introducing t-rex with t-rex enabled, greptile doesn't just review your PR, it runs your branch in a sandbox to find bugs. it mocks api calls, clicks around the UI, and writes + runs unit tests in our benchmarks, it caught ~20% more bugs than base greptile. most of the... new bugs caught could not have been caught with more inference, they required code execution. t-rex is available in beta to all greptile usersshow more

learn more:

Super excited about this release! We've worked hard to push this out and we're super excited for you to try it out in public beta today!

Shlok is the goat

Super interesting, we've been using Graphite for over a year now...but this is making me want to try Greptile.

goated name

See this is an actual good product

How dare you! LOL

🦖💀🐛

no way what!!? cant wait to try it

love the name! 🦖

t-rex test elease execute Perflectly named

T-Rex about to make all your bugs extinct

I like the image of a dinosaur stomping around your app to find bugs.

crazyy launch video mate

I built a similar sandbox environment for scaling UI verification with agents. The challenge is how do you spin up the additional services needed for your app? I'm not sure how you guys are doing it. For anyone curious, here's my post

Great name and glad you see the oppty here too. Try if you want to see the next 6 months of Greptile's roadmap

This is really useful. Running code to find bugs is powerful.

The hard part wasn't getting the agent to catch bugs. It was making it tell apart existing bugs from the ones it made while running.

i have been using greptile for a long time now & honestly a massive upgrade technically but not sure how much more money i have to burn.

What types of bugs did T Rex catch that reviews missed?

@hampiholey This is amazing!

lgtm hits different when the reviewer actually ran your code

yooooooo

i would like to know what do you guys do about false positives? a lot of times it catches bugs that do not exist (usual greptile) but ofcourse I'm interested in this!

You guys are crushing it

this is sick, what is the most common way users interact with this system?

hey daksh you could save 6x to 25x on whatever sandbox provider you use if you switched to it's extremely simple to use, to put code in and run it (we come with most programming languages and docker pre installed, even a desktop + computer use to QA, and video recording / fullstack web demo hosting primitives) not only will the t rex reviews be better, they'll cost less I'd be happy to help

watching an ai click through my messy ui feels way too personal

That is what Devin has been doing since long time..

@ycombinator this is how you build shit not just talk about it

Greptile runs branches in sandboxes. Most PRs just need a diff and a coffee break

What was the benchmark used? How did the team evaluate to get the +20% metric?

This amazing. Congrat

this is the real line, agents that act vs agents that suggest. Suggesting is cheap, executing is the whole game

first time hearing about greptile during Chris video where he interviewed you. great accomplishments, and you look way more mature and articulate than 24 as he said. how does greptile behave in elixir codebases? I am saving this post to remember myself to give it a try.