The findings identify the developer's processing architecture as a variable the conceptual modeling tradition needs to account for in order to account for specification quality in AI development platforms.
The growing adoption of Large Language Models (LLMs) in Software Engineering has reinforced the expectation that coding activities can be largely automated. However, this perception may represent yet another historical search for a solution capable of eliminating the inherent challenges of software development. This article discusses the transition from a code-centered paradigm to Specification-Driven Development. We argue that artificial intelligence reduces some of the effort associated with writing source code, but it does not eliminate the complexity of developing professional software systems. Instead, it shifts this complexity toward domain understanding, requirements elicitation, specification development, validation, maintenance, and software evolution. Building on this perspective, we discuss the renewed centrality of Requirements Engineering, considering its implications for productivity and software quality, as well as risks associated with automation bias, ambiguity propagation, Specification Overfitting, and the accumulation of Specification Debt. Finally, we propose the Specification Paradox: the more capable artificial intelligence systems become at automatically generating software, the greater the dependence on correct, complete, verifiable, and explainable human-produced specifications. We conclude that the future of Software Engineering will depend not only on machines'ability to generate code, but also on humans'ability to correctly specify, evaluate, and evolve what is intended to be built.
Enterprise software requires specification governance to transform probabilistic AI generation into deterministic, auditable engineering, and the SGRM framework is introduced, which defines four-component specification contracts, constrains stochastic generation via deterministic validation, and integrates generation, verification, and governance into a closed-loop architecture.
: We are in a time of change in regards to the emergence of software development as we know it due to the growing number of developers using large language models (LLMs), which eventually will enable major shifts toward the "post-code" era in which software development will become less reliant on coding through using AI-driven development systems that accept natural language and high-level specifications as inputs. This research will analyze the impact of these AI assistants (e.g., GitHub Copilot, Gemini and GPT) through quantitative data collected from Stack Overflow Developer Surveys, GitHub Octoverse Reports, and JetBrains Developer Ecosystem Survey regarding how developers are currently embedding AI into their current practices and what it will look like moving forward. The research found out three things about how developers use Artificial Intelligence. These things are adoption of Artificial Intelligence satisfaction, with Artificial Intelligence the different ways developers are using Artificial Intelligence is changing. The results indicate that there is a distinct directional trend toward AI-native development environments, and that developers are in the midst of rapid change to adopt these tools.
P. Vijayakumar, Jegatheeswari Perumalsamy, Priya Ranjan Parida et al.· Proceedings of the 1st Inter...· 0 citations
Generative AI is increasingly permeating software engineering, enabling developers to generate functions, files, and even entire applications from natural language specifications. AI systems are also becoming more personalized, adapting outputs based on inferred user characteristics and interaction history. While personalization may improve the development experience, it raises concerns that generated software could be shaped by attributes of the developer rather than by task requirements alone. Prior work has shown that generative AI can produce biased software artifacts, but little is known about how developer identity can bias generated code. We characterize three dimensions through which inferred developer attributes can influence generated artifacts: interface design, template content, and code structure. First, through controlled experiments on 800 AI-generated websites, we find that age- and gender-related signals produce significant differences across all three dimensions. Second, we conduct an observational study and follow-up interviews with 20 participants who used AI to create a personal website to both examine how personalization impacts software artifacts in practice, and also to understand how programmers perceive the boundary between personalization and bias. Together, our results show that developer attributes can meaningfully influence generated software beyond stated requirements, highlighting a previously underexplored tension between personalization and fairness in AI-assisted programming.
AI assistants are changing software development, yet developers' thoughts and feelings during programming remain underexplored. To explore how hidden intentions, actions, tool choices, and emotions unfold during AI-assisted tasks, we conducted a mixed-methods study with 76 developers and propose the S-IASE model. This model captures four dimensions of a development State: Intention, Action, Supporting Tool, and Emotion. Through sequential pattern mining and interviews, we identified three key patterns. First, a "Trust but Verify" workflow reordered steps in the traditional programming paradigm. Second, developers exhibited stable emotional patterns yet revealed underlying self-criticism. Finally, text-only AI responses created a modality mismatch for procedural tasks. These patterns carry implications for training developers and designing context-aware, multi-modal AI assistants.
Yin-wei Wu, Z. Li, Kathryn T. Stolee et al.· SIGSOFT FSE Companion· 0 citations
Engineering systems are increasingly characterized by large, heterogeneous collections of technical documentation, including specifications, interface descriptions, and contribution records. While artificial intelligence techniques have been applied to document analysis, many existing approaches rely on opaque models that limit transparency and human trust. This paper presents a structured AI-based approach for deriving systemlevel understanding from engineering documentation by combining semantic abstraction, modular reasoning, confidence-aware outputs, and analyst validation. The approach emphasizes transparency and evidence-linked reasoning, enabling users to inspect intermediate representations and validate inferred relationships. A case study using a large-scale wireless systems documentation corpus and a focused Wi-Fi Aware worked example demonstrates how source-anchored reasoning can scale across extensive document sets while preserving human oversight.
Amrutha Moorthy· International Symposium on C...· 0 citations