←
Diffblue3,010 posts
Diffblue avatar

Diffblue

@diffbluehq

Diffblue Cover is an AI agent for automating the generation, maintenance and management of the Java unit testing process. #AgenticAI #aifordevelopers #Java

◷ Joined August 2016▧ 1,017 Photos and videos
Diffblue avatar
Diffblue@diffbluehq·3 Mar 2025

#AI has changed how almost everybody codes. But how useful are LLM-based AI assistants if the majority of the code generated needs human oversight, review and correction? We decided to put GitHub Copilot (Coding assistant) and Diffblue Cover (AI Agent) comparatively through their paces to see which is quicker, more accurate and reliable when testing Java code. 🔥The results are compelling. Find out what we did and read the report: 🔗diffblue.com/resources/diffb… #AI #AIforCode #GitHubCopilot #Java #CodeQuality #AIAgents #AIAssistants

Post attachmentPost attachmentPost attachment
Diffblue avatar
Diffblue@diffbluehq·19 Dec 2025

As 2025 draws to a close, the Diffblue team wants to thank our customers, partners, and community for an incredible year. To everyone navigating legacy codebases, chasing coverage targets, and shipping with confidence: we see you. Here's to a restful break and an even stronger 2026. Happy holidays from all of us at Diffblue. 🎄

0:33
Diffblue avatar
Diffblue@diffbluehq·12 Dec 2025

Just published: "Uplifting Test Coverage Out of the Box with Diffblue Cover" The article reveals why traditional approaches to test coverage fail at enterprise scale and how determinism changes everything. Key insight from our Head of Engineering, Thomas Given-Wilson : "Every test Diffblue Cover produces will compile and pass. There are no hallucinations, no failing tests, no manual fixes required. The reinforcement learning loop validates each test through actual execution before accepting it." Diffblue analyzes bytecode, not source code, discovering actual behavior rather than intended behavior. Read how you can achieve substantial coverage increases out-of-the-box, without configuration or manual intervention. hubs.ly/Q03XZ1km0 #EnterpriseJava #SoftwareEngineering #java #UnitTesting

Post attachment
Diffblue avatar
Diffblue@diffbluehq·10 Dec 2025

GitHub Chief Product Officer Mario Rodriguez, speaking at Microsoft Ignite, shared how custom agents are reshaping developer workflows and demonstrated Diffblue Cover as an example of purpose-built AI doing the heavy lifting.

1:08
Diffblue avatar
Diffblue@diffbluehq·4 Dec 2025

Diffblue is now a GitHub Copilot launch partner. GitHub just announced custom agents for Copilot and they selected Diffblue alongside HashiCorp, Dynatrace, JFrog, MongoDB, and PagerDuty to bring specialized AI capabilities directly into the developer workflow. What this means for enterprise Java testing: - Native distribution across GitHub's 100M+ developer ecosystem - Integrated into VS Code, CLI, and hubs.ly/Q03WV5yj0 - The deterministic alternative to probabilistic unit test generation methods As AI coding assistants become standard, the need for reliable, verifiable test automation only grows. We're excited to be part of this ecosystem. hubs.ly/Q03WV8mX0

Post attachment
Diffblue avatar
Diffblue@diffbluehq·1 Dec 2025

Struggling to reliably test static methods in Java? Mockito’s static mocking capabilities (introduced in 3.4.0+) finally make it possible to isolate static calls, control nondeterministic behavior, and build cleaner, more deterministic tests. This guide walks through best practices, real-world patterns, and pitfalls to avoid. ⬇️ hubs.ly/Q03WjLzT0 #java #mockito #UnitTesting #CleanCode

Diffblue avatar
Diffblue@diffbluehq·26 Nov 2025

Join us next week to see how Diffblue’s next-generation unit test platform makes enterprise test coverage systematic, autonomous, and actually achievable. This latest release includes: Guided Coverage Improvement: transform output codes into action plans with "dcover issues" Test Asset Insights: learns from your existing tests and patterns LLM-Augmented Intelligence: leverage your approved LLMs without new security reviews Building test coverage today means babysitting AI assistants and plateauing at 40%. Diffblue fixes that helping you achieve your coverage gates autonomously. 🚀 Register for Tuesday's demo: hubs.ly/Q03VXB0R0 📊 Get the benchmark study: hubs.ly/Q03VXFNN0 #TestAutomation #DevOps #EnterpriseJava #SoftwareTesting #AITesting

Post attachment
Diffblue avatar
Diffblue@diffbluehq·13 Nov 2025

What's blocking AI from transforming enterprise software development? It's trust. When AI coding assistants hover at 50-65% accuracy, developers spend more time verifying suggestions than actually coding. In this article, we explore why professional developers distrust AI tool accuracy, examine why current LLM solutions remain insufficient for enterprise needs, and reveal what it takes to achieve the 95%+ accuracy threshold needed for genuine productivity gains. Read more: hubs.ly/Q03T0PT50 #AICodeGeneration #SoftwareDevelopment #CodeQuality #DeveloperProductivity #EnterpriseAI #JavaTesting

Diffblue avatar
Diffblue@diffbluehq·11 Nov 2025

Why stay stuck at 40% coverage? 🎯 In case you missed it, we just launched Guided Coverage Improvement, transforming how enterprise teams break through coverage plateaus! Before: Hundreds of cryptic output codes (R013, E115, R026...) leaving developers asking "Where do I even start?" After: Intelligent issue aggregation that groups related problems, prioritizes by coverage impact, and delivers ready-to-execute prompts for your existing AI assistant: hubs.ly/Q03SB4dh0

Post attachment
Diffblue avatar
Diffblue@diffbluehq·6 Nov 2025

Diffblue Cover now addresses more of the key pain points Java enterprises face in unit test generation through three integrated capabilities.! Achieve your coverage goals in 90 days, leverage your existing test infrastructure, and work seamlessly with your approved enterprise LLMs. All with 20x productivity advantage over AI coding assistants: 👉 See how it all works together: hubs.ly/Q03R-zn50 #SoftwareEngineering #java #UnitTesting

Post attachment
Diffblue avatar
Diffblue@diffbluehq·30 Oct 2025

We benchmark every major LLM against our autonomous agent. Why? Because understanding architectural limits matters more than following hype cycles. But we absolutely respect what LLMs do well. They excel at understanding intent, suggesting patterns, adapting to coding styles. Recently, our benchmarks exposed what enterprises are learning the hard way: on closed-source code, even Claude's impressive 83% open-source coverage drops to 49%. Qodo generates 17.6× fewer tests in the same timeframe. Even Copilot with GPT-5 tops out at 74% coverage while requiring constant developer oversight. LLMs shine at creative problem-solving and contextual understanding. Deterministic agents excel at systematic execution and guaranteed results. The results? Tests that compile 100% of the time versus 58-88%. Coverage that scales up to 29× better annually. Unattended execution for hours versus continuous prompting. Sometimes the best AI strategy is knowing which tool fits which problem. Read the complete analysis: hubs.ly/Q03QZxgY0 #SoftwareEngineering #EnterpriseJava

Post attachment
Diffblue avatar
Diffblue@diffbluehq·7 Aug 2025

Reinforcement learning isn't just for games or chatbots. At Diffblue, it's how our AI learns to write Java unit tests. No labels. No prompts. Just feedback loops that improve coverage and catch regressions. xAI's Grok 4 shows where RL is going. We're already putting it to work. #AI #ReinforcementLearning #Java #UnitTesting #DiffblueCover

0:36