I noticed most \"prompt guides\" out there are meant for marketing copy or writing emails, which isn't very helpful for us actually building AI infrastructure. \n\nOver the last year of building, I compiled a list of 50 specific prompts designed for:\n- Code architecture and system design\n- SQL/Database generation and indexing\n- Code refactoring using specific design patterns\n- Generating high-quality few-shot datasets for fine-tuning\n- Hallucination checking for RAG pipelines\n\nHere are a couple examples from the list:\n*The Codebase Indexer*: \"Analyze the following project structure and summarize the core business logic, identifying any potential bottlenecks in the data flow between [Service A] and [Service B].\"\n\n*The Hallucination Checker*: \"Compare the following model output against the provided source documentation. Identify any factual inconsistencies or made-up details.\"\n\nIf you want the full list of 50, I send it out to anyone who joins my weekly technical newsletter. We also cover deep dives on the shift to SLMs and building reliable agentic workflows.\n\nYou can grab it here: https://project-1960fbd1.doanything.app\n\n(Hope this helps some of you building this weekend!)
1 comments