Prompt: the core primitive of an LLM application[section]In the channel's material, a prompt is treated as one of the three core primitives of an LLM application — alongside the eval and the data set. Together, these make up what the source calls the prompt engineering workflow: iterating on an LLM application through its prompts, evals, and data rather than one-off trial and error. [^01a0f15b-05e8-7884-88e4-0649ae182ec0@120]### More than a sentenceThe material frames the prompt as “the source code” of an LLM application — the artifact you deliberately iterate on. [^01a0f15b-05e8-7884-88e4-0649ae182ec0@120] Strong results, the source argues, come less from clever single sentences and more from building a workflow around the prompt: encoding domain knowledge and style into reusable prompt artifacts, and correcting the system when it misses. [^01a0f15b-05e8-7884-88e4-0649ae182ec0@151]### Why prompting enduresThe source's position is that prompt engineering endures because its real job is putting domain knowledge into your LLM system. The pleasantries of prompting may eventually be automated away, but the core primitives — the prompt among them — remain. [^01a0f15b-05e8-7884-88e4-0649ae182ec0@151]- A prompt is one of three workflow primitives of an LLM application, with the eval and the data set.- Treat the prompt like the application's source code: an artifact you iterate on deliberately, not a one-off sentence.- Winning prompting means encoding domain knowledge and style into reusable prompt artifacts.- The prompt persists because its job — injecting domain knowledge into the system — can't be fully automated away.