Rootly

Join Us at #AWS re:Invent! Stop by our #HappyHour and connect with builders, partners, and leaders over drinks and good conversations. While you're at the conference, make sure to visit our booth for live demos and the latest updates from our team. Dec 1: https://luma.com/k81v2256 Dec 2: https://luma.com/klbqmxf3

As developers become more dependent on code assistants, they risk becoming less familiar with their own codebase. Sylvain Kalache, Head of Rootly AI Labs, raises the critical question: what happens when those systems fail in that case? Rob Zuber, CTO at CircleCI, argues the answer is more AI, not less. Agents can lean on your observability stack and system context to run a diagnosis and surface a trustworthy path to a fix. Watch the full discussion: https://lnkd.in/g5_jPh8p

📱 Rootly Mobile 2.7.0 is live with a redesigned on-call experience focused on speed, signal, and execution. We’ve rebuilt the home screen to act as a real-time command center—not just another menu. When you open the app, you immediately see: * The schedules you’re currently on call for, plus what’s coming up next * Open coverage requests tied to your rotations, surfaced before they become a problem * Active work (alerts and incidents) pinned to the top so you can jump in without hunting through tabs * Personal performance insights with week-over-week trends across incidents and alerts, so teams can reason about load, responsiveness, and on-call health over time We’ve also added a global Action button to remove navigation overhead when seconds matter. From anywhere in the app, you can: * Page responders into an incident * Request coverage to fill gaps in your rotation Update to Rootly Mobile 2.7.0 from your app store and read more in our changelog link below in the comments.

I'll be in Vegas for AWS re:Invent with Rootly! Shoot me a dm if you’re stopping by the booth (#1530) We also have 2 happy hours, sign up below and dm me to get off the waitlist! • Dec 1: https://luma.com/k81v2256 • Dec 2: https://luma.com/klbqmxf3 See you soon ♠️

What’s ahead for MLOps? For Maria Vechtomova, MLOps Tech Lead and co-founder at Cauchy, it's standardization. If we want AI apps to scale and get more ambitious, we need observability that actually means something across every stage of the pipeline. Clear signals, consistent metrics, and a shared language for quality.

I'll be at NeurIPS next week to present recent benchmarking work from the AI Labs team at Rootly (YC S21)! At Rootly, we have been evaluating frontier models on 'multi-hop' reasoning capabilities, which is the ability to perform reasoning with hidden intermediary steps. We recently introduced a new reasoning benchmark based on real-world code sourced from open source projects. We'll be presenting a poster on December 6th at the Efficient Reasoning workshop, where we'll share recent results and how this can benefit SRE and software developers. It's been a thrill to be part of the team at Rootly that makes this type of work possible. This is the third workshop that we've presented at in under 7 months (following ICML 2025 and ACL 2025), and it means a lot to have been able to be a part of the team. A huge thank you to JJ Tang, Sylvain Kalache, Adam Frank, Quentin Rousseau, Jorge Lainfiesta and the team for your support, it means a lot! We are also progressively releasing more benchmarking frameworks from Rootly, with links joined in the comment section. Our poster will be on Saturday, Dec. 6th at the Efficient Reasoning workshop, please feel free to swing by! (Graphics credits to Giang Tran!)

Generic coding benchmarks are table stakes. We validate frontier models on real SRE work before we believe the hype. And this time Gemini 3 Pro was a standout exception. After running it through SRE-skills-bench (our open benchmark for SRE tasks), Gemini 3 Pro came out... → ~4% more accurate on average than the next best frontier model → +13% on AWS S3 security configuration → +8% on IAM configuration tasks That’s a meaningful jump at a moment when model performance is starting to converge. Tradeoff worth noting: cost. Gemini 3 Pro is roughly ~20% higher per million tokens than GPT-5, which may or may not matter depending on your volume. We’ve published the full methodology and results including our work with Groq OpenBench and the SRE-specific task suite. Everything is reproducible. AI for reliability isn’t about vibes. It’s about whether a model can reason with operational context under pressure. Full benchmark + results in the comments.

In this video, Sylvain Kalache (Rootly) and I chat about the reliability of GenAI systems: ➡️ Why the reliability fundamentals haven’t changed ➡️ Why identical models behave differently across providers ➡️ The realities of fast-moving open-source stacks ➡️ The rising role of ML Ops ➡️ How enterprises navigate the open-source vs. closed-source debate. As usual, zero BS and a few hot takes. You should enjoy this one 😉 Thanks again to Sylvain for the opportunity. Link: https://lnkd.in/eiATr5_B

Planned maintenance in Rootly gives engineering teams a structured, auditable, and automated way to coordinate scheduled work. Whether you're rolling out upgrades, running migrations, or making infra changes, Rootly keeps communication clear, execution predictable, and visibility consistent across every stakeholder. Read more in our changelog (linked below).