Supermaven: Fast AI Code Completion and What Happened

Supermaven is a speed-focused AI code completion tool that offered near-zero latency inline suggestions and a 300,000-token context window. Founded by Jacob Jackson, the creator of Tabnine, Supermaven launched in February 2024 with a single premise: autocomplete should feel instant. In November 2024, Anysphere acquired Supermaven and began folding its technology into Cursor. By late 2025, the standalone product was sunset. If you built with Supermaven or considered adopting it, here is what happened and what it means for your workflow.

What Supermaven did differently

Most AI code completion tools introduce a slight delay between your keystroke and the suggestion. Supermaven eliminated that gap. Its custom model architecture, called Babble, delivered inline completions with claimed zero-millisecond latency. The result felt less like waiting for an AI and more like the editor reading your mind.

Beyond speed, Supermaven distinguished itself with context. At launch, its 300,000-token context window was larger than what Copilot or Codeium offered. A larger window means the model sees more of your codebase when generating suggestions, which reduces irrelevant completions and improves accuracy in projects with many interconnected files.

Core capabilities before the acquisition:

• Inline code completion with near-zero perceived latency
• 300,000-token context window (later expanded toward one million tokens)
• Editor support for VS Code, JetBrains IDEs, and Neovim
• Free tier with a Pro tier at roughly ten dollars per month
• Custom model architecture (Babble) built for speed over generality

Supermaven did not offer chat, agentic workflows, or multi-file editing. It completed code. That narrow focus was its strength.

Why Supermaven appealed to vibe-coding founders

Founders building with AI generation tools — Lovable, Bolt.new, Replit — often move to a code editor when the no-code tool hits its limits. At that point, fast completions matter. You are working in unfamiliar code, making small fixes, and trying to keep momentum. Supermaven made that transition smoother because its suggestions arrived before you finished thinking about them.

For vibe-coded projects specifically, the large context window helped. AI-generated codebases tend to be verbose, with repeated patterns and long files. A tool that sees more of that context produces better completions than one working with a narrow view.

What happened: Supermaven joins Cursor

In November 2024, Anysphere — the company behind Cursor — acquired Supermaven. Jacob Jackson and his team joined Cursor to integrate Babble into Cursor’s Tab completion model. The goal: combine Supermaven’s speed with Cursor’s multi-file editing and chat capabilities.

Anysphere announced that the Supermaven plugin would remain maintained, but the team’s focus shifted entirely to Cursor. By late 2025, the standalone Supermaven service was sunset. The technology lives on inside Cursor’s Tab feature, which now benefits from the speed and context-awareness that made Supermaven distinctive.

For founders, this means Supermaven is no longer an independent option. If you relied on it, you need to make a choice.

Supermaven vs Copilot, Cursor, and Cline

Understanding where Supermaven sat on the spectrum of AI coding tools helps clarify your options now:

• Supermaven focused exclusively on inline completion speed. No chat, no multi-file edits, no agentic behavior. Pure autocomplete, optimized for latency.
• GitHub Copilot offers inline completions plus chat, code explanation, and workspace features. Broader scope, slightly higher latency, deeper GitHub integration.
• Cursor provides inline completions (now powered partly by Supermaven’s technology), multi-file editing, chat, and composer mode for larger changes. It is the spiritual successor to Supermaven’s speed promise.
• Cline is an autonomous VS Code agent that creates files, runs commands, and executes multi-step tasks. Different category entirely — agentic, not assistive.

If speed was your reason for choosing Supermaven, Cursor is the natural next step. If you preferred Supermaven because it was lightweight and stayed out of your way, Copilot in its default suggestion mode may feel closer to that experience.

Signs your Supermaven workflow needs attention

These symptoms suggest the Supermaven sunset is affecting your development velocity:

• Completion quality dropped. The plugin still works but suggestions feel less accurate or slower than before. Maintenance has shifted to Cursor; the standalone model is no longer improving.
• Context window feels smaller. You notice more irrelevant suggestions in large files or monorepo projects. Your replacement tool may have a narrower context window than Supermaven offered.
• Editor switching friction. You migrated to Cursor for the Supermaven technology but your project setup, extensions, or debugging workflow relied on VS Code or JetBrains.
• Vibe-coded foundations are exposed. Losing a strong autocomplete tool reveals how much you depended on it to navigate AI-generated code you do not fully understand.
• Regression frequency increased. Without fast, context-aware completions guiding small edits, you make more mistakes in unfamiliar parts of the codebase.
• Onboarding stalled. New team members who expected Supermaven as part of the dev setup now need a different tool, and the transition has not been planned.

Any two of these together indicate the transition deserves deliberate attention rather than hope that things settle on their own.

Checklist: before you migrate away from Supermaven

Use this before choosing a replacement tool or changing your development workflow:

• Inventory your editor setup. List the editors your team uses. If everyone is on VS Code, Cursor is a low-friction migration. If you use JetBrains, Copilot or Codeium may fit better.
• Measure what you actually used. Supermaven was completion-only. If that is all you need, avoid paying for a full AI editor. If you want chat and multi-file editing, Cursor or Copilot give you more.
• Test with your codebase. AI completion quality varies by language, framework, and project structure. Run your replacement tool for a week on real work before committing.
• Check context window requirements. If your project has large files or monorepo structures, verify that the replacement tool handles long context without degrading.
• Audit your AI dependency. If losing one autocomplete tool disrupted your team, your workflow may depend too heavily on AI tooling you do not control. Diversify or reduce reliance.
• Review the codebase itself. A tool transition is a good moment to ask whether the code needs a steady hand — structural cleanup, naming consistency, test coverage — before layering on new tooling.

What Supermaven’s sunset means for AI-built products

Supermaven’s story illustrates a pattern founders should expect. AI developer tools are consolidating. Standalone features get acquired and absorbed into larger platforms. The tool you depend on today may not exist independently next year.

This matters most for teams that built quickly with AI tools and now maintain a product they did not fully write. When one piece of the AI toolchain changes — a completion tool sunsets, a model gets deprecated, a pricing tier disappears — the impact is larger if the codebase itself is fragile.

The practical defense is straightforward: keep your codebase clean enough that any competent tool or engineer can work in it. Clear naming, consistent patterns, reasonable test coverage. These fundamentals outlast every tool transition.

When Supermaven’s exit signals deeper problems

If the Supermaven sunset exposed deeper problems — a codebase that is hard to navigate, features that regress with small changes, an architecture that only made sense while the AI was guiding you — the right response is not another tool. It is engineering support.

Spin by Fryga works with founders who built fast using AI tools and now need stability. We step into vibe-coded and AI-generated projects, stabilize the core flows, and restore shipping confidence. No rewrites. No lectures about technical debt. Just a steady hand that gets the product moving again.

Supermaven was a sharp tool for a specific job. Its technology continues inside Cursor. For the code it helped you write, make sure the foundation holds — with or without the tool that built it.