Spatial coding might be the next big thing. Our HAX (Human-AI Experience) researchers have been mapping the future of AI and XR for tech creators. See what 13 expert interviews revealed across 5 major themes, and explore the potential for future work 👉 jb.gg/pkrm49
Hot take: ! is one of the most dangerous characters in TypeScript.
Non-null assertions silence type errors instead of fixing them. They pile up, they hide bugs, and the ESLint rule for it often gets turned off because it can't tell a legit assertion from a dangerous shortcut.
Prompt agent like caveman?
Our evaluation shows that in real tasks, this technique does not compromise quality, but only manages to reduce token use by about 8.5% – much less than advertised.
Detailed analysis here: blog.jetbrains.com/ai/2026/07/spe…
Benchmark scores are tempting, but they can mislead. In our new post, we evaluate common benchmarks to demonstrate where they fall short and propose practical improvements for more meaningful evaluation. Read the full analysis and recommendations: jb.gg/weekpq
A new Air release is out
From now on: add your favorite agents like Github Copilot, Opencode, and 17 others, and run local models through the ACP.
With this release Air also gets Java language support, powered by IntelliJ IDEA code engine.
Let us guide you through it🧵
Codex by @OpenAIDevs is now the recommended agent in JetBrains AI.
A recommended agent is just a starting point – switch anytime. As models evolve, we'll re-evaluate next month. Here's how we made the decision: 🧵
AI ROI isn’t just about individual developer productivity.
Kris Kang, Chief Product Officer for JetBrains Agent Systems, unpacks why we’re looking at ROI measurement at the organization level, tailored to each enterprise.
Dive deeper into JetBrains AI for Teams and
AI governance is a bit of a hot potato right now.
At London Tech Week, JetBrains Agent Systems CPO Kris Kang broke down the AI trends people are overhyping, what they’re completely overlooking, and why automation increases the need for skilled developers. Watch the clip below.
The smartest models don’t necessarily guarantee the best results.
Vladislav Tankov, our CTO, explains how with JetBrains AI for Teams and Organizations, you can maintain quality and manage costs with smaller models through context management, code verification, and reinforcement
Privacy risk in LLMs is more serious than most people think — especially for fine-tuned models.
With nothing but API access, membership inference attacks can reveal whether specific data was used to train a model. We've been studying this risk, and our new paper proposes EZ MIA:
We could have made another shiny Three.js demo.
Instead, we gave @OpenAI’s GPT 5.6 a vague CEO message from @kskrygan, a JetBrains IDE in the hands of @wordmandotdev , and five minutes to build something useful.
GPT 5.6 is now available in JetBrains IDEs, Junie, and Air.
Will AI make Java IDEs obsolete? @martinlippert (Spring Tools) joins The Marco Show to discuss AI coding agents, developer tools, and the future of Java development.
🎥 youtu.be/gUm6DOf1NHI