I recently canceled my JetBrains subscription. I'd renewed it every year for more than 10 years. I have a bitter feeling about it, so I want to explain how I got here, and say thank you on the way out.

Why I stayed for more than 10 years

For me, JetBrains made the best IDEs on the market, by a large margin and for a long time. Better than Sublime Text, and better than VS Code, which I could never fall in love with. I tried to switch many times, and JetBrains won every time.

The main reasons were Search Everywhere and code intelligence. Press Shift twice, and one box finds any file, class, symbol, or action. No other editor has copied it properly, even now. And the code intelligence let me dig deep into any codebase.

The second reason was the deal. For €177 a year, I got the whole suite, and I used it: PyCharm, PhpStorm, and WebStorm. You pay once a year and you're done, with nothing else to worry about.

And those tools helped me build a lot of projects. Projects that earned good money and gave my clients real value.

But even when I was a happy customer, something nagged at the back of my mind. The IDEs are built on Java, and sometimes they felt heavy. Each one took about 3 GB of disk space. Per editor! I always wished they were snappier and lighter, but that was never a dealbreaker. The interface looked good enough to me, and nothing else came close. I had nowhere to go.

That's how it stood for years. Then AI came and started to challenge the status quo, and the problems started piling up, one on top of the other, like a snowball.

The first red flag: AI

The first red flag appeared a couple of years ago, when coding agents started arriving.

With a slight delay, JetBrains began adding its own AI to the IDEs: AI Assistant first, and later the Junie agent. At first, I was excited. AI built into my favorite IDE by the people who made it. What could be better?

But unfortunately, despite having complete control over the IDE, its UX and its UI, they dropped the ball. They couldn't build AI into it in a way that worked properly, and its performance was sluggish. It looked like they just couldn't figure it out.

I wasn't the only one who felt it. AI Assistant collected so many one-star reviews on the JetBrains Marketplace that, in April 2025, JetBrains deleted a batch of them. Their own explanation: "Nuking several reviews at once without a heads-up looked shady." Today the plugin is rated 2.1 out of 5, and 848 of its 1,439 ratings are one star.

Then, in August 2025, came the new AI quota model. JetBrains' own announcement admitted that "the quotas for some plans are getting smaller." People reported burning through a month's quota in hours. One user called it "a price hike with a blindfold." JetBrains' answer on Reddit was painfully honest: "we had little to no data to come up with a proper pricing model."

Meanwhile, Cursor and Windsurf showed up and did it properly. Both are forks of VS Code, an interface I'm not the biggest fan of. And just like JetBrains, they're middlemen selling other companies' models. But they did it right, and the market noticed.

Windsurf was valued at $1.25 billion in 2024, and Google later paid $2.4 billion just to hire its leaders. Cursor went from $9.9 billion in June 2025 to $29.3 billion five months later, and in June 2026 SpaceX agreed to buy it for $60 billion. Not bad for a fork, huh? JetBrains stayed where it was.

Then there's Claude Code. In VS Code, it runs as a proper extension with a real interface, and it works far better. In JetBrains IDEs, the official Claude Code plugin has the worst UX you could wish for. The top review sums it up: "This is a button that opens a terminal and types 'claude', useless!" It's still marked beta, and it hasn't been updated since December 2025.

To be fair, Anthropic makes it, not JetBrains. I can't say why Anthropic made this plugin so bad. Maybe it's internal API limitations, maybe something else. But even then, I'm surprised JetBrains didn't notice that the plugin for the coding agent everyone's talking about looks like crap, and did nothing about it.

JetBrains could have built a proper native version itself. It didn't. Its own Claude Agent won't run on a Claude subscription, and in June 2026 JetBrains made OpenAI's Codex the default agent in AI Chat instead. A JetBrains developer advocate explained why in the comments: "The Anthropic models are just too expensive."

The thought I keep coming back to is: why don't they just copy good ideas from others? It's so easy now. They probably have an army of developers and a whole platform of products to build on. Instead, they keep reinventing the wheel, and their wheel comes out square.

Then Fleet died

Remember Fleet? It was JetBrains' answer to VS Code: a light, fast editor with the JetBrains brains behind it.

In December 2025, JetBrains killed it. In their own words: "Fleet did not succeed as a standalone product." Starting December 22, 2025, you couldn't even download it anymore.

But Fleet didn't fully die. It turned into Air. Dead Air, I might say.

The last nail: Air

Air is JetBrains' attempt at an agentic IDE, and they dropped the ball again.

Once more, the problem is Claude Code. At launch in March 2026, JetBrains said Air works with your existing agent subscriptions, "except Anthropic." To use Claude in Air, you need API tokens.

Then, on August 19, 2026, JetBrains announced: "You can now use Air with your existing Claude Pro, Max, or Team subscription." The announcement didn't last. JetBrains quietly retitled the post, cut the Claude section, and deleted the matching tweet. The only record left is a copy on the Wayback Machine.

A JetBrains engineer later explained on Hacker News that they "hit a hard stop with Anthropic." Maybe so. But then why announce it first? The feature request on YouTrack collected 220 votes, and JetBrains closed it as "Obsolete" on October 13. The official workaround is to run Claude Code in a terminal inside an editor tab. One user replied: "In 2026, I don't see the ability to use something in a terminal surface as good news." Neither do I.

And even setting Claude aside, Air feels clunky and awkward next to Conductor, Superset, Warp, OpenCode, Paseo, and so on, most of which are made by small open-source teams.

Now Air is going away too, at least as a standalone app. In September 2026, JetBrains CEO Kirill Skrygan wrote on X about the desktop app: "Yes, we are suppressing it in favor of Air in Intellij IDEs." JetBrains staff confirmed on Reddit that Air Desktop is "in maintenance mode." That's seven months after launch. To me, it looks like they were afraid Air would cannibalize their lineup of IDEs. Sounds like Kodak and digital photography.

So in less than a year, JetBrains dropped Fleet, turned it into Air, and then sidelined Air too.

The Kodak of IDEs

JetBrains is turning into the Kodak of IDEs. Their IDE moat is pretty much gone. People use IDEs less and less as a tool for writing and inspecting code, and what they expect now is a good agentic workflow management tool. JetBrains couldn't propose anything special here with their Air attempt.

The numbers point the same way. In the Stack Overflow Developer Survey, every JetBrains IDE I used lost share between 2025 and 2026. PyCharm went from 15.0% to 13.5%, WebStorm from 7.6% to 6.7%, and PhpStorm from 5.8% to 5.6%. Zed grew from 7.3% to 11.4%.

The money looks worse, too. JetBrains' Czech company, JetBrains s.r.o., went from a profit of CZK 2.48 billion in 2024 to a loss of CZK 315 million in 2025. And in October 2025, JetBrains raised prices: the All Products Pack for companies went from €779 to €979 a year per user. Good luck justifying that raise.

What I use now

Recently, I started using Zed to replace the JetBrains IDEs, for the times I still need to change code by hand or dig in and read it.

It does a great job. It works flawlessly, it's free, it's fast, it's written in Rust, and it doesn't eat gigabytes of disk space. It just kicks ass.

How fast? One Hacker News user timed opening the same small Python project: 60 seconds in a clean PyCharm, 20 seconds in VS Code, and 4.4 seconds in Zed.

A sad goodbye

It's all sad to witness. JetBrains is a big company, and it moves like a big, clunky elephant. Changing course is hard at that size, especially while the revenue is still large. I expect things to get worse, and unless they reinvent themselves, I don't think they'll make it.

I know I'm not the only one canceling. "I've been a happy paying JetBrains customer for 10+ years. This year I decided not to renew," wrote one developer under the Air announcement. Another: "Last week i cancelled my Jetbrains sub after a decade of daily driving it." For me, too, there's no value left in their products.

So, JetBrains, thank you. For more than 10 years, your products solved real problems and made my life easier. You sponsored open source and did a lot of good for the community. That's why this is sad to write. You stopped evolving, and AI is eating you.

I wish you all the best, though. Don't give up.