--- title: "I walked away from WaveBox and JetBrains in the same month" url: "https://alex.zappa.dev/blog/tool-migration-wavebox-jetbrains/" description: "Ten years of WaveBox, JetBrains paid through 2028, and why I switched to Zen and Zed anyway." --- # I walked away from WaveBox and JetBrains in the same month August 14, 2026 * [#tools](/tags/tools/), * [#browser](/tags/browser/), * [#ide](/tags/ide/), * [#oss](/tags/oss/) Ten years is a long time to use one browser. I know because I just stopped. Two major tool switches in three weeks. Not because anything broke. Because everything got worse, month by month, and I finally stopped lying to myself about it. But the real catalyst wasn’t WaveBox or JetBrains. It was Claude. ## The Claude moment Back in June 2026, Anthropic dropped Fable 5. It was the superior model, a genuine leap forward, and it ran fast. For about three days. Two or three weeks before the launch, Claude Code started crawling. My hardware was the same. Nothing on my end changed. But their API was dragging: requests queuing, latency spiking, the whole thing felt like wading through mud. I checked my network. I checked my machine. Everything was fine. Then Fable 5 and Opus Fast landed, and suddenly the same API was flying again. Call it a coincidence if you want. I don’t. It felt like the old models got throttled right before the new ones shipped so everyone would step on the gas and feel the difference. Oh, so this is what we were missing. Sign here. That’s either brilliant product management or the kind of dark pattern that makes you rethink who you’re building on top of. Maybe both. On June 12, the US government applied export controls to Fable 5 and Mythos 5. Anthropic suspended access for everyone, globally, because they had no way to verify nationality in real time. One minute the model was there. The next minute it was gone. That was the wake-up call. I’m in the EU. I ship products under the FreshJuice brand. None of them are AI products. They’re regular software. But their development pipeline runs on Claude: PR reviews, changelogs, documentation, the boring stuff that happens between writing code and shipping it. If Claude goes down, the products keep working just fine. The velocity doesn’t. The PRs pile up. The changelog sits empty. The docs fall behind. That’s not “oh no, my Copilot is gone.” That’s your delivery rhythm breaking for a week because one vendor flipped a switch. I have ADHD. My brain doesn’t move in straight lines. It jumps between ten things at once and half of them don’t make sense even to me. AI doesn’t write my code. It polishes the chaos into something coherent. It groups the noise into something useful. Without it, I’d ship a lot more garbage. With it, the garbage gets filtered before it lands. Fable 5 came back on July 1 after the export controls lifted. But the lesson didn’t go away: this whole ecosystem is a single point of failure. If the one company you depend on sneezes, you’re out of commission for weeks. That thought stayed with me. It made me look at every other tool I was locked into and ask the same question: what happens when this one goes sideways? Spoiler: this post covers the browser and the IDE. The AI agent itself, the thing I use to write, review, and ship, is a separate story. That one’s coming. The answer, for WaveBox and JetBrains, was already obvious. I just hadn’t acted on it yet. ## WaveBox: day one to done Thomas Beverley built WaveBox on the ashes of WMail. I was there for the funeral. WMail was his email client, and when he decided it was time, he sunset it and launched WaveBox instead. The pitch was exactly what I wanted: a fast, responsive desktop app that ran web services as PWAs with real integration between them. Minimal, not minimal-looking. Functional, not feature-bloated. It felt like a genuine breath of fresh air. The Electron years were solid. Then came the Chromium migration, and honestly, that was the golden era. Suddenly WaveBox was a real browser. Chrome extensions worked. Everything was fast. I was a paying subscriber from day one and never questioned it. Then the AI update landed, chasing tomorrow. I wanted to love it. Native AI inside the browser I already live in? Sign me up. What shipped was a janky sidebar that got in the way more than it helped. And that pattern became the rule: every release added something I didn’t ask for. The app got heavier. Glitchier. Optimizations that somehow made everything slower. Features nobody requested while the core experience rotted around the edges. I didn’t stay for the grandfathered subscription. I stayed because I didn’t want to leave my comfort zone. The Veteran plan, or Grandpa plan, whatever they called it, was a classy touch, and I genuinely tip my hat to Thomas for that kind of respect. But the plan wasn’t the anchor. The anchor was ten years of muscle memory. The problem is, the comfort zone stopped being comfortable. Every update added more weight, more glitches. Optimizations that somehow made everything slower. The zone wasn’t cozy anymore; it was a swamp. ## Zen: the browser that doesn’t want to be everything I switched to [Zen](https://zen-browser.app), a Firefox-based open-source browser. It’s fast. It’s light. It doesn’t have an AI sidebar pitching me summaries I don’t need. The first week was rough. Muscle memory kept pulling me back to WaveBox. I’d open Zen, do one thing, then catch myself back in WaveBox for the next five. By week three I stopped opening it entirely. A decade of habit, gone in fourteen days. I haven’t uninstalled WaveBox yet. It’s sitting there, grandfathered subscription still active, like a gym membership you keep paying for because canceling means admitting you’re not going back. I’ll cancel it. Soon. Probably. ## JetBrains: paid through 2028, walked away mid-2026 I used JetBrains IDEs across every language I touched. WebStorm for frontend. PHPStorm when PHP paid the bills. RustRover when I picked up Rust. I was deep enough in the ecosystem that I paid three years in advance, through September 2028. That’s how confident I was. And it’s not that my machine is weak. It isn’t. It’s that [Zed](https://zed.dev) exists now, and the gap got impossible to ignore. Zed is an IDE written from scratch in Rust. Not a lightweight editor wearing an IDE hat. It does everything I actually need: LSP, terminal, Git, and the Copilot/Zed AI integration that’s fast and stays out of the way. It launches instantly. It doesn’t index my entire project for thirty seconds every time I open a folder. It doesn’t re-index every time I switch branches. WebStorm did that constantly, and even with 64 GB of RAM, JetBrains always found a way to fill it up. My MBP’s fans would spin up to vacuum-cleaner levels and just stay there. Apple Silicon is supposed to run cool. JetBrains made it thermal-throttle like a gaming laptop from 2015. ### The IDE migration Getting used to it felt like learning to ride a bike. Awkward for a day or two. Then your hands know where everything is and you stop thinking about it. About two weeks in, I realized I hadn’t opened WebStorm all day. Haven’t looked back. One thing I did lose: Gerry Themes Pro. It’s a paid JetBrains plugin with gorgeous color schemes. I used a dark one called Deep Dark Space or something close to that. Losing it stung a little. So I baked my own theme for Zed: [freshjuice-dev/zed-themes](https://github.com/freshjuice-dev/zed-themes). My daily driver is CyberBanana. Yes, that’s the actual name. No, I’m not changing it. ## The part nobody talks about People treat switching browsers like surgery. “But my bookmarks! My passwords! My extensions!” I’ve never relied on any of that. I don’t trust browser sync and never have. The switch was clean because the hard part was the habit, not the data. Here’s what I actually care about: memory. JetBrains IDEs are memory hogs. Everyone knows this. WaveBox on Chromium isn’t far behind. Between the two of them, a fat slice of my RAM was spoken for before I even opened a terminal. Now I run Zen and Zed side by side and I have enough headroom to run [qwen3.8 27B](https://huggingface.co/Qwen/Qwen3-8B) locally. Not quantized to death. Not swapped to disk. Actually running. The tools I paid for were eating the resources I needed for the tools I actually wanted to use. ## Decade done WaveBox was a great ride. Thomas built something that genuinely changed how I worked on the web, and I was happy to pay for it for ten years. The Veteran plan was a classy touch. But great rides end. Zed and Zen won’t be the last tools I switch to. The world moves faster now. Something new drops every few months. The difference is I’m no longer paying for software that gets slower with every update while chasing features I didn’t ask for. The next switch won’t take me two weeks. May the 4th be with you, Alex * [PreviousClaude Code signs every commit you make. Here's how to stop it.](/blog/claude-code-coauthored-commits/)