Greptile

Ben recently created an onboarding agent for our new engineers and presented it at Postman. Check out his talk:

we’ve caught auth bypass issues in 9 PRs this week on openclaw/openclaw always need to be on high alert in large projects like this

We have been working with Nvidia on their new 120B model - Nemotron 3 Super. Bug detection is a particularly nebulous task for language models. To evaluate Nemotron 3 Super, we tested it with our harness to detect bugs on a number of real-world PRs. We were particularly impressed by this model and felt that punched above its weight class. Read more about what we found. https://lnkd.in/gdNjY5yT

GitHub has two ID systems running side by side. We stumbled onto this while building a feature and couldn't help but go down the rabbit hole.

I'll be at Georgia Tech on Wednesday the 21st ~6-8pm. If you're a CS/Math undergrad, grad student or PhD candidate interested in learning about the types of problems we're working on at Greptile, I'd love to meet you. Comment here and I will DM details.

We've been using Cursor and Claude Sonnet to write SQL, create data pipelines, and upgrade the open source libraries we use. But the bigger surprise this year has been using the following AI tools to REVIEW code: Greptile and GitHub Copilot. They aren't advanced enough yet to consider the big picture - what you're solving, why, or whether there's a better architecture to the solution. But they've been excellent at catching tiny logical mistakes and avoiding the small bugs that humans would never notice. It's saved us countless hours. Worth every dollar. Curious to know what tools you're using for code review, if any?