Activity
7K followers
- Krista Letz reposted thisKrista Letz reposted thisCursor can now attach demos and screenshots of its work to PRs it opens. Your team can review artifacts created by cloud agents directly in GitHub. Enable this feature in the Cursor dashboard: https://lnkd.in/gH2ec5fm
- Krista Letz reposted thisKrista Letz reposted thisCursor's code review agent can now learn from activity on PRs to self-improve in real time. 78% of issues found are resolved by the time the PR is merged. Learn more: https://lnkd.in/gWdbYG65
- Krista Letz reposted thisKrista Letz reposted thisMatt Maher tested frontier models in Cursor v. other harnesses. Cursor boosted model performance by 11% on average: Gemini: 52% → 57% GPT-5.4: 82% → 88% Opus: 77% → 93% His benchmark measures how well models implement a 100-feature PRD. Cursor consistently outperformed.
- Krista Letz shared thisA new way to orchestrate many agents in parallel between local and cloud, Cursor 3 is out now. All local and cloud agents appear in the sidebar, including the ones you kick off from mobile, web, desktop, Slack, GitHub, and Linear. Cloud agents produce demos and screenshots of their work for you to verify.Krista Letz shared thisWe’re introducing Cursor 3. It is simpler, more powerful, and built for a world where all code is written by agents, while keeping the depth of a development environment. The new interface is available as a separate window that complements the IDE. Update Cursor to try it. Learn more here: cursor.com/blog/cursor-3
- Krista Letz reposted thisKrista Letz reposted thisWe turned on AI code review using Cursor Bugbot for 18 engineers. Their PR throughput more than doubled. 📈 Not because they worked more. Not because they wrote more code. Because they started shipping it differently. That result is easy to misinterpret. More PRs does not automatically mean more productivity. So instead of guessing, we designed an experiment to understand what actually changed. 🧪 Find out the results in this article by Jellyfish Senior Software Irene N.: https://lnkd.in/g3bZNdDgAt Jellyfish, Turning On AI Code Review Doubled Our PR Throughput - JellyfishAt Jellyfish, Turning On AI Code Review Doubled Our PR Throughput - Jellyfish
- Krista Letz reposted thisKrista Letz reposted thisWe're releasing a technical report describing how Composer 2 was trained. Composer 2 had three main efforts: continued pretraining, reinforcement learning, and benchmark development. The goal of each was to closely emulate the Cursor environment to produce a highly intelligent coding model. 1. We show how continued pretraining results in consistent improvements in downstream coding performance. 2. The reinforcement learning phase is critical for final performance. We discuss the algorithms we apply for this stage. We find that simple approaches often work best, and improve performance broadly. 3. We describe our internal benchmark CursorBench which represents a more realistic sampling of coding problems. We discuss why we think it is important to include the complex problems software engineers see everyday. 4. We go into detail about the infrastructure behind large scale training including the kernels we developed and open-sourced for the project. We also discuss distributed training and environment scaling for RL. Thank you to the companies and open-source communities behind Kimi K2.5, Ray, ThunderKittens, PyTorch, and more. We'd also like to thank Fireworks and Colfax for their collaboration and partnership. Read the full report: https://lnkd.in/gvKHE5j8
- Krista Letz reposted thisKrista Letz reposted thisCursor cloud agents can now run on your infrastructure. Get the same cloud agent harness and experience, but keep your code and tool execution entirely in your own network. https://lnkd.in/gGNaQp-vRun cloud agents in your own infrastructure · CursorRun cloud agents in your own infrastructure · Cursor
- Krista Letz reposted thisKrista Letz reposted thisCursor can now create new components and frontends in Figma using your team's design system.
- Krista Letz reposted thisKrista Letz reposted thisCursor can now search millions of files and find results in milliseconds. This dramatically speeds up how fast agents complete tasks. We're sharing how we built Instant Grep, including the algorithms and tradeoffs behind the design. Learn more: https://lnkd.in/gXXzmCVm
- Krista Letz liked thisKrista Letz liked thisI filmed this at 11pm right after Cafe Cursor Singapore because I didn't want to forget why we do this. None of us are getting paid to show up. We have regular jobs and run our own businesses. We believe in what's being built here and want to bring it to others. 150 people on a Monday night said the same thing just by showing up. Cursor's own Nick Miller and Frances Thai flew in from the US to take every question the room threw at them. The room did not hold back. That is the culture we want. Watching Singapore's builder community grow has been one of the most energising things I've been part of. Thankful for my fellow agents of chaos and friends Sherry Jiang 🥃 Agrim Singh Ivan Leo Kaspar Hidayat. This was just the start... More to come over the next two months leading up to AI Engineer Singapore in May. Ps. Don't say bojio. 📍Cursor Kopitiam, Singapore 🇸🇬
- Krista Letz reacted on thisKrista Letz reacted on this𝗛𝗼𝘄 𝗖𝘂𝗿𝘀𝗼𝗿 𝗥𝗲𝗮𝗱𝘀 𝗬𝗼𝘂𝗿 𝗘𝗻𝘁𝗶𝗿𝗲 𝗖𝗼𝗱𝗲𝗯𝗮𝘀𝗲 𝗶𝗻 𝗦𝗲𝗰𝗼𝗻𝗱𝘀 (𝗪𝗶𝘁𝗵𝗼𝘂𝘁 𝗔𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗥𝗲𝗮𝗱𝗶𝗻𝗴 𝗜𝘁) When you first open a project in 𝗖𝘂𝗿𝘀𝗼𝗿 𝗜𝗗𝗘, it doesn’t “read” your entire codebase every time you ask a question. That would be far too slow. Instead, it does something much smarter. It begins by creating 𝗲𝗺𝗯𝗲𝗱𝗱𝗶𝗻𝗴𝘀 of your codebase, turning portions of code into numerical representations(vectors) that capture their meaning. But the real magic lies in how it keeps this index up to date without reprocessing everything. Cursor uses a 𝗠𝗲𝗿𝗸𝗹𝗲 𝘁𝗿𝗲𝗲 (used in git, cryptography and torrenting) like structure to track changes efficiently. Each file (and even parts of files) is hashed and stored in a tree. When you modify a file, Cursor doesn’t rebuild the entire index it simply compares hashes, detects exactly what changed, and updates only those parts. This means: • No full re-indexing • No wasted computation • Near-instant updates, even in large codebases When a file does change, Cursor breaks it down into 𝘀𝘆𝗻𝘁𝗮𝗰𝘁𝗶𝗰 𝗰𝗵𝘂𝗻𝗸𝘀 which are logical units like functions, classes, or blocks (using 𝐀𝐒𝐓). These chunks are then converted into embeddings, enabling fast and accurate semantic search(via cosine similarity). To make things even faster, Cursor caches embeddings based on content. If a chunk hasn’t changed, its embedding is reused instantly. No recomputation needed. The result? A system that avoids redundant work at every step: • Only changed code is reprocessed • Only relevant chunks are retrieved • Only necessary context is sent to the model This is why Cursor can understand massive codebases in seconds not by brute force, but by being incredibly efficient about what it chooses not to do. Refrences: https://lnkd.in/g_CSJDGR https://lnkd.in/gGQ9NVwD #ai #cursor #tech #coding
- Krista Letz reacted on thisKrista Letz reacted on this📣 Finally, we get to tell this story. How did history's fastest growing tech company tackle deal velocity? It starts with sending every MSA redline back to its prospects in hours, no matter what. We are very lucky to work with Tina Soppitt, Dan Riley and the relentless Cursor team. They pushed us to make Crosby better, and faster, for all of our clients. Link to the full story in comments.
- Krista Letz reacted on thisKrista Letz reacted on thisRecently, I have been working on setting up Cursor AI inside our engineering workflows to make development more structured, faster, and consistent across teams. Many people think AI coding tools simply take a prompt and generate code. In reality, a well-configured setup works very differently. Behind the scenes, there is a context orchestration pipeline that ensures the AI understands the system, architecture, and engineering standards before generating any output. This is what actually happens internally: User Prompt ↓ Cursor Context Builder ↓ Load User Rules (local preferences and development style) ↓ Load Project Rules (.cursor folder inside the repository) ↓ Load Team Rules (organization standards from Cursor dashboard) ↓ Merge rules and resolve precedence ↓ Construct the final LLM prompt with full context This means the model is not responding only to the developer prompt. It is responding to a context-aware prompt that includes architecture rules, coding standards, and domain knowledge. To make this effective, we are building a repository-level knowledge structure like this: .cursor/ rules/ – architecture rules, coding standards, security guidelines commands/ – reusable engineering workflows skills/ – domain and technical knowledge packs This transforms Cursor from a simple coding assistant into a context-aware engineering collaborator. The goal is not just faster code generation, but consistent architecture, better onboarding, and automated development workflows across the team. AI-assisted engineering works best when the system is designed to guide the AI, not when developers rely only on prompts. Excited to see how this evolves as we continue integrating AI deeper into the development lifecycle.
- Krista Letz liked thisKrista Letz liked thisHere's my philosophical framing for building with Cursor. If you are ~vibe coding~ you are doing it wrong. Instead, think of yourself as a director with an entire team of agents working alongside you. A good director shows up with a clear vision, gives their team the right context, and sets guardrails so everyone can do their best work. Same principle applies here. Before I let agents touch code, I write detailed specs. I use rules files to define boundaries. I write hooks to ensure agents don't do anything dangerous without me knowing. I break work into scoped tasks with clear objectives. Then I get out of the way. The reality is agents are remarkably good at reading codebases, writing clean code, and finding creative solutions - and they're only getting better. Micromanaging them defeats the purpose. Your job is the stuff only you can do: defining the problem, scoping the work, making architectural calls, and deciding what ships. Stop vibe coding. Start directing.
- Krista Letz reacted on thisKrista Letz reacted on thisShameless final BSidesSF panel plug 🔌 🎤 This Saturday PM, join me and friends Travis McPeak, Jackie Bow, Drew Hintz, and Kyle Polley to talk about all things AI for Security! We’re at a moment in time where AI is simultaneously being pitched as the greatest security defense ever built — and the greatest offensive capability attackers have ever had! Is #AI for #Security Friend or Foe? AMC Theatre 12 Saturday March 21, 2026 1:45pm - 2:30pm Would love to hear some questions from my network that you would ask the panel given the opportunity? #BSidesSF #SanFrancisco
- Krista Letz liked thisKrista Letz liked thisExcited to share I’ve joined Cursor! Two things stood out to me before joining, and even more so now: 1. The talent density is unreal. Everyone here is exceptional at their craft, deeply ambitious, and just as humble and collaborative. I’m excited to learn from this team. 2. The opportunity ahead is massive. Cursor is transforming how software gets built. It’s already indispensable for many engineers, and we’re still early. We’re hiring—come build with us! I’m also incredibly grateful for my time at Mistral and the amazing people I worked with. I’ll be cheering you on always. Reach out if you want to chat :)
- Krista Letz reacted on thisKrista Letz reacted on thisCat's out of the bag! Thrilled to share that I've joined Cursor to build out the Revenue Marketing function. 🚀 Massive thank you to everyone involved in the recruiting process... of course to the team within Cursor, but equally to the external folks who have provided support, guidance and references along the decision path (you know who you are and I appreciate you so much!) ❤️ I'm so excited and grateful for this opportunity and after being on the inside for just five days, I'm incredibly humbled by the talented group of builders that have come together for this insane ride. Special thank you to Jordan MacDonald, Patty Koenigs, Jordan Topoleski, Roman Ugarte, Brian McCarthy, Tido Carriero, Vibhav Joopelli, Tahirih Skolnik, Karl Oskar S., Jonathan Krangel, and Michael T. for the trust and partnership. Looking forward to full onboarding immersion next week with the newest Global GTM Leadership class 👏
- Krista Letz reacted on thisKrista Letz reacted on thisCursor isn’t just another tool, it’s a shift in how we approach development. Had the chance to go behind the scenes at the Cursor meetup in Singapore last week! Big thanks to Daryl Lee and Tommy Tien for the class that paved the way for this. It was a rare opportunity to see how the Cursor team scales AI internally and moves beyond the typical hype. The most valuable part wasn't the roadmap, it was the raw interaction with practitioners who are actually shipping with AI, not just talking about it. Even better. Bumping into former colleague who is now on the front lines of AI development. The density of the tech ecosystem in SG is hitting a new level.
Experience & Education
-
Cursor
View Krista’s full experience
See their title, tenure and more.
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Volunteer Experience
-
-
Volunteer
Phoenix Guide Dogs
- 1 year 4 months
-
Volunteer
Guide Dogs of the Desert
- 2 years 5 months
Recommendations received
1 person has recommended Krista
Join now to viewOther similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content