Andrej Karpathy has introduced a new concept called 'Context Engineering' which involves setting the stage for AI models by preparing the situation, facts, tone, intent, examples, and more to ensure better performance.
After context engineering, the next steps involve designing software that can work across multiple interactions and inputs, teaching the model how to think rather than just what to say.
Context engineering is crucial to feed AI models the right amount of information to prevent them from flailing or getting overwhelmed, akin to carefully packing a hiking bag with the essentials.
Context engineering shapes the model's behavior by providing system instructions, user profiles, previous conversations, document summaries, and more to enhance the tool's intelligence and human-like interaction.