Why Every System Integration Needs a Named Failure Owner Before Go-Live

·5 views

Most integration designs focus on the success path, leaving ambiguous failure states unassigned and unresolved. A timeout or partial response does not confirm whether a business action — such as an order being scheduled or stock allocated — actually completed. Engineers are advised to model each handoff as a series of distinct business states rather than a single boolean flag like 'synced'. Using stable idempotency keys tied to the intended action, not the network attempt, helps prevent duplicate effects during retries. Reconciliation processes and clearly separated error categories — transport failure, incomplete acceptance, and business rejection — should be built in before a system goes live.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Developer Builds Self-Hosted Ticket System to Test Real-World AI Coding Agents

A developer created Lutions, a self-hosted web application for project and ticket management, partly driven by rising software licence costs and concerns over digital sovereignty. The project grew beyond its original scope into a personal laboratory for testing agentic AI in a realistic, mature codebase rather than a simplified demo environment. The developer argues that useful AI coding agents require more than better prompts — they need access to a project's full context, conventions, history, and constraints. Drawing parallels to benchmarks like SWE-bench, the article highlights that real-world agentic development demands agents navigate multi-file changes, existing rules, and organisational expectations. The piece frames Lutions as a field study in what conditions must exist for AI agents to contribute meaningfully to established software projects.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Engineering Discipline Can Help AI Generate Production-Quality Code

A software consultant and AI advisor describes a methodology for producing production-ready code using AI without directly writing or reviewing the code during development. The approach stems from a personal experiment he began roughly two years ago, driven by concerns that rapid AI advancement would erode his technical expertise. He argues that while the method consumes more tokens upfront, it reduces rework and lowers the total cost of ownership over time. The technique is rooted in engineering discipline, requiring attention to correctness, security, performance, and observability rather than relying on 'vibe-coding.' He has published a free tutorial and an open-source skills framework on GitHub to help others adopt the approach.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

EU Cyber Resilience Act Mandatory Vulnerability Reporting Begins September 11

The European Union's Cyber Resilience Act (CRA) activates its mandatory incident reporting requirements on September 11, 2026, affecting software makers and device manufacturers selling products in the EU. Under the new rules, companies must issue an early warning within 24 hours of discovering an actively exploited vulnerability, followed by a formal notification within 72 hours and a final report within 14 days. Reports must be submitted to the relevant national CSIRT and to ENISA through a Single Reporting Platform, with the clock starting the moment a team becomes aware of an issue. The regulation targets any manufacturer placing a product with digital elements on the EU market, including apps, firmware, and connected devices, though pure SaaS services generally fall under the separate NIS2 directive. Full CRA compliance, including CE marking, is not required until December 2027, but the reporting obligations take effect immediately this Friday.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Guide: Building an AI Agent That Completes and Submits Gig Platform Jobs

A developer tutorial published on DEV Community outlines how to build an autonomous AI agent capable of completing paid tasks on freelance platforms like Upwork and Fiverr. The system connects a large language model (LLM) chain — built with tools like LangChain and OpenAI's API — to a gig platform's REST or webhook interface to receive, process, and submit jobs automatically. The guide recommends keeping the agent's scope narrow, such as writing SEO blog posts or converting design files to code, to reduce errors and simplify pricing. Sample Python code is provided for the LLM prompt chain and a FastAPI-based webhook handler that listens for new job postings and returns completed work. The tutorial also addresses cost trade-offs, noting that managed APIs like GPT-4o-mini cost roughly $0.002–$0.01 per call, while self-hosted models offer savings at the expense of greater infrastructure complexity.

0 comments Read more at DEV Community