JetBrains

With Central CLI, users of JetBrains AI for Teams and Organizations can monitor AI activity by agents running outside of IDEs. With most of our developers regularly working with 3–5 different AI tools, being able to track what’s being used for different jobs has led to better control over costs, streamlined admin, and granular usage limits. And JetBrains Central CLI has a surprising backstory. Learn more about it now. https://lnkd.in/e9ECubjw

The AI world loves buzzwords – but few of these terms are genuinely worth knowing. In this video, Kristel Cocoli walks through four concepts everyone should know: • Prompt engineering – how you phrase what you ask. • Context engineering – what the model actually sees, and why more context isn't always better. • Harness engineering – giving the model tools so it can read files, execute code, and react to results, turning a model that talks into an agent that works. • Loop engineering – the leap from prompting your agent yourself to designing the system that prompts it for you. Each step doesn't replace the one before it – it builds on it. The prompt is still inside the loop. The context is still assembled on every run. The harness still gives the agent its tools. What moves is the leverage point.

Still thinking about KotlinConf? So are we! As we begin planning KotlinConf 2027 (yes, really!) let's take a moment to reflect on this year's event: From launching Kotlin 2.4 with ergonomics, safety, and trust in mind, to introducing the Kotlin Toolchain for a unified entry point into Kotlin, there were many must-see moments over the two days. Google spoke about their use of Kotlin since the early days, with 92% of pro developers relying on Kotlin to build pure Android apps (plus the increasing adoption of Kotlin beyond Android). Spring took attendees through how to build production-worthy agentic systems using Spring AI and Spring Boot. None of this would have been possible without the 100 speakers, 17 partners, and 44 volunteers who supported 108 sessions and the 2,200 attendees who visited from all over the world. And now, on to shaping the agenda for next year 🚀 Did you attend KotlinConf 2026? What were your most valuable takeaways?

🚀 Build with Kotlin. Ship it everywhere. Join Shipaton 2026 and compete for the Ship Kotlin Everywhere Award. Reuse the Kotlin you know to build and launch a real app for Android, iOS, desktop, and web. Get support from idea to launch with: 💻 A 90-day subscription to IntelliJ IDEA Ultimate 🤖 60 days of free access to Junie, JetBrains’ AI coding agent 📘 A starter guide covering KMP basics, project structure, code sharing, libraries, and publishing 💬 Support from JetBrains Developer Advocates on Discord 🎙️ A live Q&A webinar on August 25 Build and ship between August 1 and September 30. Share your journey through a blog post, devlog, or social media updates to earn extra points. Learn more 👇 https://jb.gg/4lflo0

"Growth here isn't a straight line, it is a journey." Constant challenges makes for a more interesting career and that's exactly how many of us have built ours at JetBrains... and stayed for the long run. Including our CEO Kirill Skrygan, hear how he went from junior developer to CEO over 16 years.

The most common way to lose a hackathon? A good project with a confusing pitch. Judges from the JetBrains x Codex Hackathon on how to avoid it: 🎯 Start with the problem, not the tech ✂️ One thing that works beats five that don’t 💻 The demo is the pitch. Show it working, fast 🎉 Have fun. Judges can tell Big thanks to our judges from OpenAI, Nebius, Scarf, and Stateshift for their notes from the judging table. Full post: https://lnkd.in/dbtFZWnd #hackathon #developers #JetBrains #AI

We asked our team to walk the Loop Lobby and answer a few questions about their work. See what they had to say now.

A prompting technique claims that speaking to AI coding agents in compressed, caveman-style language can reduce output tokens by up to 65%. The JetBrains AI team put it to the test. In 86 agentic coding tasks and roughly 240 trials, caveman-style responses used around 8.5% fewer output tokens, with no detectable drop in task quality. Why so much less than advertised? In coding-agent workflows, most tokens go toward code, diffs, tool calls, and exact error messages. Caveman-style language leaves all of that unchanged. It only shortens the narration between actions, and there is very little to compress, so it is probably not a major cost-saving technique. But there may still be the practical benefits of less narration to read, faster scanning, and a clearer view of what the agent is doing. Read the full analysis: https://jb.gg/3br5o0

One prompt. Thousands of files. A complete interactive 3D experience. Our colleague Mike Hartington asked the latest Claude Opus model to transform a massive node_modules directory into something developers could visually explore. Working inside a JetBrains IDE, the model created the data-processing script, selected an approach for representing the results, and built an interactive Three.js experience where every file and directory could be navigated. The first working version was ready in minutes. What stood out wasn’t just the speed. Using Claude Opus 5.0 inside his IDE let Mike take a broad creative idea, have the model make the architectural and implementation decisions, and rapidly turn it into a functional experience. Try Claude Opus 5.0 in Junie or any JetBrains IDE.

Have you ever wondered what goes into creating a mascot? Turns out, it takes a lot of people and a lot of listening! Kodee, the Kotlin mascot, was shaped by our designers, copyeditors, marketers, researchers, localization specialists, and, of course, the Kotlin community. In this blog post, our Strategic Research team shares how their efforts helped the other teams make sense of community feedback, explore different creative directions, and understand not only what people preferred but also why. Read the story behind Kodee's creation: https://lnkd.in/gQnnNCbF