Paul

How @better_auth uses @cubic_dev_ to keep 130 k weekly deployments secure.

over the last couple of months, I’ve been building Granola Briefs - our first proactive agent that helps you be ready for your next meeting. Granola works overnight to find the most relevant context for your meeting across your Granola notes, web, and Gmail to help you be ready for your next meeting. literally zero set up required. try it out and let us know how you find it!

@theCTO Use @cubic_dev_

you raise millions and millions and millions of dollars, yet you can't review a large PR (it's not that large)

One thing that I love about @cubic_dev_ is the: "Prompt for AI agents" option where you can copy a prompt that is generated for you, copy/paste to your LLM of choice, so good

@robinebers I just started using @cubic_dev_ when I first see your suggestion and man, same feeling over here it is so fast compare to bugbot, also I like the information and the dashboard

Not getting @cubic_dev_ reviews because Github is down? Use the cubic CLI to run reviews locally: docs.cubic.dev/ide/cli-review

starting to love @cubic_dev_ more than cursor's bugbot every time I run PRs, cubic is just as good at literally half the time it takes for the review that's my biggest issue with bugbot - how extremely slow it is starting to believe this will never change because it's all-inclusive and using e.g. GPT 5.5 FAST variants would cost 2.5x more while we pay the same recommending cubic to most people now for speed

@DanielLockyer Have you tried @cubic_dev_ ? It’s #1 on code review benchmarks for a reason 👀 codereview.withmartian.com

one of the most frustrating things about AI review bots: outdated knowledge

@withmartian jump in! There’s only 2 independent benchmarks for code reviews. One of the benchmarks is an online benchmark that measures performance live across tens of thousands of open source PRs. The other benchmark is an offline one, where the lab maintains a data set of ADPRs with known bugs and runs all the code review bots on them. @cubic_dev_ is first in both, by a huge margin

@PaulSangleF @cubic_dev_ ive heard great things about cubic 👀 also how do the benchmarks get scored? i've been having a lot of success with bugbot over greptile/codex reviews but it's lower here

we cancelled greptile recently due to price increases, so we've cut down to Bugbot + Codex reviews now also finding Bugbot’s resolved rate to be ~75% which is pretty good - i wonder how it’s calculated what PR review tools are you guys having success with?

@kr0der Try cubic.dev I've been using it, works great

@steveruizok @cubic_dev_ #1 on all the benchmarks! codereview.withmartian.com

What are you guys using for automated code reviews? Cursor's gracious bugbot trial ran out and Copilot code reviews leave something to be desired

Then @cubic_dev_ Code Review goes through the code and tells the dev what needs fixing

@dominikkoch @greptile Just use @cubic_dev_ - it's also #1 (by far) on every code review benchmark codereview.withmartian.com

I think ill have to say goodbye to @greptile their new pricing sucks

@dominikkoch @greptile @cubic_dev_ is best and free for open source

@dominikkoch @greptile just use cubic, only actually good review tool.

@michael_chomsky @PaulSangleF @cubic_dev_ bro they have a cli

@michael_chomsky @cubic_dev_ docs.cubic.dev/ide/cli-review 😉 @charlietlamb knows

@PaulSangleF @cubic_dev_ bro it’s 2026 i need a cli

Ok how tf do I review this PR?

@cubic_dev_ has officially reached "people begging us to pay us" levels of PMF.

@eusouomatt @iBruno @linear @conductor_build @cubic_dev_ @coderabbitai Hey Matt, that's great feedback. I'll DM you about this

@iBruno @linear @conductor_build @cubic_dev_ @coderabbitai Tem alguns limites por assento que eu não gostei. Tinha que ter um plano mais bolado por licença ou híbrido pra agents. No mais, tá interessante

Meu dia-a-dia, no momento, fica no @linear plugado no @conductor_build rodando Codex Pro 20x com Claude Code Max 20x. Comecei a usar o @cubic_dev_ para revisar as PRs no lugar do @coderabbitai. Mesclo os modelos por tarefa onde Codex acaba mais no front e revisor do Claude Code.