Launch HN: Morph (YC S23) – Morph：以每秒 4500 个 token 的速度应用 AI 代码编辑
29 分•作者: bhaktatejas922•6 个月前
大家好，我是 Morph 的 Tejas。我们构建了一个超快的模型，可以将 AI 生成的代码编辑直接应用于您的文件，速度超过 4500 个 token/秒。不再需要缓慢的全文件重写或脆弱的查找替换技巧。
为什么？AI 生成的代码无法可靠地插入到现有代码中。全文件重写、脆弱的查找替换技巧太慢、成本高或容易出错。
Morph 的方法：
* 您的代理“懒惰地”输出编辑内容，引用现有文件中未修改的行（例如：// ...现有代码...）
* Morph 使用我们的 Fast Apply 模型 + 针对原始文件的推测解码，立即将这些编辑应用于文件，使 AI 补丁快速、可靠且可用于生产。
这种方法是 Cursor 去年首创的，但他们的模型无法作为 API 使用——所以我们为各地的开发人员构建了 Morph（提供大型免费套餐！）
实时演示（无需注册）：[https://morphllm.com/dashboard](https://morphllm.com/dashboard) 和文档：[https://docs.morphllm.com/quickstart](https://docs.morphllm.com/quickstart)
我们有两个 Fast Apply 模型：morph-v3-fast - 4500+ tok/秒，和 morph-v3-large - 2500+ tok/秒。这些模型为 create.xyz、databutton、continue.dev 等提供 Fast Apply 支持！
我们还提供用于嵌入 + 重新排序的检索模型。
下一步：内联编辑模型 (Cmd-K)：极快的内联编辑 - 保持开发流程状态；以及 Morph Tab API：我们的 Next Edit Prediction 模型以低于 500 毫秒的延迟猜测您的下一个代码编辑 + 操作。它目前处于私人测试阶段，但您可以在此处申请抢先体验：[https://morphllm.com/tab](https://morphllm.com/tab)
热门观点：
1. 对于开发人员体验来说，原始推理速度比增量准确性提升更重要——同意还是不同意？
2. 前沿模型的全文件重写是过时的——Fast Apply 编辑在速度、成本和可靠性方面胜出。
3. 随着狭窄任务的基准测试饱和到 99% 以上，复杂性正在从单个前沿模型转移到专门的推理优化模型。随着前沿模型进入高端市场，它们将放弃简单的任务，并且它们将被用于执行只有前沿模型才能完成的任务
我们很乐意听取您对编码代理的想法和经验！
[https://youtu.be/LdT8epGHJPk](https://youtu.be/LdT8epGHJPk)
– Tejas & Morph 团队
查看原文
Hey HN, I’m Tejas at Morph. We’ve built a blazing-fast model for applying AI-generated code edits directly into your files at 4,500+ tokens/sec. No more slow full-file rewrites or brittle search-and-replace hacks.<p>Why? AI spits out code that can’t reliably be inserted into existing code. Full file rewrites, brittle search-and-replace hacks are too slow, expensive, or error-prone.<p>Morph's approach:<p>- Your agent outputs edits “lazily”, referencing unmodified lines in the existing file (ex: // ...existing code...)<p>- Morph instantly applies these edits to a file using our Fast Apply model + speculative decoding against the original file, making AI patches fast, reliable, and production-ready.<p>This approach was pioneered by Cursor last year, but their models aren’t available as APIs—so we built Morph for developers everywhere (with a large free tier!)<p>Live demo (no signup): <a href="https://morphllm.com/dashboard">https://morphllm.com/dashboard</a> and docs: <a href="https://docs.morphllm.com/quickstart">https://docs.morphllm.com/quickstart</a><p>We have 2 Fast Apply models: morph-v3-fast - 4500+ tok/sec, and morph-v3-large - 2500+ tok/sec. These models power Fast Apply at create.xyz, databutton, continue.dev, and more!<p>We also provide retrieval models for embedding + reranking.
Next Up: Inline Edit Model (Cmd-K): Extremely fast inline edits - keep dev flow state; and Morph Tab API: Our Next Edit Prediction model guesses your next code edit + action with sub-500ms latency. It's currently in private beta, but you can request early access here: <a href="https://morphllm.com/tab">https://morphllm.com/tab</a><p>Hot takes:<p>1) Raw inference speed matters more than incremental accuracy gains for dev UX—agree or disagree?<p>2) Full-file rewrites by frontier models are legacy—Fast Apply edits win on speed, cost, reliability.<p>3) As benchmarks on narrow tasks saturate to 99%+, complexity is shifting from single frontier models to specialized inference-optimized models. As frontier models move upmarket, they'll leave simple tasks behind, and they'll be used to do tasks only frontier models can do<p>We’d love to hear your ideas and experiences with coding agents!<p><a href="https://youtu.be/LdT8epGHJPk" rel="nofollow">https://youtu.be/LdT8epGHJPk</a>
– Tejas & the Morph team