Laminar (YC S24)

Laminar had the most explosive initial growth of any other LLM observability platform. Of course we are pretty far behind in terms of total # of stars. If you like Rust, TS and high performance systems star our repo and contribute! https://lnkd.in/eFY_UvAm

We landed in sf 3 months ago to join Y Combinator S24 batch with Dinmukhamed Mailibay and Temirlan Myrzakhmetov. Today was our demo day. It was an extremely rewarding 3 months. We are still pretty much in the beginning of our journey, but we are already backed by some incredible people and are very excited about the road ahead for Laminar (YC S24)!

Laminar (YC S24) is trending at the top of Hacker News. You can think of Laminar as an open-source DataDog + PostHog for complex LLM apps. We're helping devs answer questions like "how many upsells my AI drive-through agent made". Check out our repo here https://lnkd.in/eFY_UvAm and see our full launch on the front page https://lnkd.in/eGmFuJeX.

After we changed the direction to make Laminar the observability and analytics platform for LLM applications, we've been preparing for our HackerNews launch for the last few days. Now that Laminar is open-source it fully makes sense for us to launch there. We polished our Github-repo and made our platform compatible with Opentelemetry, so that anybody can easily start in just a few lines of code with auto-instrumentation, i.e. get almost instant value. Also, added fast analytics using Clickhouse (it was fun!), so that people can instantly see metrics on how their LLM agents or apps perform Here's our HN launch, if you're interested to check it out: https://lnkd.in/eM69JTi7

Reposting my twitter thread about reshaping of Laminar in here. https://lnkd.in/evSG5AXP In the early days of photoroom the team realized that majority of usage comes from product photos and double downed on it. I had an a-ha moment at that time. In our case, I remember always complaining that users underutilize our platform - they build simple flows and just use our tracing and analytics parts of our platform. Now, I realized that I should have just focused on the usage patterns and then infer what the most valuable things about the product. In our case it's: tracing of entire LLM app and ability to quickly experiment with a chain of prompts outside of the codebase Another major flaw of Laminar was that to fully utilize all features of the platform, devs had to rebuild their entire apps as a graph in our pipeline builder. Now thinking of it, it's just ridiculous that it was the ONLY entry point to our app. Now we will double down on things which our customers use us for: tracing, analytics and LLM chain management, and remove the barriers of entries. Devs now can instrument their code directly using our API! Continue reading here -> https://lnkd.in/evSG5AXP

We are in SF to build Laminar AI (YC S24) as a part of Y Combinator S24 batch! Feels surreal... YC always felt like an unreachable dream, but now here we are. Me, Dinmukhamed Mailibay and Temirlan Myrzakhmetov haven't truly grasped the whole situation yet because we've been heads down coding since landing last Thursday, but we're getting there step by step. They say you can't build dev tools without proper twitter presence. I am really bad at social media, but learning doing it in-between coding ??. Follow me on twitter https://lnkd.in/ecmtUPAa to see what it feels like building AI startup at the heart of SF!

Laminar AI major new release: - switched to dark theme and improved UI/UX across the platform - support for cyclical workflows - sub-pipeline node - pipeline templates - public sharable pipelines - drastically faster engine performance Check it out -> https://www.lmnr.ai