ASE-26: a curriculum for agentic software engineering as a discipline

Researchers propose ASE-26, a comprehensive undergraduate curriculum for agentic software engineering, arguing that as AI agents increasingly handle 75-79% of programming tasks, the industry's critical gap is not better models but structured practitioner discipline. The curriculum represents an academic effort to formalize training in agent-directed development as a new professional discipline.

The shift from code-writing to agent-direction represents a fundamental restructuring of software engineering work. With data showing AI handles three-quarters of programmer activities and employment declining for young developers most exposed to AI, universities face pressure to redesign computer science education. ASE-26 addresses this by formalizing agentic software engineering as a teachable discipline rather than treating AI as a tool within existing curricula.

The curriculum's framing around the 'evolutionary spiral'—the co-evolution of intent and build between human and agent—signals a conceptual departure from traditional software engineering pedagogy. Instead of teaching students to write code, the discipline teaches them to effectively direct AI agents, understand their limitations, and iterate through collaborative problem-solving. This reflects industry reality: organizations like Anthropic document that most programmer work now involves directing AI rather than implementing algorithms.

For the tech industry, this curriculum addresses a critical mismatch between labor market demands and educational outputs. Companies struggle to find developers skilled in agent collaboration because no standardized training existed. By establishing pedagogy around this practice, universities can produce graduates immediately valuable to employers navigating AI-augmented development.

The emphasis on discipline outlasting 'specific capabilities of today's models' indicates forward-thinking curriculum design. Rather than teaching Claude-specific techniques, ASE-26 presumably teaches principles transferable across future models. This matters for long-term industry stability—structured frameworks allow organizations to adopt better models without retraining entire workforces. The curriculum's formalization signals that agentic development has matured from experimental frontier to legitimate professional practice requiring formal education.

• →AI handles 75-79% of programmer tasks, making agent-direction skills more valuable than traditional coding ability
• →The critical industry gap is practitioner discipline and structured frameworks, not model capability
• →ASE-26 formalizes agentic software engineering as an academic discipline with a 21-module curriculum structure
• →Young developers face 13% relative employment decline in AI-exposed roles, creating urgency for curriculum redesign
• →Curriculum design emphasizes principles transferable across future models rather than tool-specific techniques