The Semi-Executable Stack: Agentic Software Engineering and the Expanding Scope of SE

2025-07-10 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Advanced, extended

Summary

This paper, a keynote companion from the Agentic Engineering 2026 workshop, introduces "The Semi-Executable Stack," a six-ring diagnostic reference model for understanding the expanding scope of software engineering in the era of AI-based agentic systems. It argues that AI does not diminish software engineering's relevance but rather broadens its engineered object beyond executable code to include "semi-executable artifacts" like natural language prompts, workflows, controls, and organizational routines. The model spans executable artifacts, instructional artifacts, orchestrated execution, control systems, operating logic, and societal/institutional fit. The paper demonstrates the stack's diagnostic utility through three worked cases, reframes common objections to agentic systems as engineering targets, and proposes a "preserve-versus-purify" heuristic for adapting legacy software engineering processes. It emphasizes that the transition is mediated by socio-technical engineering, not solely by benchmark performance.

Key takeaway

For Software Engineers and Architects grappling with AI integration, recognize that your role is expanding, not shrinking. Focus on applying engineering discipline to semi-executable artifacts like prompts, workflows, and control systems, not just code. Your ability to design and manage these higher-level artifacts, ensuring traceability and governance, will be crucial for project success and organizational adaptation, as neglecting them can lead to project stalls despite strong underlying models.

Key insights

Agentic AI expands software engineering's scope to "semi-executable artifacts" beyond traditional code.

Principles

• Utility, not perfection, drives organizational change.
• Compounding availability beats isolated excellence.
• Broader creation multiplies the need for engineering.

Method

The Semi-Executable Stack is a six-ring diagnostic model for locating contributions, bottlenecks, or transitions in agentic software engineering, spanning from executable code to societal fit.

In practice

• Design explicit control systems for agent reliability.
• Push lifecycle discipline outward to prompts and workflows.
• Shift attention to outer-ring artifacts like specification and architecture.

Topics

• Agentic Software Engineering
• Semi-Executable Stack Model
• Semi-Executable Artifacts
• Software Engineering Scope
• Organizational Operating Logic

Best for: AI Scientist, Research Scientist, Software Engineer

Related on AIssential

• The Semi-Executable Stack: Agentic Software Engineering and the Expanding Scope of SE — Software engineering's scope expands beyond code to semi-executable artifacts, requiring new models for understanding AI's impact.
• Beyond Vibe Coding: The Artifacts Layer — Durable artifacts are essential for preserving engineering intent in agentic AI development, enabling true delegation.
• From IDEs to AI Agents with Steve Yegge — AI is fundamentally reshaping software engineering, shifting focus from manual coding to agent orchestration and rapid prototyping.
• The Agent Stack - Part 1: A Systems Map of Modern Agent Infrastructure — Deconstructing the overloaded "agent" concept into a defined stack clarifies system responsibilities and failure points.
• An Ex-Meta L8’s Agentic Engineering Setup — High-productivity agentic engineering stems from manager-like delegation to AI agents, supported by specialized, agent-agnostic tooling for planning, execution, and validation.
• Build, Reuse, or Hybrid? How Orchestration Powers Agentic AI — Effective orchestration is crucial for integrating and coordinating agentic AI components, whether built, reused, or hybrid.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.