Context Engineering: Specifics of Context Formation and Management

Modern LLM systems and agent-based solutions demonstrate that the quality and predictability of their behaviour are determined not so much by the model itself as by the context in which it operates. Using prompts without preparing and analysing the required context leads to unstable results, hallucinations, and difficulties with scaling.

This talk examines what kind of context is required for effective use of LLMs, the extent and format of that context, and why “more context” does not mean “better”.

We'll analyse the key context levels, tagging features, and the use of existing artefacts. We'll also explore techniques for working with context and discuss the role of the systems analyst in preparing context and creating effective agents.