PulseAugur
EN
LIVE 08:46:07

Microsoft open-sources code-testing-generator, outperforming GitHub Copilot

Microsoft has open-sourced code-testing-generator, a polyglot agent designed to write and verify unit tests for software development. This agent reportedly achieves a 92.1% task completion rate on Microsoft's internal benchmark, significantly outperforming GitHub Copilot's 78.9% rate on the same tasks. The key innovation lies in its Research-Plan-Implement pipeline, which allows it to analyze a repository's structure, language, and existing conventions before generating tests, leading to more reliable and context-aware results. AI

IMPACT This agent's ability to understand repository context and generate verified tests could significantly improve developer productivity and code quality, especially for complex or legacy codebases.

RANK_REASON Microsoft open-sourced a code-testing-generator tool.

Read on Mastodon — fosstodon.org →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

Microsoft open-sources code-testing-generator, outperforming GitHub Copilot

COVERAGE [2]

  1. MarkTechPost TIER_1 English(EN) · Michal Sutter ·

    Microsoft Open Sources code-testing-generator: a Polyglot Unit-Test Agent That Hits 92.1% Task Completion Versus 78.9% for Stock Copilot

    <p>Microsoft has open sourced code-testing-generator, a polyglot unit-test agent shipping in the MIT-licensed dotnet/skills repository. It reads a repository before writing anything — detecting the language, test framework, existing conventions, and the real build and test comman…

  2. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    Microsoft has open sourced code-testing-generator, a polyglot unit-test agent that achieves 92.1% task completion versus 78.9% for stock GitHub Copilot. It read

    Microsoft has open sourced code-testing-generator, a polyglot unit-test agent that achieves 92.1% task completion versus 78.9% for stock GitHub Copilot. It reads repositories to understand language, framework and conventions before writing tests. https://www. marktechpost.com/202…