queue-mode workers with KEDA, deployed through Argo CD on a k8s cluster.
Mihai Farcas·Sep 24How to Evaluate the Performance of Autonomous AI AgentsEvaluating an AI agent’s performance means checking more than whether its final output is correct, it means verifying whether the agent…
Mihai Farcas·Aug 20How to Chain Chatbot Queries to Database QueriesChaining chatbot queries to database queries and back takes four steps: the LLM reads the question, writes SQL against the described…
Mihai Farcas·Jul 21I Hired Claude, But Slower and More ExpensiveA bad hire taught me what AI did to technical interviewing.
Mihai Farcas·Jul 11Ten Correct Rules That Resulted in a Broken SystemA lesson in antifragile software design and the two techniques that will help you keep your sanity.
