Undermind’s Post

I thought this would be common knowledge by now, but many of my scientist and researcher friends were pleasantly surprised when I shared this: 𝗟𝗟𝗠 𝗔𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 𝗔𝗿𝗲 𝗥𝗲𝘃𝗼𝗹𝘂𝘁𝗶𝗼𝗻𝗶𝘇𝗶𝗻𝗴 𝘁𝗵𝗲 𝗪𝗮𝘆 𝗪𝗲 𝗔𝗽𝗽𝗿𝗼𝗮𝗰𝗵 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 Traditionally, graduate students (myself included at one point) relied heavily on subject-matter experts, such as advisors, who could recommend papers from their vast memories. From there, it often took days or weeks to sift through articles to pinpoint relevant citations and build a coherent narrative. But now, tools like Undermind, an LLM-based prompt engineering startup created by MIT graduates, are reshaping this process. Here’s how a modern, closed-loop research workflow might look: 1. Ask a research question. Start by framing an inquiry or literature review topic with Undermind. Refine the question through dialogue with Undermind if necessary. 2. Source journal articles. Let Undermind pull relevant articles from pre-print databases like arXiv and many partnering publishers, complete with summaries tailored to the research question. 3. Dive deeper with OpenAI's GPT or Anthropic's Claude. Upload the most relevant articles to these LLMs and ask the refined research question to generate a synthesized literature review. 4. The result? 𝗔 𝗹𝗶𝘁𝗲𝗿𝗮𝘁𝘂𝗿𝗲 𝗿𝗲𝘃𝗶𝗲𝘄 𝗶𝗻 𝗵𝗼𝘂𝗿𝘀 𝗶𝗻𝘀𝘁𝗲𝗮𝗱 𝗼𝗳 𝘄𝗲𝗲𝗸𝘀. 5. Of course, don't forget to check the results against common sense, first principles, and/or feedback from a subject-matter expert!