Mar 2026· arXiv.org· Vol abs/2603.11073· 0 citations· 13 references
Computer Science
TL;DR
An experience report from a small full-stack team that applied contextual prompting and explicit architectural constraints to build a multi-project agent learning platform designed for sustained, production-oriented use and an academic retrieval-augmented generation system is presented.
Abstract
Code-generating tools are increasingly used in software development, yet experience reports on conversational"vibe coding"under production constraints remain limited. This paper presents an experience report from a small full-stack team that applied contextual prompting and explicit architectural constraints to build (i) a multi-project agent learning platform designed for sustained, production-oriented use and (ii) an academic retrieval-augmented generation system. The agent platform supports multiple isolated projects, each with structured memory and background processing, thereby enforcing project-level isolation. The RAG system provides citation-grounded answers, role-based access control, and evaluation tracking. Across both systems, vibe coding accelerated scaffolding and integration. However, the generated code often under-specified isolation rules and infrastructure constraints when these were not explicitly defined. Consequently, aspects such as multi-tenancy, access control, memory policies, and asynchronous processing required deliberate architectural design and verification. We observe a shift in engineering effort from boilerplate implementation toward constraint specification and enforcement auditing. We also identify recurring architectural"non-delegation zones"where conversational code generation remains insufficient for production reliability.
An exploratory review of the emerging gray literature, which largely agrees on what a well-engineered loop contains: triggered agent runs bounded by machine-checkable stop conditions, persistent state files, verifier sub-agents, token budgets, and defined points of escalation to humans.
Jai Lal Lulla, Vahram Nersesyan, Seyedmoein Mohsenimofidi et al.· 0 citations
ICAE-Bench, a benchmark for evaluating coding agents under interactive project-building settings, starts from a fuzzy product requirement, simulating the dynamic paradigm with an automated User Agent, and introduces three key designs.
Zhongyuan Peng, Dan Huang, Chuyu Zhang et al.· arXiv.org· 3 citations· ⚡1
This paper reviews recent empirical literature to ask what the developer's job is shifting from typing code to directing agents that type code a change often summarized as a move from code generation to code orchestration.
P. N. Nesarajan, P. Thenmozhi, Shenbaga Priya et al.· International Journal of Inn...· 0 citations
The widespread availability of generative tools has weakened a long-standing assumption in computing education: that the production of working code can serve as a proxy for student competence. In resource-constrained settings, these tensions are compounded by intermittent power, high data costs, and emergent institutional governance. We report a two-site qualitative study of Nigerian computing departments (n = 20), drawing on semi-structured interviews with students and academic staff and analysing the corpus through thematic analysis to characterise assessment practice under policy-light conditions. Our findings describe a persistent detection trap, where staff rely on inconclusive software or subjective judgement, and institutional silence, where expectations for acceptable use are unevenly specified and applied. We contribute the Scaffolded AI-Verification Framework (SAVF), presented as a traceable design pattern catalogue of handset-first, low-data feasible teaching moves derived from these stakeholder accounts. SAVF comprises (i) permitted-help statements with disclosure, (ii) process-evidence bundles that foreground explanation and testing, and (iii) course-anchored prompts that require adaptation to local materials and constraints. We provide three pattern specifications, a traceability table linking themes to requirements and patterns, and adoption guidance for low-bandwidth implementation, positioning SAVF as a stakeholder-informed design contribution with a testable evaluation plan for future in-situ study rather than as an evaluated intervention.
Kehinde D. Aruleba, Kike Ladipo, I. Sanusi et al.· International Computing Educ...· 0 citations
A high-quality benchmark of 1,000 code refinement instances from 328 Python, Java, and JavaScript repositories that focused on one of the most challenging code refinement scenarios that strictly requires repository-level knowledge reasoning, and a straightforward method, RepoRefiner, which retrieves repository-level context by collecting the full file content, extracting definitions of its identifiers, and summarizing these contexts to support code refinement.
Ke Wang, Peng Lan, Jiakun Liu et al.· ACM Transactions on Software...· 1 citation
SWE-Touch is introduced, a framework that stress-tests this setting through validated Counter-Edits: plausible edits to task-relevant code that conflict with task completion, and point to detecting workspace changes, reconciling conflicting edits with the task, and verifying the affected behavior as key capabilities for future optimization.
Yuqiao Tan, Jinxiang Meng, Fangyu Lei et al.· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduAug 27, 2026
A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.
MIT News · Artificial Intelligence· news.mit.eduAug 3, 2026
