We just rolled out CursorBench 4.0! It includes new tasks for how well models follow instructions, work on challenging projects over time, and is more difficult than before (so all models score lower).

The similarities between teaching my daughters to be good humans and teaching machines to think and learn new skills continue to amaze me. And yet, watching my oldest run through our yard, laughing and experiencing life, I can't see how machine intelligence could ever compare. It's already hard to imagine life without AI, much like life without the internet. But the more time I spend talking to machines, the more I appreciate the human condition and the messiness of the real world. Being human, with all its imperfections, is awesome.

Chat → Code → @Bot​s I was skeptical at first, but I'm now Bot-pilled. It is hard to give up "the old way" of prompting agents, but I strongly believe always-running bots are the future of where all work on a computer is headed. For example, it still feels *weird* to not see the response stream back to me, or not look at model pickers, and trust in the model/harness to do compaction in these very long conversations. But once you see how simple the product UX can be, and see the results, this way of working is amazing (even with it being in early beta). Models will continue to improve exponentially in the next 3-6 months, getting better and better at proactively doing real work for you in the background. And this feels like the right interface and UX for that work. I started by using Bot for personal tasks instead of work things, but it quickly took over the majority of my tokens. Very excited about the future of Bot!

We just increased the included usage of Grok models in Cursor! Demand has been increasing with the launch of Grok 4.6. We already doubled limits last month with more compute. And now we're raising limits again permanently. Enjoy!

How can you trust @Bot​s to operate autonomously? This is something we've carefully designed. First, we use an LLM to automatically review every action your bot takes. If anything is unexpected, it will ask for your approval. But this isn't enough. We also have customizable rules to allow or deny actions. For example, I have bots that are connected to my X, Gmail, etc. I explicitly block any write/destructive actions and have it ask for my approval first. This is a good balance between YOLO mode and reviewing everything by hand. Although, we do support YOLO if you like to live dangerously!

We're making Grok Bot more widely available. All SuperGrok Plus, Cursor Pro+, and Cursor Teams subscribers now have access. We're also offering a free trial with limited usage for all other users.

Grok @Bot has made a few simple yet powerful technical decisions that I believe make it easy and enjoyable to use. 1. The best UI is none at all. The product interface is dramatically simpler than alternatives without sacrificing functionality. How is this possible? It's one of the first products designed for current frontier model capabilities and has a UI restrained enough to remain easy to use as models improve exponentially. Everyone knows how to text. 2. A thin harness for the client, a thick harness for the server. You might have noticed the app feels very fluid to use, even for a beta product. This is primarily because of everything we didn't have to build. The app harness is essentially a single tool to send messages between the client and server. The complexity moves to the server, where you can still use the coding agent harness with specialized tools as needed. This helps make the UI fast and responsive on desktop and mobile. 3. An always-on computer. Most coding agents and assistants today start fresh with every question you ask. Some of these sessions are on your local machine and others happen in the cloud. We believe strongly that cloud is the future, which is why it's the only option. Further, rather than spinning up virtual machines for every conversation, your bots connect to their own computer. This means you can still run agents on the bot's persistent filesystem. It's closer to what programmers have been doing by using Tailscale from their phones to connect to a remote computer and run an agent TUI. You get those capabilities without the hassle. 4. Your bots can use the browser. Coding agents have shown that most work on a computer can be expressed and run as code. You can ask for a task in natural language and the agent will decide to write a script to complete it. This is amazing, but there's still many tasks which can't be completed without logging into a website and clicking around the browser. Models and harnesses are now good enough to reliably handle this. The combination of writing code and using browsers means you can automate almost any task on a computer. Further, you can ask Grok Bot to record you doing the task, and then turn it into something repeatable.

Big day! Cursor has officially joined SpaceX. I’m proud of our team and what we’ve accomplished so far. Our momentum is accelerating with Grok 4.6 and Grok Bot as well. I’ll be working on making Grok useful, tasteful, and safe. Onward!

We've released the model card for Grok 4.6! It goes in-depth on the capabilities of the model across many different evals for coding, engineering, knowledge work, and more. We also discuss pre-deployment safety testing and our safeguard stack.