Answers are grounded in this paper and cite their sources.
Modern analytical instrumentation research is fundamentally built upon custom software to manage data input/output (I/O) and hardware control. However, development of such software requires extensive experience and expertise in software engineering, limiting the ability of many research groups to develop new instrumental prototypes. The advent of agentic, large-language model (LLM) Artificial Intelligence (AI) software engineering where AI handles all writing, testing, debugging, and execution of code under the direction and validation of the researcher (“Vibe Coding”) provides a mechanism to rapidly bridge this experience gap and give researchers functional, high-level code without necessitating expertise in software development. In this tutorial, we describe a lightweight, four-stage software development lifecycle (SDLC) for agentic coding (prompt engineering and architecture design, iterative agentic debugging, human-in-the-loop validation, and regression testing via simulation) and demonstrate it through the development of a control and analysis software suite for atmospheric pressure Ion Mobility Spectrometry (IMS). This process is easily expanded to any analytical instrumentation in the research laboratory through commonly available Python libraries. As the agent is guided by user prompts, prior knowledge of such libraries is helpful, but not required as modern agents can “plan” a process through both user input and open web search background information. With proper understanding of crafting AI prompts and using them in the SDLC, agentic coding promises to allow rapid prototyping in the research lab at minimal cost.
No takes yet. Share an insight, caveat, or question.
Davis et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: