Nate B Jones explains that long conversations repeatedly send the same accumulated input back to the model. That reused context can dominate token consumptionAI token efficiency measures how effectively a model or workflow turns consumed input and output tokens into useful results., so the biggest savings come from controlling what remains in the conversation rather than merely shortening the newest prompt.
Nate B Jones recommends editing an incorrect prompt instead of appending a correction, grouping related questions and starting a fresh conversation when the task changes. When work continues across sessions, preserve the accepted artifact or decision rather than the entire exploratory history that produced it.
The workflow also benefits from concise outputs, lightweight source formats and deliberate retrievalRetrieval is the process of selecting relevant stored information and returning it to an AI system for the current task.. Users should search files before loading them, keep approved answers easy to find and expose only the tools and documents neededModel selection chooses the AI model whose capability, quality, cost, speed, safety, and operating constraints best fit a task. for the current task. These practices reduce noise as well as cost.
Compaction and context editingContext engineering designs how relevant instructions, evidence, state and tool information are assembled for a model's task. can extend a session, but they are approximations rather than perfect memory. Nate B Jones therefore treats them as supporting controls and uses smaller models or prompt cachingPrompt caching reuses compatible computation from previously processed prompt content to reduce repeated model-input work. only when the task and repeated workload make those choices appropriate.
Watch on YouTube




