A new transformation is underway in software engineering, driven by the rapid adoption of generative AI in development workflows. Similar to how version control systems once automated manual coordination, AI tools are now beginning to automate many aspects of programming. For embedded software engineering organizations, however, this marks their first experience integrating AI into safety-critical and resourceconstrained environments. The strict demands for determinism, reliability, and traceability pose unique challenges for adopting generative technologies. In this paper, we present findings from a qualitative study with ten senior experts from four companies who are evaluating generative AI-augmented development for embedded software. Through semi-structured focus group interviews and structured brainstorming sessions, we identified eleven emerging practices and fourteen challenges related to the orchestration, responsible governance, and sustainable adoption of generative AI tools. Our results show how embedded software engineering teams are rethinking workflows, roles, and toolchains to enable a sustainable transition toward agentic pipelines and generative AI-augmented development.

Literate programming, a concept introduced by Knuth in 1984, emphasized the importance of combining human-readable documentation with machine-readable code as writing literate programs is a prerequisite for software quality. Our objective with this paper is to evaluate whether generative AI models, Large Language Models (LLM) like GPT-4, LLaMA or Falcon, are capable of literate programming because of their extensive use in software engineering. To truly achieve literate programming, LLMs must generate natural language descriptions and corresponding code with aligned semantics based on user prompts. In addition, their internal representation of programs should allow us to recognize both programming languages and their descriptions. To evaluate their capabilities, we conducted a study using the Rosetta Code and CodeNet repositories. We perform four computational experiments using the Rosetta Code repository, encompassing 1,228 tasks across 926 programming languages, and validate our findings on the larger CodeNet dataset, which includes 55 tasks and 52 languages. Our findings show that LLMs in the trillion-parameter class are capable of literate programming, while models in the million- and billion-parameter classes are better at recognizing programming languages than tasks. Based on these results, we conclude that modern LLMs inhibit a deeper ability to encode programming languages and the semantics of programming tasks, bringing us closer to realizing the full potential of literate programming.

Developing autonomous driving systems (ADSs) involves generating and storing extensive log data from test drives, which is essential for verification, research, and simulation. However, these high-frequency logs, recorded over varying durations, pose challenges for developers attempting to locate specific driving scenarios. This difficulty arises due to the wide range of signals representing various vehicle components and driving conditions, as well as unfamiliarity of some developers’ with the detailed meaning of these signals. Traditional SQL-based querying exacerbates this challenge by demanding both domain expertise and database knowledge, often yielding results that are difficult to verify for accuracy.

This research explores the transformative potential of generative AI in software development. Generative AI is revolutionizing the field by offering capabilities to automatically generate, refactor, and test code. Through the use of action research, new methods and tools based on generative AI models are studied and developed. The initial focus is on the models' ability to comprehend high-level design concepts. Subsequently, the research moves into the augmented generation of software artifacts. Finally, organization-specific or task-specific methods are introduced to enhance software developers' productivity and experience.