# 策展 · X (Twitter) 🔥

> 📖 本站完整內容索引（documentation index）：[llms.txt](/llms.txt)

> 作者：Morgan (@morganlinton) · 平台：X (Twitter) · 日期：2026-06-17

> 原始來源：https://x.com/morganlinton/status/2066946225837109735

## 中文摘要

Cursor 共同創辦人揭曉 Cursor 下一代大型語言模型展現智慧潛力。此模型不僅標誌著該公司在基礎建設上的重大投入，更暗示了其從單純的程式撰寫輔助工具，轉向更廣泛的智慧化應用發展。

<video src="https://pub-75d4fe1e4e80421b9ecb1245a7ae0d1a.r2.dev/curated/1781666522264-t281u7gx.mp4" poster="https://pub-75d4fe1e4e80421b9ecb1245a7ae0d1a.r2.dev/curated/a03a55a8aadd9810.jpg" controls playsinline preload="metadata" style="max-width:100%;height:auto;display:block;margin:1rem 0"></video>
> 一名講者在舞台上介紹該公司的下一代人工智慧模型。

**核心技術規格**
根據發表內容，Cursor 的下一代模型在規模與訓練資源上展現了極高的投入：
- 參數規模：該模型具備超過 1.5T（1.5 兆）個參數。
- 訓練規模：採用超過 10 萬張 GPU 從零開始進行預訓練（Pre-trained from scratch）。

**未來發展願景**
Michael Truell 在演講中強調了該模型超越程式撰寫的定位，並透過以下階段展示其發展藍圖：
- 智慧跨越：明確提出「Intelligent beyond coding」的願景，顯示 Cursor 意圖將其 Agent 能力擴展至更複雜的邏輯推理與任務處理。
- 團隊協作：透過展示團隊成員在辦公室交流的照片，暗示未來產品將更強調人機協作與工作流程的深度整合，而不僅僅是單一的程式碼生成功能。

**產業影響**
此次發表反映了頂尖 AI 開發工具正積極透過大規模算力投入，試圖在程式開發領域建立護城河，並透過提升模型的通用智慧能力，進一步擴大在軟體工程生態系中的影響力。詳細內容可參考 [Morgan Linton 的推文](https://x.com/morganlinton/status/2066946225837109735) 提供的現場影片紀錄。

## 媒體內容

**一名講者在舞台上介紹該公司即將推出的新 AI 模型及其開發願景。**

**影片中的 Prompt 與操作**

操作步驟：

1. （00:00）講者在舞台上介紹新模型的三大核心技術規格。
2. （01:56）講者說明該模型將具備超越編碼的通用智慧。
3. （03:19）講者提及與某航太公司的合作關係。
4. （04:16）講者展示團隊成員在辦公室討論的照片。

**逐字稿**

- `00:02` And it's different in three ways. First, it's big, and it is as big as Opus and GBT.
- `00:10` Second, it's trained from scratch, so it's not starting from open source bases. We love
- `00:15` open source, and we want to find more ways to contribute to open source, but starting
- `00:19` from scratch will let us control all of the behaviors of the model and tune in even more
- `00:23` for the workloads that we and you care about. Crucially, also, we are running on 10 to 20x
- `00:31` more compute than we've ever had access to. Our last model used a little bit more compute
- `00:36` than we had in the past, but this is really the first model where we're scaling up compute
- `00:40` a ton. And this is a very, very, very big deal because in the past, our Composer models
- `00:47` from the Composer 1 to Composer 2.5, they were trained on a very small set of
- `00:51` GPUs compared to Frontier Labs. And that really limits what could be done. The things that
- `00:56` we were doing to make the model better and better and better were fundamentally blocked
- `00:59` by whether we could run on more GPUs for more hours. And so this scale-up of 10 to 20x really
- `01:07` lets us, we think, get to Frontier, and hopefully soon leap from it and give you all exciting new
- `01:13` powerful capabilities. And then last, this is going to be a model that's going to be intelligent
- `01:19` beyond just coding. We think that this is important for what Crusher would like to do. We want to make
- `01:25` this so that anyone can build anything they'd like on a computer. And the bottleneck to that
- `01:29` is starting to become not just the code writing side of things, but everything that you would want
- `01:34` out of an engineer colleague. And that means using tools that engineers can use. That means
- `01:40` long-term timing. That means actually testing the software and clicking through buttons. That means
- `01:44` also having great UX around showing users exactly what was changed by the agent. But we're going to
- `01:52` take a step outside of just coding and make this model much more generally intelligent. And so that
- `01:59` regime of the training has been kicked off and will be released in the next couple of weeks. And all of this
- `02:04` is backed by our partnership with SpaceX, which as you all know is a little bit more than a partnership.
- `02:11` It's been a slow news day. But, yeah, I think that this is both exciting for this particular
- `02:20` project and kind of the next series of all the places that we're looking forward to. But also kind of
- `02:26` carving back to what we were talking about at the start, there's so few players and institutions in AI that can
- `02:33` really co-design both the product side of things and the model side of things. And I think the
- `02:38` fundamental company DNA of the other players are all that of, you know, maybe large tech companies
- `02:43` or maybe things that started out as labs and then kind of backed into the developer and builder side of things.
- `02:49` And this is a company that, you know, for better of course, I think for better, is really fundamentally about developers.
- `02:57` That's what we started with. That's what we care about. This company is just made up, you know,
- `03:03` is made up of fundamentally people that want to build a tool that's useful for them.
- `03:07` And so this next piece of the company, I think, elevates for the first time a group of people with that product ethos,
- `03:14` with that developer ethos down to its core, and gives that group the ability not just to edit the pixels on the screen,
- `03:21` but also to edit the fundamental capabilities of the models, which is an important piece of the product.
- `03:27` And so really excited about both this next release and just the series of releases we're going to see over the next few months.
- `03:32` And so there you have it. Three big announcements. The app, mobile app being released today, the start of Tiki on GitHub,
- `03:39` and a first sneak peek, the first time we've ever talked about it, of this new model.
- `03:45` And our plan for models going forward. And so really excited to spend the day with you all.
- `03:50` As mentioned, this community is an extension of how we've worked on Purser from the very start.
- `03:55` We are building this with you. And so come up to team members, come up to me, talk about what you hate about Purser,
- `04:00` what you'd like about Purser, what you'd like to see build. And, you know, hope to continue to make the product better for you all.
- `04:07` And we have some awesome content. We have a series of light game talks hosted by Rukash at the chalkboard.
- `04:15` We're a company that likes Blackboards. We have them all over our office.
- `04:17` We have a series of great talks here. We're going to have a conversation with Ed Catmull, the co-founder of Pixar.
- `04:24` And I think it will be kind of a timely conversation because Ed, you know, apart from just generally having his crazy esteemed career,
- `04:30` straddle technology and creativity at crazy industry scale.
- `04:35` And he saw an industry that changed so many times over in a way that maybe even comes to rival just the pace of change in AI.
- `04:46` And so really excited for that to close out the day. And it's great to see you all. Thanks.
- `04:50` Thank you.
- `04:51` Thank you.

## 標籤

新產品, 產業趨勢, LLM, Cursor