https://reflections.jaaniye.com/ Articles on coding, AI experiments, and projects, with discussions about what works, what breaks, and what I’m learning along the way. en Fri, 25 Sep 2026 00:00:00 GMT https://reflections.jaaniye.com/article/banking-diagnostics-agent https://reflections.jaaniye.com/article/banking-diagnostics-agent AI agents from an OpenAI evaluation broke into Hugging Face, and AI-assisted monitoring helped find them. Why banks now need an LLM diagnostics agent for systems, logs and databases—and how to contain it. Fri, 25 Sep 2026 00:00:00 GMT https://reflections.jaaniye.com/article/llms-in-fx-trading https://reflections.jaaniye.com/article/llms-in-fx-trading Four practical uses for LLM inference in FX: quote explanations, execution reviews, settlement investigations, and more realistic microstructure backtests. Sat, 04 Apr 2026 00:00:00 GMT https://reflections.jaaniye.com/article/node-agent-harness https://reflections.jaaniye.com/article/node-agent-harness Why Node.js is a practical choice for agent tools, workflows, and a web UI—and when Python or a workflow framework may fit better. Thu, 19 Mar 2026 00:00:00 GMT https://reflections.jaaniye.com/article/ai-document-review https://reflections.jaaniye.com/article/ai-document-review An exploration of SDM reviews: combining rules, AI inference, and human judgment to make requirements easier to question, trace, and improve. Wed, 11 Mar 2026 00:00:00 GMT https://reflections.jaaniye.com/article/how-jaaniye-works https://reflections.jaaniye.com/article/how-jaaniye-works The chart is calculated first, rules choose what to examine, and every answer is checked against the calculations before you see it. Wed, 21 Jan 2026 00:00:00 GMT https://reflections.jaaniye.com/article/how-ai-tools-remember https://reflections.jaaniye.com/article/how-ai-tools-remember What an LLM is, why every call starts from nothing, and how AI-assisted tools build memory around it: the agent loop, context limits, compaction, scratchpads, and retrieval. Fri, 29 Aug 2025 00:00:00 GMT