MKMudassir Khaninmudassirworks.hashnode.dev·Sep 9 · 10 min readLLM Context Window Management in Production: A Governance ChecklistYour model advertises a 200K token window. Your agent starts producing worse answers somewhere around 90K, and nobody notices for three weeks, because nothing throws. That gap between advertised capac01A
MKMudassir Khaninmudassirworks.hashnode.dev·Sep 7 · 9 min readGraphRAG vs RAG: The Break Even Analysis Before You SwitchThe GraphRAG vs RAG question usually gets asked backwards. Teams ask whether a knowledge graph produces better answers, and the honest answer is that it depends entirely on the shape of the query you 01A
MKMudassir Khaninmudassirworks.hashnode.dev·Sep 2 · 8 min readAI Agent Tool Selection: Why Agents Break When You Give Them Every ToolYou registered every tool the team asked for, the agent got worse, and nothing in the logs explains it. AI agent tool selection is the part of agent engineering that fails quietly, because every tool 00
MKMudassir Khaninmudassirworks.hashnode.dev·Sep 1 · 12 min readRAG Retrieval Latency in Production: Where the Time Actually GoesIf you have shipped a retrieval augmented generation system and users keep telling you it feels slow, the most useful thing you can do is stop treating RAG retrieval latency as a single number. It is 00
MKMudassir Khaninmudassirworks.hashnode.dev·Aug 21 · 9 min readSession vs JWT in Next.js 16: What Changed and What Most Guides Get WrongEvery time someone asks whether to use sessions or JWT for authentication in Next.js, the honest answer is "it depends on what you're building," and then the conversation stalls because most guides ne011
