Jake Beyer

I recently joined @cursor_ai . It felt like stepping onto a rocket ship… I didn’t expect to feel this energized again so quickly. After taking some much-needed time off earlier this year, I spent time being intentional about what I wanted next. Not just the role, but the people, the pace, and the problem space. The combination of humility, authenticity, and raw talent here is something special. The talent density is real. I’ve joined the Go-To-Market team focused on AI deployment. After 12+ years in product management, this is a deliberate shift into a GTM role, and it feels like the right move at exactly the right time. It’s wild to think back to 2014 and my time at Y Combinator as a solo, non-technical founder. Then, my options for building were incredibly limited. My first MVP was literally a Wufoo form, and I remember getting estimates north of $100K from dev shops to build a very simple web app. It’s mind-blowing how much has changed. With tools like Cursor and today’s LLMs, what once took months and a huge budget can now take minutes, hours, or days. Going from that experience as a non-technical founder to being at Cursor now, helping teams deploy AI and empowering both developers and non-technical builders to create complex products, feels incredibly full circle. We’re entering a moment where both software and the way we interact with it are being fundamentally reshaped. I believe Cursor and its partners will be a leading force in redefining how humans interact with technology for years to come. The scale of what’s happening here is hard to ignore. 64% of the Fortune 500 use Cursor, more than 50,000 enterprises are built with it, and over 100 million lines of enterprise code are written in Cursor every day. Feeling lucky, grateful, and very ready to build. If you’re thinking about deploying AI in your org, let’s talk. We’re also hiring. cursor.com/careers

@aaron_epstein @ycombinator Very cool @DraftedAI. Very relevant for maximizing FAR on small lots. Would love to try this out.

Designing your dream home just got even faster

Do you ever just need to pull your agent aside to have a quick chat without breaking the flow? Well, now you can…

Introducing side chats, a new way to ask questions and explore ideas without interrupting your main conversation. Each side chat is a durable agent conversation you can @-mention to bring context back into the main thread.

Booking out a local cafe on July 20th in San Francisco Grab coffee, Cursor credits, meet the team, and build Limited spots for founders, engineers, designers, PMs

Announcing Grok 4.5, our first model trained specifically for coding and agents. It was trained with Cursor and offers frontier intelligence at leading speeds and cost efficiency. x.ai/news/grok-4-5

I was clearly wrong about Anthropic. They are obviously currently the leader in AI. No company has released a model as good as Mythos/Fable and they will undoubtedly have Mythos 2 ready soon. And I would never cut them off in a way that hurt them badly, even as a competitor. That’s not my style. Tesla open sourced its patents and we made the Supercharger network available to all competitors, even though we could have made it a walled garden. SpaceX launches competing satellite systems with no increase in price or use of unfair terms. Even my worst enemies can attack me on this platform. …

Grok has always been very strong on law

SpaceXAI just released Grok 4.5, and it ranks #4 on GDPval-AA v2 with an Elo of 1543 - behind only the latest Claude releases from Anthropic on real-world agentic knowledge work tasks Grok 4.5 achieved this score at a cost of $0.49 per GDPval task to sit clearly on the Pareto frontier for performance versus cost. This cost is lower than GLM-5.2 and Kimi K2.6, and nearly 90% cheaper than the models ahead of it on our leaderboard. We’re finalizing the remaining Artificial Analysis Intelligence Index evaluations and will share final results soon. Thanks to @SpaceXAI and @elonmusk for their collaboration testing this model ahead of release, and congratulations on the launch!

Try it out in Cursor with double usage for the first week. cursor.com/blog/grok-4-5

We just dropped our latest model at @cursor_ai 🔥 We partnered with SpaceXAI to train Grok 4.5, and it’s already looking like the new king of coding agents. Terminal-Bench leader, crushing multilingual SWE, and built for way more than just software engineering. The future of AI coding just got a massive upgrade and this is just the beginning.. 🚀

We've partnered with SpaceXAI to train Grok 4.5. It’s our most powerful model yet and the first we've built for more than software engineering.

@gokulr I’ll float this to the team! I know we are tight on bandwidth right now so we may have to keep this on our radar.

@wjbeyer Thanks @wjbeyer ! Would love to see if Cursor might be interested in co-sponsoring this open source effort. The more heavyweight AI companies put their weight behind a standard, the easier it will become to get it adopted.

PRODUCTSPEC: OPEN STANDARD FOR SOFTWARE INTENT tl;dr ProductSpec is the open standard for software intent before implementation. The more I worked on PRDs, the more obvious one thing became: Product specs need an open standard. Why? Because the PRD has become an overloaded artifact. Every company has its own template. Every team has its own preferred format. Every PM has their own way of writing. That was manageable when the only readers were humans sitting in the same org context. AI changes the requirement. A Product Spec now has to be readable by humans and executable by AI agents. That means the spec has to carry intent clearly enough for a designer, engineer, product leader, and coding agent to understand the same thing: • What problem are we solving? • What is the product bet? • What is in scope? • What must be true before this ships? • What metrics tell us whether the bet worked? This is why I open-sourced ProductSpec. ProductSpec is a Markdown standard for software intent before implementation. The core sections are simple: • Problem • Hypothesis • Scope • User Experience • Acceptance Criteria • Success Metrics The deeper design principle: Structure the parts machines must execute or compare. Leave readable the parts humans must reason about. That is why ProductSpec keeps Problem and Hypothesis as readable prose, while giving structured formats to the parts agents and tools need to parse: • Scope: what is in, out, and deliberately cut • Acceptance Criteria: what must pass before launch • AI Evals (within Acceptance Criteria): the evals an AI feature must pass before shipping • Success Metrics: what should be measured after launch When to use ProductSpec ProductSpec is not for every act of building. It is for consequential software work where intent needs to survive handoff. For an individual builder, a Product Spec is useful when the work is complex, risky, long-lived, or being handed to an AI agent loop. For quick experiments, one-off scripts, or throwaway prototypes, it may be faster to brainstorm, build, and iterate directly. For a team or organization, ProductSpec is most useful when coordination cost appears: multiple people, multiple agents, design and engineering handoffs, customer-facing launches, AI features with evals, or decisions that will need to be revisited later. ProductSpec does not replace Git, Jira, Linear, Figma, analytics tools, OpenSpec, Spec Kit, or AI coding agents. It sits upstream of them. ProductSpec -> Engineering Spec -> Tasks -> Code -> Evaluation -> Learning -- Git stores implementation history. A Product Spec can live beside code in Git, but code commits should not be the first durable record of why the work exists. -- Jira and Linear store work history. A Product Spec can become epics, tickets, or tasks, but it should remain the durable statement of intent behind those tasks. -- Figma stores design artifacts. A Product Spec can link to prototypes, mockups, or screenshots through user_experience, but it does not replace the design source of truth. -- Analytics tools store outcome data. -- OpenSpec and Spec Kit turn intent into engineering plans. -- AI coding agents execute implementation tasks. -- ProductSpec stores the software intent behind the work: the problem, hypothesis, scope, acceptance criteria, and success metrics that downstream tools should preserve. I'd love for this standard to be broadly adopted, which means it must be broadly owned by the builder community. Founders, PMs, engineers, designers, researchers, AI builders: please contribute examples, critiques, section changes, parser implementations, validator improvements, and integrations with GitHub, Jira, Linear, Figma, OpenSpec, Spec Kit, and agent workflows. (link below on how to contribute) If you have scars from writing product docs that looked aligned but failed during execution, those scars belong in the standard. My goal is for ProductSpec to become the open source format for software intent before implementation. (links below)

@elonmusk Finally!

Model Y long wheelbase now available to order in the US

Completely agree!

It is unbelievable to me that this wasn’t a red card + penalty kick Balogun was completely obstructed here and still hit the crossbar He earned two PKs that weren’t called This didn’t get VAR but an irrelevant midfield tangle led to a red card Unreal

You can now try Kimi K2.7 in Cursor! Results from our evals ↓ Interesting to see the comparison with GLM 5.2.

Claude Fable 5 is available again in Cursor. It leads all models on CursorBench, but is the most expensive per task.

Cursor for iOS!

Orchestrate with Cursor on the move!