When “Good Enough” Stops Being Good Enough
Cursor arrived at exactly the right moment – an AI coding assistant that felt genuinely useful rather than gimmicky, built on top of VS Code so developers didn’t have to abandon their existing habits. It captured attention fast, became the default recommendation in developer Slack channels, and convinced a meaningful slice of professional engineers that AI-assisted coding wasn’t just a party trick. For a while, that momentum looked self-sustaining.
But momentum has a shelf life, and Cursor’s is showing cracks. A quiet frustration has been building among the power users who adopted it earliest – the ones who pushed it hardest and wrote the glowing Twitter threads that drove its initial growth. They’re not rage-quitting. They’re just… drifting. And increasingly, the direction they’re drifting toward is Zed.
What Cursor Got Right, and Then Stopped Improving
Cursor’s core value proposition was speed and integration. You could stay inside a familiar editor, get context-aware suggestions, and ask questions about your codebase without switching tools. That combination was genuinely novel when it launched, and it pulled developers away from GitHub Copilot without requiring them to change their entire workflow. The switching cost felt low because the editor itself was VS Code underneath.
The problem with building on VS Code is that you inherit its weight. VS Code is beloved by millions, but it is not a fast editor. It is not a lean editor. For developers who care about startup time, memory footprint, and keystroke latency – the kind of developers who still argue about Vim versus Neovim at 11pm – VS Code has always been a compromise. Cursor didn’t fix that compromise. It added more on top of it.
Feature releases from Cursor have continued, but the pattern that frustrated users describe is one of AI feature additions layered over a foundation that hasn’t materially improved. The editor itself – the raw act of opening a file, navigating a large repository, handling a complex multi-file refactor without lag – hasn’t gotten noticeably better. For developers who spend 8 to 10 hours a day inside their editor, that stagnation is felt in the body, not just benchmarks.
Zed’s Actual Advantage
Zed is built in Rust. That’s not a marketing point – it’s an architectural reality that produces a measurably different experience. The editor opens instantly. Large files don’t cause jank. The interface responds to keystrokes with a precision that VS Code-based editors can approximate but not match. For developers who’ve normalized a certain amount of friction as the price of modern tooling, using Zed for the first time can feel slightly disorienting, because the friction they’d learned to ignore is simply gone.
Zed added AI features more recently than Cursor, but it did so without sacrificing the thing that made it worth using in the first place. The sequencing matters. Cursor started as an AI product and inherited an editor. Zed started as an editor and added AI. That order shapes everything about how each tool feels to use under pressure.
The Power Dev Migration Pattern
The migration happening right now isn’t a mass exodus – it’s a specific demographic moving first. Power developers, the ones who customize their setups obsessively and blog about their workflow, are the cohort switching to Zed. They’re the same people who drove Cursor’s early word-of-mouth. When they leave, they don’t just take their own usage – they take their recommendations with them.
This creates a compounding effect that isn’t visible in aggregate user numbers yet. Developer tools spread through credibility, and credibility flows from the people whose opinions other developers trust. A senior engineer at a mid-size startup switching from Cursor to Zed and writing about why carries more weight in a developer community than a thousand casual users staying put out of inertia. Cursor can retain its broader user base while quietly losing its most influential ones.
What makes Zed’s position interesting is that it doesn’t need to win the whole market to win this specific contest. Cursor’s growth was never really about capturing every developer – it was about capturing the developers other developers listen to. If Zed owns that tier, it controls the direction of future adoption in a way that raw install numbers don’t reflect. This dynamic shows up in other software markets too: the tool that wins the influential minority often ends up winning the broader conversation, even if it takes time for the install base to follow. A similar pattern played out in how Causal cornered Runway’s most analytically sophisticated users before the broader finance tool market noticed the shift.
Cursor isn’t standing still – it’s raising money, shipping updates, and still converting new users who have never tried anything else. But the feedback loop that originally built its reputation is weakening. The developers posting about Cursor now tend to be newer users discovering it for the first time, while the developers who built its early mythology are quietly posting Zed screenshots instead. A product can survive that shift for a while on inertia and marketing. It cannot survive it forever.
Zed still has real gaps – collaborative features are less mature, plugin ecosystems are thinner, and some language server integrations require more manual setup than VS Code offers out of the box. For developers on large teams with established tooling decisions, switching carries genuine cost. But for individual contributors and small teams – exactly the profile of the power users who are already moving – those gaps are narrower than they used to be, and they’re closing. The question isn’t whether Zed is perfect. It’s whether Zed is better enough, for the people who matter most to Cursor’s reputation, that the math on switching has already changed.
