Develocity (@gradle) header
Develocity (@gradle) profile picture

Develocity

@gradle

Release announcements, tips, and events focused on boosting developer productivity with Develocity and Gradle.

 San Francisco, CA
develocity.ai
Joined May 2009
  • Tweets6.2K
  • Following234
  • Followers38K
1.3KMedia
Photo by Develocity (@gradle)Photo by Develocity (@gradle)Photo by Develocity (@gradle)Photo by Develocity (@gradle)Photo by Develocity (@gradle)Photo by Develocity (@gradle)Photo by Develocity (@gradle)Photo by Develocity (@gradle)Photo by Develocity (@gradle)Photo by Develocity (@gradle)
Gradle 9.8.2 is out 🐘 🔒 Fixes 3 high-rated security vulnerabilities (upgrade recommended) ☕ Java 27 support 🪞 Reuse Maven mirror settings ⚡ Up to 45% faster builds on some Windows machines docs.gradle.org/9.8.2/releas…
Gradle 9.8.2 Release Notes

Gradle 9.8.2 Release Notes

We are excited to announce Gradle 9.8.2.

docs.gradle.org
1
4
69
2,858
Gradle 9.9.0 deprecates the built-in Ant integration, and Gradle 10 removes it. Never call `ant`, use `AntTarget`, or import org.apache.tools.ant? You're done. Otherwise: blog.gradle.org/side-quest-r… #WhatsNew #Gradle
Gradle blog image, Side Quest: Removing Ant. A terminal shows gradle --version output with the Ant line struck through in red; beside it, a grid labelled 58 imports, 12 features.

ALT Gradle blog image, Side Quest: Removing Ant. A terminal shows gradle --version output with the Ant line struck through in red; beside it, a grid labelled 58 imports, 12 features.

9
1,123
📣 Gradle 9.9.0 RC1 is out and ready for testing 🔬 docs.gradle.org/9.9.0-rc-1/r… Bug fixes and… ⚡Performance improvements for Kotlin DSL compilation 🎯 Test filtering for non-method JUnit Platform tests 📦 Wrapper settings preserved on upgrade 🗂️ Predefined groups in the Problems API Please report any 🐛 on GitHub!
1
9
1,181
Your agent tweaks one file and triggers the full test suite. What’s the daily cost? A full run can take 15 minutes, and more agents mean more runs. Test history picks what your change can break. Skip the rest, finish sooner. develocity.ai/product/predic… #AskingNow #Develocity
Screenshot of a Hacker News comment: an agent ran full test suites over and over, each costing about 15 minutes, adding up to about 2 hours. A reply suggests telling it in AGENTS.md not to run the full suite unless asked.

ALT Screenshot of a Hacker News comment: an agent ran full test suites over and over, each costing about 15 minutes, adding up to about 2 hours. A reply suggests telling it in AGENTS.md not to run the full suite unless asked.

533
You wait for Gradle's configuration phase every build. But why? New project? A single line, org.gradle.configuration-cache=true, lets unchanged builds skip it. docs.gradle.org/current/user… #GradleDidThat #Gradle
4
642
Fixing a bug you can't reproduce just hides it until it returns, worse. So what should the agent do? Reproduce it from build data first, prove the fix. develocity.ai/blog/debugging… #BossFight #Develocity
8-bit game screen. Boss: Phantom Executor, full HP bar. Appears only when all 3 line up: composite build, config cache hit, test selection on. Log: P1 try 1 miss, try 2 miss, try 3 unclear, tries 4-5 miss, saves a handoff. P2 all 3 at once: hit.

ALT 8-bit game screen. Boss: Phantom Executor, full HP bar. Appears only when all 3 line up: composite build, config cache hit, test selection on. Log: P1 try 1 miss, try 2 miss, try 3 unclear, tries 4-5 miss, saves a handoff. P2 all 3 at once: hit.

1
1
552
Release 1.1.0 of the Gradle Skills plugin is out. It fixes an issue when rolling back a Gradle Wrapper upgrade and adds minor improvements to help weaker models. github.com/gradle/gradle-ski…
Release 1.1.0 · gradle/gradle-skills

Release 1.1.0 · gradle/gradle-skills

Minor improvements to make the Best Practices skill work better on weaker agents. Fixes potential copy problems when rolling back using the Wrapper Upgrade skill.

github.com
11
1,112
Every Gradle 7 release has a known high-severity advisory, and your build has no patch. Now what? A rushed major upgrade costs time. The Security Subscription patches 7.6.6, so teams patch where they are and upgrade later. blog.gradle.org/gradle-secur… #StateOfTheBuild #Gradle
The gradle.org Security Subscription coverage table: versions, release lines, open source vs Subscription fix coverage.

ALT The gradle.org Security Subscription coverage table: versions, release lines, open source vs Subscription fix coverage.

1
641
Your root build script configures every subproject, and Isolated Projects fails the build. What replaces it? Convention plugins: each project applies shared logic to itself. docs.gradle.org/current/user… #ReadingTheBuild #Gradle
5
885
Your Docker build just redid work you thought was cached. The log doesn't say why. One line change? Every layer after it rebuilds. Every time. Develocity 2026.3 shows which step missed the cache so you can fix it. develocity.ai/releases/2026.… #ShipNote #Develocity
Develocity 2026.3

Develocity 2026.3

Release notes and updates for Develocity 2026.3

develocity.ai
3
656
"So we call it dependency hell." Sean Sullivan of Grubhub at a recent summit. His defenses: `failOnVersionConflict()`, dependency locking, and a bill of materials to keep libraries aligned. Which one does your build use? #FromTheStage #Gradle
Sean Sullivan of Grubhub presenting his dependency management talk at Code Remix Summit, speaking from the podium.

ALT Sean Sullivan of Grubhub presenting his dependency management talk at Code Remix Summit, speaking from the podium.

1
718
Two clean PRs pass, get merged, and break main. Everyone's stuck. Why? Each was tested alone, not together. A broken main halts all merges. Uber's merge queue tests them as a batch first: 53% less CI usage, 37% shorter waits. dpe.org/sessions/dhruva-julo… #SummitTape #Develocity
1
771
Build failed, but your code is fine? Don't debug for hours. Flaky tests trick agents into breaking working code. Failure Handler shows the agent what's wrong so it only fixes real bugs. develocity.ai/releases/2026.… #ExplodedView #ContextEngineering
Screenshot of the Failure Handler run from the 2026.3 release notes: checks the last 7 days of builds, finds the root cause, fixes it and verifies. Numbered markers 1 to 3 point to each.

ALT Screenshot of the Failure Handler run from the 2026.3 release notes: checks the last 7 days of builds, finds the root cause, fixes it and verifies. Numbered markers 1 to 3 point to each.

1
661
Your IDE sync crawls, but turning on Isolated Projects feels too risky for your CI. Can you split it? Set the flag in gradle.properties and turn it off on the CI command line, just like the Gradle team does. blog.gradle.org/isolated-pro… #OneFlag #Gradle
Screenshot of GitHub issue #37026, Isolated Project Support for Syncs Only, opened Mar 10: asks for a flag that enables Isolated Projects for syncs only, because org.gradle.unsafe.isolated-projects=true is currently a build-wide switch.

ALT Screenshot of GitHub issue #37026, Isolated Project Support for Syncs Only, opened Mar 10: asks for a flag that enables Isolated Projects for syncs only, because org.gradle.unsafe.isolated-projects=true is currently a build-wide switch.

3
889
Your fresh CI runner downloads hundreds of megabytes before any test runs. Why? Stalled downloads fail jobs. Develocity 2026.3 caches them nearby, Python setup is up to 40% faster. develocity.ai/releases/2026.… #ShipNote #Develocity
Develocity 2026.3

Develocity 2026.3

Release notes and updates for Develocity 2026.3

develocity.ai
707
You're on Gradle 9.4.0, and the security fix ships only in the latest minor. Now what? Open-source fixes only reach the latest minor. The Gradle Security Subscription covers 9.2.x to 9.8.x: critical fixes within 7 days, high within 30 days. gradle.org/security-subscrip… #AskTheDocs #Gradle
gradle.org Security Subscription: the open-source promise (fixes limited to the latest minor), the coverage table showing 9.2.x to 9.7.x covered only by the subscription with a 7/30-day SLA, and what's in the subscription.

ALT gradle.org Security Subscription: the open-source promise (fixes limited to the latest minor), the coverage table showing 9.2.x to 9.7.x covered only by the subscription with a 7/30-day SLA, and what's in the subscription.

4
958
Your agent passes its checks, but bugs still ship. More eyes on the PR won't help. An AI that writes its own tests can pass them and still be wrong. Use the test suite it didn't write, so green means go. develocity.ai/blog/regressio… #FromTheFounder #ContextEngineering
Why Regression Tests Will Remain Essential, Even with Smarter AI and Formal Verification | Develo...

Why Regression Tests Will Remain Essential, Even with Smarter AI and Formal Verification | Develo...

No AI, however smart, can guarantee that a change breaks nothing that worked before. With agents making more changes, regression tests matter more, not less.

develocity.ai
1
2
605
New IntelliJ IDEA Tech Talk episode! @wolfs42, Principal Software Engineer at Develocity, joins @antonarhipov to investigate unnecessary task runs, check test history, and see how AI handles build problems in IntelliJ IDEA. 👉️ youtu.be/xBNBjYH-foE 👈️
7
36
4,783
Agents can now upgrade your Gradle wrapper correctly using our new skill. We tested 7 models. How did they score? 0 of 7 models pinned it alone. With our skill, 4 scored 29/29. See the ranking: blog.gradle.org/introducing-… #HeadToHead #Gradle
Bar chart titled wrapper-upgrade skill test results: 7 models, each with a without-skill bar (grey) and a with-skill bar (teal). Claude Opus 5.5 16 and 29, Gemini 3.8 Flash 15 and 29, DeepSeek V4 Pro 13 and 29, GPT 5.6 Luna 13 and 29, Claude Sonnet 5 12 and 28, Claude Haiku 4.5 10 and 26, Qwen 3.6 35B 12 and 17.

ALT Bar chart titled wrapper-upgrade skill test results: 7 models, each with a without-skill bar (grey) and a with-skill bar (teal). Claude Opus 5.5 16 and 29, Gemini 3.8 Flash 15 and 29, DeepSeek V4 Pro 13 and 29, GPT 5.6 Luna 13 and 29, Claude Sonnet 5 12 and 28, Claude Haiku 4.5 10 and 26, Qwen 3.6 35B 12 and 17.

2
1
17
963
Your build passes, but it misses the cache on work that didn’t change. Logs don’t say why. Two bytes change upstream and every downstream cache key resets, so you pay on every run. Build Scans show exactly which input changed. develocity.ai/blog/develocit… #WhatWeLearned #Develocity
Screenshot of the Build Scan comparison from the javac blog: same task in both builds, 2 class files differ, so the cache keys differ. Numbered markers 1 to 3 point to each.

ALT Screenshot of the Build Scan comparison from the javac blog: same task in both builds, 2 class files differ, so the cache keys differ. Numbered markers 1 to 3 point to each.

623
Load more