On benchmarks @cubic_dev_ is #1 best by far! The gap between cubic and #2 is bigger than the gap between #2 and #8. codereview.withmartian.com/

I love seeing @cubic_dev_ customers winning

built on cubic.dev

Hey Peter! - Most teams start in shadow mode with the default, lower approval threshold until they’re comfortable with the risk classification. - Shadow mode doesn’t approve PRs - it leaves a comment explaining whether cubic would have approved the PR and why - Once they’re comfortable, they switch it to live mode. - Some larger teams do create a service account, especially if they want a clear audit trail or tighter control over which repositories it can approve in.

Introducing cross-repo reviews in @cubic_dev_. During beta, customers fixed over 1,200 bugs flagged by cross-repo reviews. cubic checks how changes affect related repos, using the relevant branch or open PR. Findings appear in your original review. Available on every plan.

Usamos @cubic_dev_ y super contentos

I use three Agentic PR reviewers all as reviewers that my agent then addresses. If I could only choose one it would be Cubic. @cubic_dev_ @QodoAI @devindesktop

Imo you are good to go , when you are approved by @cubic_dev_

im a big fan of @greptile and @cubic_dev_

I don’t know what model @cubic_dev_ is using to review PRs… Got 11 issues flagged on a PR that @coderabbitai marked as good to go with zero issues. AI code review is pure chaos right now!

my agent loves @cubic_dev_ reviews

i tried @cubic_dev_ in a oss repo today. it's review progress % and updating comments with progress of fixes is something i wish existed in greptile and coderabbit.

I already have a pretty solid workflow for reviewing my own code before opening a PR. But what surprised me that after a couple of months using @cubic_dev_ is that it still catches important things I and my agents miss

Looks like @cubic_dev_ is hard at work 😉

cubicがいない！

we use @cubic_dev_ for reviews, is really good, for test github actions, you can see the code in our open source repo

AI wrote 90% of the code cubic reviewed last quarter. Today we're releasing our @cubic_dev_ State of AI Coding report: which model actually writes the best code, measured on hundreds of thousands of real production commits instead of a benchmark. Turns out it's GPT-5.5. And almost nobody uses it as their main model, 80% of devs run Opus. Report link 👇

so @cubic_dev_ just sent me $210 for referring friends 👀 might use that to give away a few more Claude subs soon why @opencode, @cursor_ai and @OpenAIDevs refuse to offer gifted subs, I'll never understand such a missed opportunity that I keep ringing the bell on (!!!)

Super happy with the outcomes when using compound engineering, intent engineering, @cubic_dev_ & @ClaudeDevs

Daytona is one of the largest sandbox providers for AI agents, powering teams like LangChain and Writer. Every Daytona PR now follows one rule: no human review until every cubic comment is resolved. Full case study: cubic.dev/customers/daytona