CodeRabbit
@coderabbitai.bsky.social
Supercharge your dev team with AI code reviews. Get your free trial now: https://coderabbit.ai!
Our very own Director of AI, David Locker, will be taking to the stage at AI Dev 25 x NYC this Friday!
He'll be talking all things CodeRabbit, Context Engineering & Prompt Engineering!
🐰🎙️
He'll be talking all things CodeRabbit, Context Engineering & Prompt Engineering!
🐰🎙️
November 12, 2025 at 4:12 PM
Our very own Director of AI, David Locker, will be taking to the stage at AI Dev 25 x NYC this Friday!
He'll be talking all things CodeRabbit, Context Engineering & Prompt Engineering!
🐰🎙️
He'll be talking all things CodeRabbit, Context Engineering & Prompt Engineering!
🐰🎙️
A CodeRabbit review per day, keeps the Merge Cop away!
November 10, 2025 at 4:00 PM
A CodeRabbit review per day, keeps the Merge Cop away!
The problem with 👍👎 is they optimize LLMs for approval, not truth.
Stop training dev tools to chase treats like golden retrievers. Give high-signal, contextual feedback instead.
Heres how!
Stop training dev tools to chase treats like golden retrievers. Give high-signal, contextual feedback instead.
Heres how!
November 7, 2025 at 4:27 PM
The problem with 👍👎 is they optimize LLMs for approval, not truth.
Stop training dev tools to chase treats like golden retrievers. Give high-signal, contextual feedback instead.
Heres how!
Stop training dev tools to chase treats like golden retrievers. Give high-signal, contextual feedback instead.
Heres how!
Devs have always followed one core rule: faster is better.
But AI isn’t magic. It’s inference.
That's why engineering AI pipelines for accuracy, instead of low latency is crucial wherever correctness matters more than speed.
Introducing... Slow AI.
Move slow and fix things.
But AI isn’t magic. It’s inference.
That's why engineering AI pipelines for accuracy, instead of low latency is crucial wherever correctness matters more than speed.
Introducing... Slow AI.
Move slow and fix things.
November 5, 2025 at 7:10 PM
Devs have always followed one core rule: faster is better.
But AI isn’t magic. It’s inference.
That's why engineering AI pipelines for accuracy, instead of low latency is crucial wherever correctness matters more than speed.
Introducing... Slow AI.
Move slow and fix things.
But AI isn’t magic. It’s inference.
That's why engineering AI pipelines for accuracy, instead of low latency is crucial wherever correctness matters more than speed.
Introducing... Slow AI.
Move slow and fix things.
Theres still time to enter! 🗣️
> Hackathon Starts: October 29th
> Submissions due: 12:00 PM PT on November 17th
> Judging duration: November 17th to November 24th
> Winners announced: November 24th
Build a full-stack app and compete for $140k in cash prizes and credits! 💰
> Hackathon Starts: October 29th
> Submissions due: 12:00 PM PT on November 17th
> Judging duration: November 17th to November 24th
> Winners announced: November 24th
Build a full-stack app and compete for $140k in cash prizes and credits! 💰
November 3, 2025 at 4:15 PM
Theres still time to enter! 🗣️
> Hackathon Starts: October 29th
> Submissions due: 12:00 PM PT on November 17th
> Judging duration: November 17th to November 24th
> Winners announced: November 24th
Build a full-stack app and compete for $140k in cash prizes and credits! 💰
> Hackathon Starts: October 29th
> Submissions due: 12:00 PM PT on November 17th
> Judging duration: November 17th to November 24th
> Winners announced: November 24th
Build a full-stack app and compete for $140k in cash prizes and credits! 💰
Attention Gemini lovers! 🔔
You can use CodeRabbit with Google’s Gemini CLI in a full agentic loop!
Code with Gemini, review the changes with CodeRabbit, and fix the issues with Gemini.
You can use CodeRabbit with Google’s Gemini CLI in a full agentic loop!
Code with Gemini, review the changes with CodeRabbit, and fix the issues with Gemini.
October 31, 2025 at 4:53 PM
Attention Gemini lovers! 🔔
You can use CodeRabbit with Google’s Gemini CLI in a full agentic loop!
Code with Gemini, review the changes with CodeRabbit, and fix the issues with Gemini.
You can use CodeRabbit with Google’s Gemini CLI in a full agentic loop!
Code with Gemini, review the changes with CodeRabbit, and fix the issues with Gemini.
CodeRabbit is proud to be mentioned in the NVIDIA GTC keynote as part of the AI startup ecosystem building on NVIDIA!
Catch the full keynote from Jenson Huang below!
Catch the full keynote from Jenson Huang below!
October 29, 2025 at 4:02 PM
CodeRabbit is proud to be mentioned in the NVIDIA GTC keynote as part of the AI startup ecosystem building on NVIDIA!
Catch the full keynote from Jenson Huang below!
Catch the full keynote from Jenson Huang below!
We’ve been named in the 2025 Disruptors 60 list!
Curated by Greenfield Partners and live on TechCrunch this list names the most groundbreaking startups turning AI into real-world impact!
Our mission is to scale code quality with AI, and this recognition validates that work!
Curated by Greenfield Partners and live on TechCrunch this list names the most groundbreaking startups turning AI into real-world impact!
Our mission is to scale code quality with AI, and this recognition validates that work!
October 27, 2025 at 3:07 PM
We’ve been named in the 2025 Disruptors 60 list!
Curated by Greenfield Partners and live on TechCrunch this list names the most groundbreaking startups turning AI into real-world impact!
Our mission is to scale code quality with AI, and this recognition validates that work!
Curated by Greenfield Partners and live on TechCrunch this list names the most groundbreaking startups turning AI into real-world impact!
Our mission is to scale code quality with AI, and this recognition validates that work!
CodeRabbit is now available on the Vercel Marketplace!
> Works natively with Vercel projects.
> 3‑step install from Marketplace.
> Manage invoices and seat allocation directly from Vercel.
Install link below!
> Works natively with Vercel projects.
> 3‑step install from Marketplace.
> Manage invoices and seat allocation directly from Vercel.
Install link below!
October 23, 2025 at 5:21 PM
CodeRabbit is now available on the Vercel Marketplace!
> Works natively with Vercel projects.
> 3‑step install from Marketplace.
> Manage invoices and seat allocation directly from Vercel.
Install link below!
> Works natively with Vercel projects.
> 3‑step install from Marketplace.
> Manage invoices and seat allocation directly from Vercel.
Install link below!
If you're a funded startup looking to supercharge your engineering team, you can apply to our Startup Program!
Get 50% off CodeRabbit Pro for 3 months.
To be eligible, your startup must be:
> VC- or accelerator-backed
> Funded with $25 million or less
Get 50% off CodeRabbit Pro for 3 months.
To be eligible, your startup must be:
> VC- or accelerator-backed
> Funded with $25 million or less
October 15, 2025 at 3:48 PM
If you're a funded startup looking to supercharge your engineering team, you can apply to our Startup Program!
Get 50% off CodeRabbit Pro for 3 months.
To be eligible, your startup must be:
> VC- or accelerator-backed
> Funded with $25 million or less
Get 50% off CodeRabbit Pro for 3 months.
To be eligible, your startup must be:
> VC- or accelerator-backed
> Funded with $25 million or less
Earlier this week, our CEO, Harjot Gill, took the stage at OpenAI's DevDay 2025 to showcase CodeRabbit.
During the talk, Harjot discussed:
> Second-order effects of Gen AI
> Agentic verification within our analysis chain
> Understanding PR intent based on repository context
Full vid below 👇
During the talk, Harjot discussed:
> Second-order effects of Gen AI
> Agentic verification within our analysis chain
> Understanding PR intent based on repository context
Full vid below 👇
October 10, 2025 at 3:37 PM
Earlier this week, our CEO, Harjot Gill, took the stage at OpenAI's DevDay 2025 to showcase CodeRabbit.
During the talk, Harjot discussed:
> Second-order effects of Gen AI
> Agentic verification within our analysis chain
> Understanding PR intent based on repository context
Full vid below 👇
During the talk, Harjot discussed:
> Second-order effects of Gen AI
> Agentic verification within our analysis chain
> Understanding PR intent based on repository context
Full vid below 👇
Fancy rounding out SF Tech Week with some good vibes?
Bring your board, borrow one, or just come for the coffee!
We'll be unplugging after a long week of talks, talks, and more talks with some nice cold waves 🤙
See you there?
Bring your board, borrow one, or just come for the coffee!
We'll be unplugging after a long week of talks, talks, and more talks with some nice cold waves 🤙
See you there?
October 9, 2025 at 2:17 PM
Fancy rounding out SF Tech Week with some good vibes?
Bring your board, borrow one, or just come for the coffee!
We'll be unplugging after a long week of talks, talks, and more talks with some nice cold waves 🤙
See you there?
Bring your board, borrow one, or just come for the coffee!
We'll be unplugging after a long week of talks, talks, and more talks with some nice cold waves 🤙
See you there?
One trillion OpenAI tokens.
One trillion tokens used to clean and ship quality code across the planet.
We’re truly honored to receive this, it’ll take pride of place in our new SF office!
🧡🐇
One trillion tokens used to clean and ship quality code across the planet.
We’re truly honored to receive this, it’ll take pride of place in our new SF office!
🧡🐇
October 7, 2025 at 3:39 PM
One trillion OpenAI tokens.
One trillion tokens used to clean and ship quality code across the planet.
We’re truly honored to receive this, it’ll take pride of place in our new SF office!
🧡🐇
One trillion tokens used to clean and ship quality code across the planet.
We’re truly honored to receive this, it’ll take pride of place in our new SF office!
🧡🐇
100k GitHub installs?
Completed it ✌️
Completed it ✌️
October 6, 2025 at 3:36 PM
100k GitHub installs?
Completed it ✌️
Completed it ✌️
If you want decisive, patch-like feedback, pick Opus 4.1 (or GPT-5 Codex).
But if you want broader bug coverage and better value, Sonnet 4.5 is the smart choice, it's more human, more cautious, and sometimes spots what others miss.
But if you want broader bug coverage and better value, Sonnet 4.5 is the smart choice, it's more human, more cautious, and sometimes spots what others miss.
October 3, 2025 at 3:58 PM
If you want decisive, patch-like feedback, pick Opus 4.1 (or GPT-5 Codex).
But if you want broader bug coverage and better value, Sonnet 4.5 is the smart choice, it's more human, more cautious, and sometimes spots what others miss.
But if you want broader bug coverage and better value, Sonnet 4.5 is the smart choice, it's more human, more cautious, and sometimes spots what others miss.
Sonnet 4.5 “hedges,” often using language like “might” or “consider".
It thinks aloud, unlike the direct style of Opus 4.1.
But 87% of its critical comments come with actionable code, and it’s strong at finding concurrency bugs.
It thinks aloud, unlike the direct style of Opus 4.1.
But 87% of its critical comments come with actionable code, and it’s strong at finding concurrency bugs.
October 3, 2025 at 3:58 PM
Sonnet 4.5 “hedges,” often using language like “might” or “consider".
It thinks aloud, unlike the direct style of Opus 4.1.
But 87% of its critical comments come with actionable code, and it’s strong at finding concurrency bugs.
It thinks aloud, unlike the direct style of Opus 4.1.
But 87% of its critical comments come with actionable code, and it’s strong at finding concurrency bugs.
Benchmarks show good progress: 41.5% of Sonnet 4.5’s comments flagged important bugs (vs. 35% for Sonnet 4).
That means about 4 in 10 of its comments either solved the key bug or flagged another truly significant issue.
That means about 4 in 10 of its comments either solved the key bug or flagged another truly significant issue.
October 3, 2025 at 3:58 PM
Benchmarks show good progress: 41.5% of Sonnet 4.5’s comments flagged important bugs (vs. 35% for Sonnet 4).
That means about 4 in 10 of its comments either solved the key bug or flagged another truly significant issue.
That means about 4 in 10 of its comments either solved the key bug or flagged another truly significant issue.
We benchmarked Claude Sonnet 4.5, and here is what we found!
Sonnet 4.5 gets you close to Opus-level performance at a fraction of the price, making it a perfect sweet spot for teams.
Here’s what’s new, surprising, and different👇
Sonnet 4.5 gets you close to Opus-level performance at a fraction of the price, making it a perfect sweet spot for teams.
Here’s what’s new, surprising, and different👇
October 3, 2025 at 3:58 PM
We benchmarked Claude Sonnet 4.5, and here is what we found!
Sonnet 4.5 gets you close to Opus-level performance at a fraction of the price, making it a perfect sweet spot for teams.
Here’s what’s new, surprising, and different👇
Sonnet 4.5 gets you close to Opus-level performance at a fraction of the price, making it a perfect sweet spot for teams.
Here’s what’s new, surprising, and different👇
Our Director of AI and King of Pop Culture, David Loker was recently on the Super Data Science Podcast!
He covered:
> CodeRabbit (Of course)
> Context Engineering
> Data security at CodeRabbit
> Vibe Coding
You can check out the full 1hr 17 episode in the thread below!
He covered:
> CodeRabbit (Of course)
> Context Engineering
> Data security at CodeRabbit
> Vibe Coding
You can check out the full 1hr 17 episode in the thread below!
October 2, 2025 at 3:16 PM
Our Director of AI and King of Pop Culture, David Loker was recently on the Super Data Science Podcast!
He covered:
> CodeRabbit (Of course)
> Context Engineering
> Data security at CodeRabbit
> Vibe Coding
You can check out the full 1hr 17 episode in the thread below!
He covered:
> CodeRabbit (Of course)
> Context Engineering
> Data security at CodeRabbit
> Vibe Coding
You can check out the full 1hr 17 episode in the thread below!
Find bugs, Fix bugs!
October 1, 2025 at 1:18 PM
Find bugs, Fix bugs!
Plus, Codex is faster!
Pipeline changes and leaner reasoning mean you get earlier, more helpful feedback.
Our north star: catch the bugs that matter, quickly, and with less noise!
Pipeline changes and leaner reasoning mean you get earlier, more helpful feedback.
Our north star: catch the bugs that matter, quickly, and with less noise!
September 30, 2025 at 3:30 PM
Plus, Codex is faster!
Pipeline changes and leaner reasoning mean you get earlier, more helpful feedback.
Our north star: catch the bugs that matter, quickly, and with less noise!
Pipeline changes and leaner reasoning mean you get earlier, more helpful feedback.
Our north star: catch the bugs that matter, quickly, and with less noise!
We also added severity tags and stricter filtering:
> Critical issues float to the top
> Refactors are gated
> Nitpicks are opt-in (you're welcome)
> Duplicates and “nice to have” suggestions get filtered
Less scanning, more fixing!
> Critical issues float to the top
> Refactors are gated
> Nitpicks are opt-in (you're welcome)
> Duplicates and “nice to have” suggestions get filtered
Less scanning, more fixing!
September 30, 2025 at 3:30 PM
We also added severity tags and stricter filtering:
> Critical issues float to the top
> Refactors are gated
> Nitpicks are opt-in (you're welcome)
> Duplicates and “nice to have” suggestions get filtered
Less scanning, more fixing!
> Critical issues float to the top
> Refactors are gated
> Nitpicks are opt-in (you're welcome)
> Duplicates and “nice to have” suggestions get filtered
Less scanning, more fixing!
Codex’s comments are action-forward, often including direct patch suggestions!
It’s great at tricky bugs like condition variable misuses, deadlocks, and subtle API traps & provides fewer filler comments.
It’s great at tricky bugs like condition variable misuses, deadlocks, and subtle API traps & provides fewer filler comments.
September 30, 2025 at 3:30 PM
Codex’s comments are action-forward, often including direct patch suggestions!
It’s great at tricky bugs like condition variable misuses, deadlocks, and subtle API traps & provides fewer filler comments.
It’s great at tricky bugs like condition variable misuses, deadlocks, and subtle API traps & provides fewer filler comments.
In our toughest PR benchmarks, Codex delivers 35% higher precision per comment vs. plain GPT-5, covers the same errors, and cuts comment volume by a third.
Reviews are now way snappier and more focused!
Reviews are now way snappier and more focused!
September 30, 2025 at 3:30 PM
In our toughest PR benchmarks, Codex delivers 35% higher precision per comment vs. plain GPT-5, covers the same errors, and cuts comment volume by a third.
Reviews are now way snappier and more focused!
Reviews are now way snappier and more focused!
After rolling out GPT-5 within CodeRabbit, we (and you) noticed reviews became too pedantic, dropping the signal-to-noise ratio.
The release of GPT‑5 Codex, plus the product changes we made, brings our ratio back home without sacrificing the ability to find the hard bugs!
🧵
The release of GPT‑5 Codex, plus the product changes we made, brings our ratio back home without sacrificing the ability to find the hard bugs!
🧵
September 30, 2025 at 3:30 PM
After rolling out GPT-5 within CodeRabbit, we (and you) noticed reviews became too pedantic, dropping the signal-to-noise ratio.
The release of GPT‑5 Codex, plus the product changes we made, brings our ratio back home without sacrificing the ability to find the hard bugs!
🧵
The release of GPT‑5 Codex, plus the product changes we made, brings our ratio back home without sacrificing the ability to find the hard bugs!
🧵