Tag
Kent Beck argues that software engineering is not going away even as AI advances, because coding involves building trust, understanding, and connections that cannot be automated.
The article discusses common misconceptions about the true purpose of code review in software development, emphasizing it's more about knowledge sharing and team collaboration than just finding bugs.
The author emphasizes that the key to making AI agents write good code lies in following classic software engineering practices from 1990s textbooks: writing tests, doing CI/CD properly, top-down design, and modular decoupling.
A discussion on where to place guardrails to prevent AI coding agents from making unauthorized changes, exploring friction points at various stages of the deployment workflow.
Senior SWE-Bench is an open-source benchmark that evaluates AI agents on software engineering tasks requiring senior-level skills.
Criticizes the Vibe Coding style, pointing out that the imprecision of natural language requirements causes AI Agent outputs to require extensive rework, and advocates for spec-driven development to ensure safety and production quality.
Senior SWE-Bench is a new open-source benchmark designed to evaluate AI agents on realistic, underspecified software engineering tasks, emphasizing skills like intent alignment and code quality rather than overly detailed specifications.
A podcast interview with Kent Beck covers his career from early programming to Agile and TDD, emphasizing the importance of human skills in software engineering.
SWE-rebench leaderboard updated with new models (GLM-5.2, Qwen3.6, Gemma 4 31B, etc.) and an improved UI, showing performance rankings on software engineering tasks.
Explores the debate over whether AI will eventually replace human software developers, examining current capabilities and future trends.
An article discussing how software engineering best practices are evolving with the integration of LLM-based coding tools, offering guidance for developers.
Article comparing old C++ performance tricks with modern compiler capabilities, showing that compilers can now optimize naive code better than hand-tuned hacks. Includes benchmarks on AMD Zen 5 with Clang 21.
The author argues that ORMs are more detrimental than beneficial and that developers should focus on learning SQL directly to avoid issues like attribute creep and inefficient queries.
This paper introduces InterFLOPBench, a benchmark for evaluating LLMs on detecting floating-point errors in C code, finding that recent models achieve high F1-scores, though performance varies by error type.
Gergely Orosz announces an upcoming event with Kent Beck tomorrow.
This paper audits three performance-optimization benchmarks (GSO, SWE-Perf, SWE-efficiency) for coding agents, finding that runtime instability, scoring rules, and task coverage significantly affect reliability, and that many tasks are already solved by at least one public submission.
This paper catalogues five recurring MCP server architectural patterns observed across fifteen independently developed servers, providing a taxonomy with context, problem, solution, and consequences. It also documents anti-patterns, cross-cutting concerns, and quantitative evaluations including inter-rater reliability and transport overhead.
IBM Research introduces ScarfBench, an open benchmark for evaluating AI agents on cross-framework Java migration tasks, focusing on Spring, Jakarta EE, and Quarkus. The benchmark assesses whether migrated applications build, deploy, and preserve behavior, unlike traditional code generation benchmarks.
A tweet recommends watching Angie Jones' keynote on building AI agents, highlighting core engineering principles and mentioning Python frameworks like DSPy, Functai, and Mirascope.
The article explores how AI code generation excels at local code chunks but struggles with global program understanding, leading to excessive defensive checks. It examines whether programming language design can help, using an example of local reasoning assuring global properties.