Software Engineering 3.0 (SE 3.0) is the name Towards AI-Native Software Engineering (SE 3.0): A Vision and a Challenge Roadmap gives to the era of software engineering it proposes, which it calls AI-native in contrast to the AI-assisted era before it. In SE 3.0 development is intent-centric and conversation-oriented: it is driven not by code but by intents — desired outcomes or goals a developer wishes to achieve — which human developers and their AI teammates clarify through back-and-forth conversation, after which the AI drives the code-creation loop by synthesising the intents into runnable software. The same paper presents this as a throwback to the essence of the discipline, which has always been to turn intents into high-quality software, with code only a means to that end.
Usage
The term belongs to a three-era periodisation the paper sets out. Software Engineering 1.0 was code-centric, with tools supporting traditional process activities and humans driving the entire process. Software Engineering 2.0, the present era, keeps that code-centric approach while adding AI models to support traditional activities (AI4SE); AI coding assistants are its prime example, and the human still drives the code-creation loop. SE 3.0 is set against both: AI is no longer an assistive tool but a transformative element, the AI becomes a teammate rather than a task-driven copilot, and models are expected to be knowledge-driven and efficient rather than data-driven and costly to train.
The paper grounds the term in a technology stack of five components, each named with a .next
suffix as an allusion to the future: Teammate.next for adaptive, personalised AI partnership,
IDE.next for intent-centric, conversation-oriented development, Compiler.next for multi-objective
code synthesis, Runtime.next for SLA-aware execution with edge-computing support, and FM.next for
knowledge-driven models trained with Curriculum Engineering. Because it is a
vision, the paper describes these by their desired attributes rather than by a concrete
implementation, and it names commercial vibe coding platforms as very early glimpses of the era
rather than as instances of it.
The Rise of AI Teammates in Software Engineering (SE) 3.0: How Autonomous Coding Agents Are Reshaping Software Engineering takes up the term from that vision and uses it for empirical work. It glosses SE 3.0 as agentic software engineering, in which developers collaborate with autonomous AI teammates through an intent-driven, conversational process; the agents work at the task level — reading codebases, planning changes, executing tools, running tests and submitting pull requests — while the developer’s role shifts to orchestration: setting goals, constraints and permissions and reviewing the final changes. It names Devin, GitHub Copilot Workspace, Google Jules, OpenAI Codex, Claude Code, Cursor DeepAgent and Genie as typical tools of the era. The same paper adds an intermediate step, SE 1.5 or predictive coding, for token-level assistance such as autocomplete that sits between SE 1.0 and SE 2.0, and treats autonomous coding agents opening pull requests on GitHub as evidence that SE 3.0 is already unfolding rather than still ahead. It argues that practices such as Agile, Scrum and DevOps were built for all-human teams and that SE 3.0 calls for new methodologies for orchestration, review and governance.
Related Terms
- AI-Native Software Engineering — a separately proposed account of AI-native software engineering, organised around the unit of work, the correctness model and the accountability model
- Software 3.0 — a similarly numbered but distinct term about programming large language models through prompts and context
- Curriculum Engineering — the training approach the SE 3.0 stack relies on for its models