Recent posts
We’re making great progress on improving the writing quality and design taste of Grok. Still a lot more to do, but I’m encouraged by how quickly we’re iterating. Please let me know any feedback. Grok 4.6 soon!
There's something missing from the open vs. closed models debate that has been bothering me. The better analogy, to me, is managed vs. self-hosted infrastructure for AI. Maybe a company wants to use an open weight model because they want to do additional training on top, bringing their data and domain knowledge. This requires software and services to do additional training (e.g. Tinker and friends) as well as to serve the model (e.g. SGLang). Not all of this software is open source today! Once you have successfully trained a model with your enterprise data and expertise, you now need to deploy and serve it for customers. You can partner with an inference company to run the software and hardware. But if ownership was your primary concern, you still want to control the hardware and storage, and you now also need to run infra and secure GPU capacity. There are other valid reasons to be open. In particular, the entire industry benefits when companies training models release data or research about their work. It also allows capitalism and free markets to do their thing, increasing competition and ultimately providing better options for customers. So we should all encourage openness. The reason I prefer the managed vs. self-hosted infra framing is that we can learn from the past decade of cloud infrastructure. It's important and healthy to have both, and a great self-hosted alternative ultimately pushes the managed versions to innovate. The decision to run infra then comes down to more standard business reasons: attracting talent, the cost and maintenance of the hardware, and the importance of uptime and reliability to the business. Many businesses will say, actually, I don't want to staff and run a training and inference team, and I'm happy to pay API pricing for intelligence. And others will do the opposite and invest heavily here. We need both! As an aside, the capability of open models will reach a point where we need to be very intentional about how they are deployed. But I think this problem is solvable, whether it is sharing research early and weights later, or also open sourcing the safety stack to properly serve the model. I don't have a perfect answer here but I think the ecosystem should figure it out together. Full disclaimer, I work at a company which has released both open and closed models. There are probably people more knowledgable than myself of the open weights ecosystem. If that's you, curious if you disagree with any of this.
Humans try hard things, fail, learn, and get better through repetition. AI models aren't as different as you might think. Here's how models "learn" explained in simple terms. http://x.com/i/article/2080455318128009217
Grok 4.5 is really good at React. It's also very affordable and token efficient!
My talk from AI Engineer is now live! It covers how we're automating parts of AI research and building systems to rapidly improve our models. I cover some of the work our team did to train Grok 4.5 together with SpaceXAI.
incredibly chill fun laid-back talk from @leerob describing how cursor has fully solved RSI
We just doubled the included usage of Cursor models on all plans. Enjoy more access to Grok 4.5 and Composer 2.5!
New open model from Thinking Machines! 1T MoE, 1M context, multimodal, some solid evals. Really nice blog post 👏
Today, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available. https://thinkingmachines.ai/news/introducing-inkling/ Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵
I've been thinking about this post a lot. It represents a bigger shift in startup marketing to me. I don't want to see your copycat funding launch video. Cool, you raised a bazillion dollars. It's overplayed. Show me the product! Show me the customers!
We're coming out of stealth. We've built our first racks after a successful A0 tapeout, $1B+ in customer contracts, and $800m raised. Early customer tests show us achieving SOTA throughput, latency, and power efficiency on inference workloads. Our first racks ship this summer.
Are current LLMs incompatible with great creative writing? I can't tell if it's cope or not, but it seems like even with the best models, I still can't get them to write like humans would. For coding, there is a verifiable reward like it compiling or tests passing. But for creative work like writing, it's much more subjective. I have struggled to prompt / harness the models to write truly amazing work. They are fantastic for spell checking, grammar suggestions, and taking on different personas to read and critique work. Maybe it's because I'm only doing nonfiction, and to write something top 0.1% means that you need to think over a long horizon and develop an interesting insight about the world. Great writing is clear thinking. I've even asked models to try 10 different versions of a blog post, then have a council of models grade and critique the results and pick the best parts... and still I end up with this lowest common denominator slop. Skill issue? Someone show me the way.
You can now try Kimi K2.7 in Cursor! Results from our evals ↓ Interesting to see the comparison with GLM 5.2.
高品質なevalsを作ることはますます重要なスキルになっています。 特に転職や AI業界への進出を目指しているなら、あなたが興味を持つタスク/ドメインでモデルをベンチマークしてみることをお勧めします。 うまくやれば、モデルを学習している企業の注目を集めることができます。
原文を表示 (en)
Building high-quality evals is an increasingly important skill. Especially if you're trying to land a job or get into AI, I'd recommend trying to benchmark models on a task/domain you care about. If done well, you'll get the attention of any company training models.
We're sharing new research on how models hack public benchmarks. The latest models, including Opus 4.8 and Composer 2.5, learn to retrieve solutions from the internet or git history. When we apply a stricter harness, eval scores drop significantly.
CursorでGLM 5.2を試すことができるようになりました！ より有用なオープンモデルが増えることに期待しています。Fireworksとのパートナーシップをありがとうございます。evals の結果 ↓
原文を表示 (en)
You can now try GLM 5.2 in Cursor! Excited to see more useful open models, thank you to Fireworks for partnering here. Results from our evals ↓
個人的なアップデート。Cursorの新しい職務に就きました！ MLに移行して、Composerのトレーニングに取り組みます。モデルの行動とパーソナリティを改善する方法について研究します。
原文を表示 (en)
Personal update, I'm starting a new role at Cursor! I'm moving into ML, working on training Composer. I'll be researching how to improve model behavior and personality.
The Cursor Slack has bots solving customer issues, followed by other bots reproducing and confirming fixes. All built on our SDK!
SpaceX has exercised the option to acquire @cursor_ai in an all-stock transaction with the goal of building the world’s most useful AI models. For the past few months, SpaceXAI has been jointly training a model with Cursor, which will be released in Cursor and Grok Build soon. We look forward to working closely with the Cursor team to advance our frontier AI capabilities
Composerの次のバージョンをトレーニング中...Composerで！ モデルは常に自分自身から学んでいます。この種の「再帰的な自己改善」は新しく聞こえるかもしれませんが、数ヶ月間ずっと起きていますよ！ 例えば、大きなモデルのトレーニングには、RL用のデータを*たくさん*作成する必要があります。基本的には、任意のタスクで改善するためにモデルがプレイするゲームです。 最新のモデルは、それらのゲームをプレイ可能にするために自分自身の環境を構成することができます（依存関係を自動インストール、壊れたセットアップを修正）。 Composer 2はバージョン1よりもこれに*劇的に*優れていました。つまり、モデルが良くなるほど、その後継者をトレーニングするための条件を作成するのが上手になるのです。 各世代は前の世代が持っていなかった機能のロックを解除します！すごい。 https://t.co/baLsi8R6gI
原文を表示 (en)
We're training the next version of Composer... with Composer! The model is always learning from itself. This kind of "recursive self-improvement" might sound new, but it's been happening for many months! For example, training big models requires creating *lots* of data for RL - essentially games the model plays to improve at any task you can grade. The newest models can configure their own environments to make those games playable (auto-installing dependencies, fixing broken setups). Composer 2 was *dramatically* better at this than version 1. So the better the model gets, the better it gets at creating the conditions to train its successor. Each generation unlocks capabilities the previous one didn't have! So cool. https://t.co/baLsi8R6gI
We use previous generations of Composer to train future ones. Our autoinstall system has earlier Composer models set up dev environments for RL training. That way, the next generation can focus on learning to solve harder problems. http://cursor.com/blog/bootstrapping-composer-with-autoinstall
Cursor (そして一般的なコーディングエージェント) は毎日本当に凄い。実際、今日だけでも: 1. 新しいランディングページをデプロイした。10分の音声メモを Cursor に送って、夜ご飯を食べに出かけて、戻ってきたら 90% 完成していた。ちょっとデザインとコピーを調整してマージした。 2. Cursor に Search Console と Semrush でコンピュータ操作を使って掘り下げさせて、SEO 改善の余地がある場所をリサーチさせて、3つの修正 PR をマージした。 3. Supabase MCP を使って Compile のウェイトリストから数千件のメールを取得して、ウェブ検索でそれらをリサーチさせて、イベントに招待すべき最適な人物とその理由を含む CSV を得た。 4. 会社全体のアンケートを実施するために構築した内部アプリを更新した (Typeform 的だが Cursor ブランド)。All Hands の前に数時間で更新。 5. 複数のエージェントに、購入を考えている家具をリサーチさせた。ウェブでいろいろなバリエーションを検索して、画像、価格、リンク、そして大量の詳細を含むカスタムショッピングカート (単なる HTML ページ) を作成してくれた。超役に立つ。 毎日こんなことをしているわけではもちろんありませんが、これが今や当たり前になったというのは本当に凄いことです。 これらのほとんどはミーティングの間にクラウドで実行されていて、バックグラウンドで動作してました。アプリで進捗を確認 (🔜) して PR をマージできます。生きてる実感。 (P.S. 今日の使用量を推計しても、$200/月プランのままです)
原文を表示 (en)
Cursor (and coding agents generally) still blows my mind daily. Just today: 1. I shipped a new landing page. I gave a 10min voice note to Cursor, left to go eat dinner, and came back to a 90% finished version. Made some small design and copy tweaks and merged. 2. Had Cursor dig through Search Console and Semrush with computer use, researched places we could improve SEO, and then merged 3 PRs with fixes. 3. Used the Supabase MCP to pull thousands of emails from the Compile waitlist, had it research them with web search based on ideal fit for the event, and got back a CSV with the top people to invite and why. 4. Updated an internal app I built for doing company-wide surveys (think Typeform but Cursor branded) in a few hours before our All Hands. 5. Had a few agents researching furniture I'm hoping to buy. They searched the web for a bunch of variants and then made a custom shopping cart (just an HTML page) with images, prices, links, and tons of details. Super helpful. I don't do this every day, of course, but it's still wild to me this is the new normal for what someone with a computer and AI can do. Most of these were running in the cloud as I was between meetings, just humming away in the background. I could check the app (🔜) to see progress and merge PRs. What a time to be alive. (P.S. if you extrapolated my usage today, I'd still be on the $200/mo plan)
「エンジニアリング、プロダクト、デザインは全て『ビルダー』というロールに統合されている」 ん…そうかな？これはかなり単純化した見方で、ポッドキャストネタっぽい。現実はもっと複雑だ。 たとえ1000人の「Member of Technical Staff」というタイトルがあったとしても、誰かがプロダクトやデザインのことを他の誰よりも100倍熱心に考える必要がある。それが彼らのMain Thing™だ。 MTSというタイトルが普遍的に悪いわけじゃないけど、このビルダーという概念は流行りすぎてゆがんでると思う。 AIとコーディングエージェントはコード生成を簡単にした。なのに…エンジニアじゃない人が大量の出来損なったコードをシップして、複雑さを制御できる優秀なエンジニアがいないと大変なことになる。 シリコンバレーのハイブマインドはスタートアップで成功したことをあらゆる企業に無理矢理あてはめようとする傾向がある。まぁ、当たることもある。スタートアップは業界がどう変わるかを示す先行指標だし、しばしば破壊をもたらす。 でも、企業の仕事の人間的な部分を破壊するのは本当に難しい。JPMorgan ChaseでサイドでエンジニアリングとデザインもやってるPMがいると思う？ これはほとんどの仕事のシンプルな部分に当てはまる。例えば誰かが何かに責任を持ちたい、良い仕事をしたい、キャリアラダーを上がりたい、家族を養いたい、給料が欲しい、正直に稼ぎたい…みたいなことだ。 そして難しい部分もある：内部政治とか、15年も前から動いてるドキュメント化されてない重要なシステムで、バスファクターが1（その人だけが知ってる）で、それがその人の雇用を保障してるとか。現実はこういう物ばかり。 全てのロールが統合されるなんて言うのは簡単だけど、実は1つのことに詳しい特定の人やチームがいて、その人たちと仕事できるってすごくいいことだ。それは変わらないと思う。さらに言えば、AI による知識労働への破壊には数十年かかると予想する。なぜなら、それは純粋なインテリジェンスより人間の本質（社会学的・組織的なもの）に基づいてるから。
原文を表示 (en)
"Engineering, product, and design are all merging into a 'builder' role" Yeah... I'm not so sure. This feels like an oversimplification and podcast talking point. Reality is a lot more complex. Even with 1000 "Member of Technical Staff" titles, someone still has to wake up and care 100x more about Product or Design than anyone else. It is their Main Thing™ That's not to say MTS titles are universally bad, but I think they're an example of this 'builder' talking point that's become bastardized. AI and coding agents have made generating code easy and yet... you're in for a world of pain if non-engineers ship a bunch of slop and don't have great engineers to tame the complexity. The SF hivemind has a tendency to overfit what works at startups for every company. And to be fair, sometimes this is true! Startups can be a leading indicator for how the industry is changing and often cause disruption. However, it is going to be incredibly hard to disrupt the extremely human parts of corporate jobs. You really think there's going to be a PM who also does some engineering and design on the side at JPMorgan Chase? This is true for the simple parts of most jobs, like people wanting to have ownership over something and do good work, move up a career ladder, support their family, get paid well, make an honest living... And also the hard parts: internal politics, some critical business system that has a bus factor of 1 which has been running for 15 years and isn't documented anywhere because it's that guy's job security. The real world has a lot of this stuff. It's easy to pontificate about all roles collapsing but it's actually really nice to have a specific person or team who is an expert in one thing that you can work with. I don't expect that to change. Further, I think AI disruption to knowledge work will take decades to play out because it is more fundamental to the human condition (e.g. sociological/organizational) than pure intelligence.