Sept 2026LLMOps
Context Engineering Is the New Prompt Engineering
In 2024 we tuned prompts. In 2025 we built RAG pipelines. In 2026 the job is managing everything the model can see — context windows, retrieval layers, tool layers, agent state. Here's the full stack.
Aug 2026Agentic AI
LangGraph + LangMem: Building Agents That Actually Remember You
Every demo agent forgets you the moment you close the tab. Production agents cannot. Here is how to wire LangMem's three memory types into a LangGraph pipeline — and what each one costs you.
Jul 2026Data Engineering
Why Text-to-SQL Gets It Wrong (And What to Use Instead)
Frontier models answer roughly 84–90% of business metric questions correctly via raw SQL generation. The dbt Semantic Layer pushes that to 98–100%. The difference is not the model — it is the interface.