A Q1 2026 developer survey found that experienced developers run an average of 2.3 AI coding tools simultaneously. The most common stack: Cursor for daily editing, Claude Code for complex agentic work, and Copilot for projects locked into the GitHub ecosystem.

That number — 2.3 — is worth sitting with. It’s not 1.0. It’s not 5.0. It’s 2.3.

Why 1.0 hasn’t happened

The obvious prediction for a maturing market is consolidation. One tool wins across enough dimensions that developers stop maintaining separate subscriptions. That prediction hasn’t come true, and the 2.3 average tells you why.

Each of the three dominant tools has a genuine structural advantage that the others don’t have:

Cursor has the best IDE integration. Tab completion that knows your codebase. Visual diffs. Composer sessions that stay in context with your open files. Keyboard shortcuts that feel native. Developers who use Cursor as their primary editor aren’t choosing it because it has the best underlying model — they’re choosing it because the editing experience is genuinely better. You feel this in the latency of completions, the accuracy of context retrieval, the quality of the diff presentation.

Claude Code has the best agentic capability for complex multi-step tasks. Large refactors. Codebase exploration with a fresh pair of eyes. Tasks that require combining reading, writing, and running code over many iterations. Claude’s long context window and instruction-following mean it handles the kinds of tasks where Cursor’s editor-focused approach adds friction rather than value.

Copilot has GitHub integration that neither of the others can replicate. If your workflow runs through GitHub — pull requests, code review, Actions pipelines, Codespaces — Copilot is woven into the fabric in ways that require effort to replace. The code review feature, in particular, does something the other tools don’t: it reviews your PR directly in the GitHub UI, with access to the full diff and discussion context.

These aren’t interchangeable advantages. A developer who drops Cursor loses the IDE experience. A developer who drops Claude Code loses the agentic depth. A developer who drops Copilot loses the GitHub thread. So they keep all three.

The most common stack

The three-tool stack I see most often among active practitioners:

• Cursor for daily editing — it’s the default editor for most AI-forward developers I talk to
• Claude Code for anything complex — kicked off from the terminal when Cursor isn’t the right shape for the task
• Copilot for GitHub — kept active because the PR review and Actions integration are too useful to give up, even if the developer’s primary coding tool is Cursor

This stack costs around $70-80/month assuming base plans across all three. That’s not nothing, but for a developer whose daily work involves significant coding, the productivity argument is not hard to make.

The real cost isn’t money

The money is manageable. What’s harder is context.

Running three tools means maintaining three mental models. Which tool should I use for this task? Which one has my current codebase indexed? Did I set up the context in Cursor, or was that my Claude Code session from this morning? The cognitive overhead of deciding which tool to reach for is small per decision but adds up across a full day.

There’s also workflow fragmentation. A coding session might start in Cursor, escalate to Claude Code when the task gets complex, and end with a Copilot code review on the resulting PR. Each handoff is a context switch. Each tool has its own session state, its own notion of what you’re working on.

The 2.3 average probably understates this friction for the developers who are most invested in AI-assisted workflows. The developers running 3+ tools are also the ones doing the most context-switching.

Will this consolidate?

The 2.3 number will probably not converge quickly to 1.0. The structural advantages of each tool are real and maintained by teams that are well-funded and actively developing. Cursor won’t be better than Claude Code at long agentic tasks until Anthropic lets them use the same underlying capability — which would require a licensing or integration arrangement neither party has announced. Claude Code won’t displace Cursor’s editor experience until Anthropic builds and ships an IDE, which is not on any known roadmap. Neither will displace Copilot’s GitHub integration until Microsoft decides to open that integration to third parties.

What I expect instead: the number hovers around 2-3, with slow consolidation driven by whoever builds the best integration story rather than the best model. The tool that wins on the “works everywhere” dimension — editor, terminal, GitHub, CI — doesn’t have to be the best at any one of them. It just has to be good enough at all of them that the switching costs of maintaining a multi-tool stack stop being worth it.

That tool doesn’t exist yet. Until it does, 2.3 stays where it is.