Sharan Harsoor·Nov 5Why I Built Runkite: From Building AI Agents to Operating ThemThe problem I saw, the assumptions I changed, and the self-hosted control plane I built around the agent loop.A response icon1A response icon1
Sharan Harsoor·Oct 14AI Learned to Write Code. It Never Learned to Hold a Wrench.Code, writing, and web AI already have a fair, shared test that everyone compares their work against. Industrial maintenance never did…A response icon1A response icon1
Sharan Harsoor·Sep 12Your RAG is only as good as a cut you never look atYou’re building a RAG pipeline. Embedding model: picked. Vector store: up. Retrieval: written. The chatbot is fine until a user asks about…A response icon1A response icon1
Sharan Harsoor·May 28Agent Skills: The Quiet Standard That’s Changing How We Build AI AgentsEverything I learned going from complete beginner to publishing my first skill — including all the things I got wrong along the way
Sharan Harsoor·Apr 13The Algorithm That Saves Your Photos in Space — And on S3How a “trivial” math paper written in an afternoon became the backbone of every storage system on the planet.
Sharan Harsoor·Jan 21The Rise of MCP: How a “USB-C for AI” Is Quietly Reshaping the Future of Intelligent SystemsA Complete Journey from Zero to Multi-Agent Orchestration
Sharan Harsoor·Nov 22, 2025Cost Optimization in LLM Observability: How LangFuse Handles Petabytes Without Breaking the BankPart 3 of 3: Storage Economics, Compression Magic & The Art of Scaling Cheaply
Sharan Harsoor·Nov 21, 2025OpenTelemetry for LLMOps: How LangFuse Achieved Universal Multi-Language Support Without Building…Part 2 of 3: Standards, Interoperability & The Future of LLM Observability
Sharan Harsoor·Nov 20, 2025From 50 Seconds to 10 Milliseconds: Inside LangFuse’s Journey to Zero-Latency LLM ObservabilityPart 1 of 3: Event-Driven Architecture & Async SDKs
Sharan Harsoor·Jun 2, 2025The Complete Guide to Embeddings and RAG: From Theory to ProductionA comprehensive deep dive into embeddings, tokenisation, vector databases, and Retrieval-Augmented Generation architectures for modern AI…A response icon6A response icon6