skip to content
Video cover art for CodeRabbit with Santosh Yadav
Video

CodeRabbit with Santosh Yadav

Santosh Yadav discusses CodeRabbit’s AI code review, Change Stack, triage, security scans, and context-aware workflows with Anthony Campolo.

Open .md

Episode Description

Santosh Yadav discusses CodeRabbit’s AI code review, Change Stack, triage, security scans, and context-aware workflows with Anthony Campolo.

Episode Summary

Anthony Campolo welcomes Santosh Yadav of CodeRabbit to discuss AI-assisted development and code review. Santosh traces his path from .NET development and engineering management to Angular teaching, NgRx contributions, and open-source community work. He recalls experimenting with AI tools, helping colleagues adopt them, and joining CodeRabbit in January 2026. Presenting its broader change-management approach, he demonstrates configurable pull-request reviews, Change Stack explanations of intent, and triage workflows for prioritization, reminders, and autofixes. The conversation turns to security concerns and barriers facing defenders. Santosh says CodeRabbit’s deep scans can identify vulnerabilities, including previously unknown ones, and describes plans for scan credits and access for open-source maintainers. He outlines findings dashboards and scan scheduling before discussing accessibility checks and community MCP connections. Santosh emphasizes linking repositories and issue trackers to improve context, adjusting settings, and teaching review preferences through feedback. After discussing recent conferences, upcoming GitHub Universe plans, and community events, he introduces Angular Inspector DevTools and Monorepo Doctor. Anthony connects those projects to his own repository-management challenges, and they close by discussing future streams and staying in touch.

Speakers

  • Anthony Campolo
  • Santosh Yadav

Chapters

00:00:01 - From .NET to Angular and Open Source

Anthony Campolo reconnects with Santosh Yadav after appearing on his stream about Redwood several years earlier. Santosh describes nearly 18 years in engineering, beginning with .NET desktop applications before moving into web development. He recalls working at a Mumbai startup, becoming an engineering manager, and changing jobs to return to an individual-contributor role in Pune.

Teaching Angular online and in person becomes a turning point. Santosh explains how learning from free resources motivated him to give back through open source. A conference encounter preceded his first NgRx pull request in 2019, followed by Angular contributions and writing for Angular in Depth. He also notes moving from India to Germany in 2022.

00:06:09 - Adopting AI Tools and Joining CodeRabbit

Santosh recalls gaining early access to GitHub Copilot through the GitHub Stars program and experimenting with AI outside work while his employer restricted its use. Once company policy changed, he helped colleagues adopt the tools, although he remained critical of early generated tests. He describes preferring an editor-based workflow where he can inspect the code and supplement usage limits with his own API access.

The discussion shifts to his relationship with CodeRabbit, which began before he joined. Santosh describes meeting the team at GitHub Universe, connecting them with open-source maintainers, and considering an AI-focused role while living in Germany. Conversations around Universe in 2025 led to his January 2026 start.

00:13:04 - Configurable Reviews and Change Management

Santosh describes CodeRabbit’s shift from a code-review product toward what he calls agentic change management, encompassing reviews, triage, security, and Change Stack. In the dashboard, he shows repository onboarding and pull-request comments, then walks through configurable summaries, changed-file lists, walkthroughs, sequence diagrams, and estimates of human review effort.

He says reviews can surface related issues and pull requests, suggest labels and reviewers, and include optional poems or prompts for coding agents. Anthony appreciates the concise, scannable structure. Santosh explains organization-level settings with repository overrides and identifies supported Git platforms, including GitHub, GitLab, Azure DevOps, and Gerrit, while noting some self-hosted usage.

00:18:56 - Change Stack Explains Pull-Request Intent

Santosh introduces Change Stack as a response to larger AI-generated pull requests whose authors may not fully understand the changes. Rather than requiring reviewers to reconstruct purpose from code alone, he says the interface explains intent and context. He compares this with the way formatters and linters previously reduced style debates, leaving reviewers more room to discuss what a change should accomplish.

The demonstration brings explanations and review comments into one view, with filters for comment severity and controls for submitting reviews, closing requests, and merging. Santosh also describes asking a CodeRabbit agent for clarification, although he does not demonstrate that interaction. Both speakers discuss how large migrations make understanding changes increasingly important.

00:23:26 - Triage Rules, Priorities, and Autofixes

Santosh frames triage as a response to teams facing more pull requests than human reviewers can comfortably process. He distinguishes low-risk changes from critical work that still needs human attention, then shows rules for merging lower-risk changes, closing stale requests, sending Slack reminders, and highlighting urgent fixes that have gone inactive.

Anthony asks about agents resolving review feedback before a person returns to the pull request. Santosh says autofix workflows can address CI failures and comments from CodeRabbit or human reviewers. He then demonstrates a prioritized inbox with explanations, assignment controls, and review actions. Anthony reflects on how these interfaces could reshape workflows traditionally handled inside GitHub.

00:28:51 - Deep Security Scans and Defender Access

The speakers discuss reports of recent security incidents and Anthony’s experience rotating API keys, using those concerns to introduce security scanning. Santosh says CodeRabbit’s deep scans can identify vulnerabilities, including previously unknown ones. He previews up to $50 in scan credits and plans to extend access to open-source maintainers.

Anthony raises concerns about defensive security prompts being refused and access requirements creating friction for legitimate users. They argue that protecting shared dependencies benefits downstream users. Santosh discusses scan costs and scheduling, while a dashboard label prompts a brief clarification about the product’s credit units.

00:35:26 - Maintainer Access and Scheduling Security Scans

Asked about Solid, Santosh says he has contacted its maintainers and describes working through trusted relationships to provide security access responsibly. He also mentions conversations with Nuxt and pnpm maintainers.

Santosh walks through security findings, repository dashboards, and scan history. He says enabling security review can add checks for issues such as exposed secrets. Because deep scans incur costs, he suggests choosing a schedule based on repository activity and security needs, offering monthly or twice-monthly runs as examples rather than presenting daily scanning as necessary.

00:38:16 - Accessibility Checks and Community MCP Connections

Anthony asks whether the review infrastructure could support specialized checks for accessibility or performance. Santosh says accessibility findings are already part of CodeRabbit’s reviews, then introduces MCP connections as a way to supply additional tools and information. He highlights E18E’s ecosystem-performance work while discussing the problem of AI selecting outdated or poorly maintained dependencies.

Santosh describes using E18E and internal connections involving Linear, BigQuery, and Google Cloud, with examples such as retrieving data or inspecting logs. The dashboard also includes Google Analytics and options for adding connectors. He clarifies that the listed MCP servers are community projects surfaced through CodeRabbit’s interface, rather than all being owned by CodeRabbit.

00:41:56 - Better Context and Learning Review Preferences

Santosh identifies missing context as a common weakness in AI-assisted reviews. He recommends linking related repositories, especially when frontend and backend code live separately, and connecting systems such as Jira or Linear. He says these inputs help CodeRabbit assess changes more meaningfully. The settings discussion covers premerge checks, organizational configuration, and optional presentation choices such as tone, poems, and emojis.

Anthony begins connecting his own repositories and asks about upcoming work. Santosh identifies security as the immediate focus, then explains feedback-based learning. He says developers can reply to an unwanted suggestion with their team’s preferred approach so CodeRabbit records that preference, emphasizing that teams should actively communicate their engineering conventions.

00:47:21 - Conference Travel and Developer Community

Santosh recounts attending All Things Open the previous week and meeting maintainers and students. Speaking from San Francisco, he discusses attending AgentCon and MCPCon, including a WebMCP talk that interested him. He then outlines plans for an in-person ODSC West workshop, other local events, and GitHub Universe the following week.

He explains that Universe matters to him because of GitHub’s role in his career and the chance to reconnect with community friends. Asked about CodeRabbit’s community, he mentions Luma events, office meetups, Discord, and a planned November 11 Berlin event. Anthony reports starting a deep scan and begins wrapping up before inviting final project links.

00:52:44 - Angular Inspector, Monorepo Doctor, and Closing

Santosh introduces Angular Inspector DevTools, an open-source project for humans and agents with MCP and CLI interfaces. He describes support across Angular-related environments and says a 1.0 release is planned after remaining fixes. He also presents Monorepo Doctor, drawing on recurring problems from years of monorepo work, and describes a cleanup skill tested against a codebase with hundreds of issues.

Anthony relates these projects to his own collection of similar repositories, duplicated agent instructions, and inconsistent dependencies, recalling earlier ChatGPT and RepoMix experiments. They discuss consolidating shared patterns, invite feedback on Santosh’s tools, and briefly consider future streams. The episode ends with final thanks and farewells.

Transcript

[00:00:01] - Anthony Campolo

All right. Hello, everyone. Welcome back to AJC and the Web Devs. We got a new guest here, Santosh. I was on your stream. You invited me on to talk about Redwood actually 3 years ago, which was one of the last streams I actually ever did introducing people to Redwood because I'd done it all throughout 2020, 2021, 2022, and then 2023 was when it started to fade out a bit. But, um, yeah, we'd love to hear about your background, um, how you got into development, how you got into open source, and then we'll talk about where you work, CodeRabbit, which is a very interesting product.

[00:00:36] - Santosh Yadav

Sure. Uh, thanks for the invite. So, uh, yeah, as you mentioned, I mean, I, I just realized it's already been 3 years like we did the stream together last time. So, wow. Um, and, uh, so a little bit about myself. My name is Santosh and I've been an engineer for like close to 18 years now, and I started my journey as a .NET developer. I was, you know, Windows guy, so I was actually building Windows desktop apps when I started. I think a lot of, a lot of people I speak— spoke to, they, they somehow started with .NET because there was only 2 options, right? .NET and Java. So a lot of people started with those, and, uh, I moved into like web, uh, because ASP.NET MVC came around 2012 and we wanted to build a new product. And then I started exploring, you know, jQuery because of course everyone was writing jQuery back in the days. Uh, still a lot of people still use it. Uh, it's, it's running the internet still after so many years. And, uh, I, I had like a short— not short, but I think close to 5-year stint at a startup in Mumbai where I ended up working with a lot of Microsoft technology and started like, you know, for a few years I was engineering manager and then I got bored. I was like, no, this is not what I want to do. I want to write code. So I went back into IC. So I had to change my job because they, they didn't— they were not like ready to actually move me back to IC. So I was like, okay, let's move on. Time to say bye and then move to another city, um, in India, uh, it's Pune. Um, and then started as an IC again and around, and around the same time I actually started teaching Angular, which was like probably the turning point for me. And I was like, okay, this Angular technology sounds good. I mean, I, I started learning and then I started teaching.

[00:02:39] - Anthony Campolo

Was this like in-person kind of meetups and like workshops or were you doing it online? Like what was the teaching environment?

[00:02:45] - Santosh Yadav

So both I was doing actually online as well as there was in person in Pune where I used to teach. And then I realized, you know, I made enough money— I think not enough, but yeah, I made decent money, I would say, uh, without, without investing much of on from my side because I was just learning everything for free, right? I mean, I was not paying anybody, uh, and I was like, oh, you know what, And then I moved to a job where I was doing Angular full-time. And then I realized, you know, oh, so Angular gave me so much. So I think I should go and, go and give back something to the community, right? And it was around 2018, I wanted to actually go ahead and contribute. But there was no clear way. So I was like, oh, how should I contribute? And because my— when I actually went to GitHub repository on Angular. And, you know, they— the GitHub page where you go and see like how many people have committed, how many contributors are there. And somehow in top 100, like I couldn't see any from anyone from India. I was like, oh, this, that is, this is disappointing. So I was like, you know what, I will do that. I will be the first one. And, uh, that was my motivation because, you know, Angular gave me so much. And then 2019, finally in 2019, actually I went to my first ever conference just to attend the conference. It was not like I was speaking or anything. It was just to attend and meet the Angular community. Uh, I ended up meeting, uh, one person from Angular team and went back home.

[00:04:23] - Anthony Campolo

AngularConf, right?

[00:04:24] - Santosh Yadav

Sorry,

[00:04:25] - Anthony Campolo

you're talking about AngularConf, like the official Angular conference? Is that what you're talking about?

[00:04:29] - Santosh Yadav

I know, no, it was—

[00:04:30] - Anthony Campolo

which conference are you talking about?

[00:04:32] - Santosh Yadav

Uh, I won't name it because, yeah, yeah, but yeah, I mean, it was back in India. And then of course, like, came back home within 5 days, like, I raised my first PR for a library called NgRx, which is a state management library.

[00:04:53] - Anthony Campolo

Sure.

[00:04:54] - Santosh Yadav

And, and then I started like doing some PRs for Angular. People saw me on internet, and then I started contributing to some some blogs. So there was a community where they used to write blogs, and it was a very I think big community because they they were doing some amazing amazing you know blogs or tutorials around around Angular. It was called Angular in Depth, which was pretty famous back in the days. I think it's it's not used anymore or. It was sold to somebody else. That's—

[00:05:28] - Anthony Campolo

oh yeah, I recognize the site. Yeah,

[00:05:31] - Santosh Yadav

yeah. And then, uh, I started writing for them, and, you know, that's how, like, people started— like, I started getting more involved with the community, started talking to more people, started building, you know, uh, more, uh, relationship with other people from other countries. So that's how it started.

[00:05:51] - Anthony Campolo

Cool. And am I right that you've lived and worked in India your whole life?

[00:05:58] - Santosh Yadav

Yeah, so I was in India till 2021, end of 2021 basically, and I moved to Germany, uh, in 2022. So, uh, yeah,

[00:06:09] - Anthony Campolo

I think— yeah, I think we talked about this very briefly when I was on the streams. That would have been like a year before, before that. Cool, very cool. Um, I'd be curious, um, where— when did you first start getting into like AI tools and things like that? Like, how did you transition into that? And, um, were you like an early adopter? Were you someone who's like, I'm a senior dev, I don't want to have a thing write code for me?

[00:06:35] - Santosh Yadav

Uh, I, I, I think that, that, uh, I got lucky. I mean, I was part of the GitHub Stars program, uh, since 20— uh, 2020. And, you know, uh, GitHub had, uh, some early advantage because they launched GitHub Copilot way early before Cursor even came. And it was just like autocomplete, right? And I had access to AI tools from that point of time, but it was just autocomplete. And then Cursor came, and then of course everyone started fighting for like their place in the ecosystem. Of course GitHub was improving as well. So I had like access to the tools before, so I was able to experiment. I was able to see like how it is going, how it is moving ahead. I was not allowed to use it because I'm sure like just let's talk about a year ago, many companies were resistant, like they don't want their developers to use this tool. And now everyone is like, no, no, no, you should be using the tool. So I was working at the— at Celonis at that point of time. And they were like, no, you cannot use the tool, or even you cannot use any AI tool because it was related to customers because our customers needed the information like more and more customers started, started asking like if you're using AI, Uh, to generate the code. And company don't want you to be in that weird situation, right, where they've come to know from some third party that, you know what, people, developers are using AI. So it was like restricted. I was using it like for my own purpose, like, you know, open source side, or maybe, uh, just to build some, some fun stuff on the side, but not for work. And last year things changed. I think more and more developers wanted to use AI. So my company said, yes, uh, let's start using AI. And, uh, I was involved in the early adoption phase where I was actually teaching other developers like how to use it, uh, in a good way, how you can use it properly. But models were not good. I remember like trying out Sonnet when it came out and I was like, no, I mean, I, I would still prefer writing code by hand rather than using this, this model because it's just, it was just shit. Uh, the kind of test I used to get as a reviewer, I was like, oh wow, I should burn this test before it goes on production. Um, but, uh, I got, I mean, I was lucky that I had access to all those tools and, uh, I was not paying anything. Like I was like getting everything for free, uh, just trying out. So yeah, that, that, that's how it started for me.

[00:09:07] - Anthony Campolo

Very cool. And do you have a certain daily driver like Codex or Claude Code, or you still in an editor or you switch around?

[00:09:15] - Santosh Yadav

Oh, so I mean, a lot of people do ask me that question. So I'm somebody who still likes to see the code. So right now, uh, what I do is I'm using GitHub Copilot app, um, with Opus 5.5, and it doesn't give me like a lot of credits because I think GitHub has— I don't know, I mean, uh, whatever limit they have is like, it's, it's probably it's gone in 10 days for me. So I'll most of the time I also use my company's because I had tried push code for CodeRabbit, right? So in that, that time I just go ahead and use Anthropic's 5.5 model with GitHub Copilot because they— you can bring your own key. So GitHub allows you to do that. You can bring your own key and just start using it,

[00:09:58] - Anthony Campolo

right? Yeah, awesome. So how did you first hear about CodeRabbit and what made you excited to work there?

[00:10:07] - Santosh Yadav

It was back in '23 because I, I was— I, I had heard about CodeRabbit, uh, when they started. And 2024 was one first time actually I came, I, I met the team. So in 2024, I was here for GitHub Universe, and, uh, one of my friend who was actually leading the DevRel at that time here, and I ended up having a chat with him. He said, yeah, he sent us like, uh, He also wanted to connect with other open source communities because, you know, CodeRabbit invests heavily in open source. We almost gave close to $1.3 million in cash since we started.

[00:10:46] - Anthony Campolo

That's awesome.

[00:10:48] - Santosh Yadav

Yeah. So he— I mean, so I was like connecting my friend with other open source maintainers so they can like, you know, sponsor them, support their work. And 2024 and then 2025, at the end, I was like, you know, I was at Celonis still doing all those stuff, and I realized, you know, this is the time I should move into AI. And but which company? Like, of course, like, I have to go ahead and figure out, like, where should I work? And I, I live in Germany, so that's also a restriction, right? Uh, probably not many AI companies, uh, are coming out of Europe, and not many companies in US will hire me, uh, if I'm in Europe. So, uh, I just reached out to my friend. I was like, hey, you know what, uh, do you think like we— is there anything for me right now? Because I see the company had raised Series C and they had just announced that they are going to do, you know, $1 million of fund for open source. So I was like, yeah, let's see. Like, I mean, why not? Like, just— I was just taking my chances basically. So, and, uh, CodeRabbit was doing great. I mean, I, I know the community around open source. I used to speak to some of my friends who were using CodeRabbit, and they're like, you know what, that CodeRabbit is amazing. And I was like, you know, I mean, if people are telling me from outside— I, of course, I knew the people. I knew our CEO, I had met them before even I had joined. I knew my friend for ages, for years before he joined. I knew some of the teammates there. So I was like, you know, people are telling me from outside as well that this company is amazing, I think I should take my chances. And as we spoke, and I was actually— it was the same time again, GitHub Universe 2025, and I was supposed to travel anyways because I don't like to miss GitHub Universe. And I met them again, I had a discussion around like what they are looking for, what they want me to do. And yeah, and then it was just a single discussion. I came to the office and ended up meeting my future teammates, and I started like this year in January then.

[00:13:04] - Anthony Campolo

Very cool. So won't you give like a high-level explanation of what is CodeRabbit for someone who's like never heard of it or maybe just knows that it's some sort of like code review thing?

[00:13:16] - Santosh Yadav

Uh, we are changing. I think we, we are not just the code review tool anymore, or not a code review company, just, just the code review company anymore, because AI is changing a lot of stuff, right? I mean, I'll of course talk about it, but, uh, right now we are calling it as, uh, calling us as an, uh, agentic change management company, because code review was one part. Like, I think a lot of, a lot of people are using code review, like you know, they started using code review at some point of time. But there are so, so much more things which needs to be solved around this entire change management process which we have. So code review is one part, then we have triage, which is again another part of the product or entire lifecycle, I would say. We have security, we have code, you know, something known as Change Stack, which can explain you the changes in the PR. So this is like 4 things which we have been building, and of course we'll build more because, you know, we want to cover the entire change management process, uh, in an agentic way.

[00:14:23] - Anthony Campolo

Interesting. Yeah, I have a bunch of questions, but I think that would be a good time to pull up the dashboard, um, so I will help kind of explain better what it does, especially now as you mentioned they actually have a bunch of different features. So, um, I would say you should start with like what's the the kind of basic things someone would do if they're like getting set up, or like, uh, the what was the code review thing, so people who already know that can kind of get, get a sense of it.

[00:14:50] - Santosh Yadav

Yeah, sure. Uh, so, uh, I think most of, most of the open source maintainers or of most of the companies who started using us, they started at code review. So, you know, you can actually, uh, connect your repositories. For example, let's see here. So I, I can connect my repositories and, uh, I can start doing the review. So if I go to, let's say, any of the Recent PRs of mine. So, and probably open one. So for example, this one, right? So as soon as someone actually raises a PR, I mean, we actually start to review, review your changes and we put some comments. You can see the comments, developers can resolve this. So this is where we started. Yeah, so, uh, and now, you know, from this dashboard you can actually just onboard any new repositories you have. And, uh, a lot of things were actually built around this. So there was a dashboard, there was a reporting thing, and, uh, there was settings. For example, I can actually control a lot of, uh, configurations. Okay, let me move this. Yeah, so I can control a lot of configuration from the screen as well. Like how I want my review to look like. For example, here, so I can see preview, and these are the things which I will probably see on my PR. So I can see what are the files which have changed. I will see a walkthrough, I will see a sequence diagram, I'll see how much the estimation is, how much generally it will take to someone as a human to actually review this particular PR,

[00:16:29] - Anthony Campolo

right? Yeah, the AI will blow through in 5 minutes, but it'll always tell you this will take a half hour.

[00:16:35] - Santosh Yadav

Yeah, for sure. And then of course we can also give you possibly related issues. So let's say if you have your Jira or Linear, you know, we can also do that, and possibly related PRs as well. I think this is one of the amazing features because it happens, right? Sometimes actually people try to fix the same issue in multiple PRs, and CodeRabbit can actually find and tell you that, okay, this This PR possibly at the same. And then we give you related labels, suggested reviewers, and poem. I think this was— this went viral when we started. A lot of people started liking poems.

[00:17:12] - Anthony Campolo

Yeah, it reminded me of Open Sauced, which was BDuggie's project. And what they used to do is every PR would have like a GIF that like represented how they felt about the GIF or something like that. But what I really like about this template is that it make sure to keep it very minimal, high level, like bullet points, short titles, because it will avoid that issue where like Claude will just write you like a 5-paragraph block of text and like, here's your PR. So like this gives you a really clear structure, so it's like scannable, which is great.

[00:17:45] - Santosh Yadav

And then of course you can always, you know, control what you want to see. Some— I speak to some of the open source maintainers, they're like, you know what, Uh, I don't want to have the prompt or AI prompt, for example. So if I go to any PR, I mean, there is a prompt for AI agents, but a lot of open source maintainers are still of the opinion that we should not give it. So you can like control all this. You can enable, disable whatever you want to see on the, on the PR. And then this configuration is something which can be extended. So it's like org-level settings. It's an— it works like an inheritance. So, you know, this, these are some common org-level settings, but you can override. So you can actually go and, uh, go to your repository, and then you can also change the settings for specific repositories as well.

[00:18:34] - Anthony Campolo

Do you have to have a GitHub repo, or are there other ways to, to get the code there?

[00:18:40] - Santosh Yadav

We do support multiple Git platforms, so GitHub, GitLab, Azure DevOps. Uh, we have Gerrit now, um, and we— I think there are some companies who are using self-hosted as well. So there are multiple ways. Yeah,

[00:18:54] - Anthony Campolo

cool.

[00:18:56] - Santosh Yadav

Uh, and yeah, so this is, this is, uh, the first thing which we started. And, you know, we— the next possible thing which we, we could build was like, you know, uh, people are raising PR, people are using AI to actually review the PR. Or, uh, sorry, to actually open the PR. And sometimes they don't even know like what is happening in the PR. Uh, people are just, you know, because they have a lot of credits, so they just want to actually go ahead and use it. Uh, so we started building our next product, which was Change Stack. I'll just open this, and this is also free for open source. So, you know, uh, what we did here is we wanted to explain what is the intent behind the changes. Because as a developer, we all know, like, if, if I'm working in a team, possibly I know like what technology is being used, and possibly I also know the technology. So I don't want to read the code to understand, uh, what's happening, because I, I can read the code and sometimes code doesn't make any sense. We have all been there, right? I mean, you see the code and it's like, oh, what is this? But if I tell you the intent behind the changes. Probably most of the time this is what we do. If you go back and think about the code review, no one was reading the code. Like, of course there are, there are some nitpicks and everything which was happening, uh, and then linters came and removed that noise. Like, no one was talking about standards and, you know, tabs versus spaces. People just started using tooling, Prettier, ESLint, ESLint. So it reduced like more, more and more noise, and then people started discussing about the intent. Okay, what you want to achieve, like why you have this OPR open, and that's, that's the discussion. So now this Change Stack actually does exactly the same thing. So, you know, you can actually go ahead and, uh, see like, oh, you know, uh, what is this, like why, uh, what it— what is happening here. So, you know, you can read the intent behind it or in AI word, you can say it's the context. It gives you the context behind these changes so you can read it and, okay, and then you can figure out, okay, if this is needed or not. And we can also see all the comments here because at some point of time we realized, uh, back in the days when people were doing code reviews on GitHub, I mean, I think it was very rare, right, where we were doing hundreds of comments on a PR. But think about now, I mean, with AI, there are times that I think frequently I would say there are more than 10, 15 comments easily.

[00:21:38] - Anthony Campolo

Yeah. And you have it do much larger migrations that have like many more lines of code added and removed at once.

[00:21:46] - Santosh Yadav

Exactly. And no one is able to figure out or read it on the GitHub UI because that's,

[00:21:53] - Anthony Campolo

and this is like, it solves a huge part of that issue right here. Now this is awesome.

[00:21:58] - Santosh Yadav

Yeah, so we said like, okay, let's bring everything together. So we started like bringing— like, you know, you can also filter out the comments based on is it a net pick or it's a major or minor. So we try to create everything a single view and you get everything. You can submit the review, you can merge from here, and you can close pull requests, merge pull requests. So single UI for doing everything with the context as well. And, uh, I don't have the access to the agent, but I can talk to CodeRabbit agent as well in this window. So, you know, you can— I can ask CodeRabbit saying, okay, probably I, I couldn't understand what is this, just explain me again. And CodeRabbit will give you all the details. So, uh, this is, this is the next thing which we started building because of course PRs are not getting smaller, like PRs are just— the size of the PRs are increasing. And we have all seen that, like, because I think, uh, it's, it's, it's good thing and bad thing as well. Good thing for people who understand what they're doing and bad thing for people who don't understand what they're doing because they just want their AI to go and go crazy and just write like million lines of, uh, code migration and just figure out—

[00:23:13] - Anthony Campolo

image of them giving the AK-47 to the monkey. I'm sure, I'm sure you've seen that. People always say like me and Claude code with that image or Claude codes with a junior dev, something like that.

[00:23:26] - Santosh Yadav

Exactly. That's the perfect way to describe it. So yeah, I mean, uh, and then we realized like now as we started working with more and more enterprises and more and more companies, they, we also realized, you know what, now I have thousands of PRs. Now, because everyone in the organization is a contributor, everyone wants to code,

[00:23:49] - Anthony Campolo

rip tokens.

[00:23:51] - Santosh Yadav

Yeah, exactly. And companies started coming to us and saying, hey, you know what, what should I review first now? Because we don't know. I mean, we are literally lost because there are just too many PRs to review. And even though CodeRabbit is able to actually go through and do the first pass and It's ready to be merged, but you know, most of the companies, I'm sure like the companies you also work with, no one is actually fully trusting the AI right now. It's like there is, there is a human in the loop, right? So someone who has access and goes ahead and merges the PR until unless it's like small changes. So if we all have this checks like, oh, you know, if this is a doc change, probably just go ahead and merge. If CodeRabbit approves it. But if it's a critical change, someone wants to have a look at it. But if that someone is actually busy and like has hundreds of PRs to review, it's like, then what happens to the critical fixes or critical features which should be delivered? So this is where we started building something called triage. So triage gives you that inbox And you know, you can actually— where is that? So, okay, so I can build the rules, uh, in— if I go to the triage, so I can build the rules like, you know, uh, much lower risk changes. This is what I was talking. So, you know,

[00:25:23] - Anthony Campolo

like rename a variable or something like that, you know.

[00:25:26] - Santosh Yadav

Yeah. And then, uh, I can also auto-fix PRs, close stale PRs. Send Slack reminders. We use it internally a lot and priority urgent fixes. So if I, if there is any P0 or P1, we can tell you, okay, these are the P0, P1, and it's not active for last 24 hours. So you can actually, you know, send it in your Slack as a message. So

[00:25:50] - Anthony Campolo

actually the, the auto fix PR one sounds really interesting because that's one thing I was thinking of is when I first saw people using this tool, it was like the human loop thing where they would get a comment, then they would like read it, respond to it. Whereas like, I want my agent to just be like fixing stuff and just like, you know, keep figuring out, oh, there's an error, you put up another PR, and then once it's kind of done all that, then you have one kind of thing at the end. And he's like, that's what the autofix PRs thing would be for, kind of get your PR through any of those weird bugs that's going to hit along the way.

[00:26:22] - Santosh Yadav

Yeah, so autofix PRs does actually exactly the same. Like now you get the comments, like you have a lot of comments. And if, uh, even your CHX, right, if you see fix spelling CHX as well, so it can actually auto-fix all the CI errors as well as any CodeRabbit comments or human review. So it will still keep working until unless everything is resolved and it finally says, okay, I'm done, there is nothing to be, nothing to be done now, just go ahead and merge your PR. So, uh, this is triage. I think, uh, there is an inbox as well for triage. Let me show you this. Uh, this is the inbox, so this is how it looks So it is right now, I think I have 4 PRs, but they are all P3. But I can go back and anyone like who has access to their teams, right, they can actually go here and see like, okay, what are the PRs which is waiting for me? If I— there is a— there are any P3 or P0s, so, you know, I can just action it. And we give you the reason like why we think this is P3 or P4 or P0. So for example, this one is— I mean, this is my open source one, so this has been waiting for your action. So I need to go ahead and do something because there is nothing done on the PR yet. So we give you everything, you just go ahead and decide like which ones you want to review first. And you can also distribute within the team, right? You— sorry, because of course there will be multiple people working on within the team, and then you can distribute like And you can assign it from here. So you can say, okay, let's— as an engineering manager, you can actually just assign it to somebody else in the team and then they can take care of it. Or you can just submit review from here as well, or merge it if you think it's very important.

[00:28:08] - Anthony Campolo

It's really interesting how your platform is in a position where you can kind of rethink a lot of the things that is making people want to drop off of GitHub, because GitHub has so much legacy and it's really hard to review code in this new stage environment, it's like crashing all the time. Whereas like you just create a couple like streamlined things that just like fix so many issues with that whole workflow, and you're kind of slowly taking up more and more of that real estate, which makes a lot of sense because you got in at the PR level. So it's like you're already hooked into everyone's repos, so anything the GitHub platform could do in theory, CodeRabbit could do if you would build out the feature. Yeah, it's very cool.

[00:28:51] - Santosh Yadav

Uh, and then the other thing which we started building was security. So of course we realized like now, um, and we— I think we were talking offline as well, right? I mean, we're talking about, uh,

[00:29:03] - Anthony Campolo

yeah, I'm curious about this aspect for sure.

[00:29:06] - Santosh Yadav

Yeah, so I think, uh, we saw the last week— I think it was not last week, right? This— it was this week.

[00:29:13] - Anthony Campolo

So yeah, 2 days ago was the PHP one, and then I think yesterday or today there was like a Ruby As well. And the day before, the PHP was Railway getting hacked. Like, I have a bunch of Railway apps, so I had to roll all my API keys, and it was, you know, not a big deal. Um, nothing catastrophic happened. But yeah, it's, uh, it's been a weird, weird week for security. Yeah,

[00:29:33] - Santosh Yadav

I mean, just, just think about it. I mean, I was thinking like maybe this, this happened within a few weeks, but no, it was actually just a few days, and every, uh, the entire world went crazy. I mean, I was, uh Uh, I think even the Prime Minister of Japan, he posted like, you should go ahead and update your PHP version. And I was like, what the hell?

[00:29:54] - Anthony Campolo

Yeah, the crazy thing was the NSA tweet about it. Like, why is the NSA tweeting about PHP, you know? But it makes sense. I'm sure there's all these government sites that were written on PHP when no one had any other idea how to make a website in government, you know, 15 years ago. And that's how all the sites are now, I bet. That's a lot of stuff, a lot of critical data probably.

[00:30:15] - Santosh Yadav

Yeah, exactly. So, uh, so this— we started actually working on security. We launched our security version 1.0 this year, uh, you know, and then we have been improving the product. We, uh, you can actually do, uh, scans, you can run deep scans. This is very interesting. I think this is one of the things which I love. Uh, I think I can maybe, uh, do Uh, let's see, I can select because, uh, for now I think this is, this is something which we are launching. So, you know, you get $50, up to $50 for free. You can run this deep scan. Sorry, uh, you can—

[00:30:54] - Anthony Campolo

that's tight. That's great to have. No, I'm just saying that's good getting the $50 credit.

[00:30:59] - Santosh Yadav

Yeah, yeah. So, uh, we will be announcing it soon, uh, but if you are on stream, I think you, you got early access now. You can actually just go ahead and run this deep scan and it can actually actually find vulnerabilities in repositories. And this is very, uh, pretty, uh, you know, thorough scan, deep scan. Uh, it can actually go ahead and find day-zero vulnerabilities as well. So be— I mean, just, just to share, I mean, of course, like, if someone else gets access, uh, we have, we have seen people actually abusing this thing, but we are actually putting this in hands of uh, developers or enterprises now where they can actually figure out day-zero vulnerabilities before someone hacks their system.

[00:31:43] - Anthony Campolo

Yeah, I mean, that should be how any of these models should be able to do it. And something I was talking to you before the stream is that just recently, um, because there's this new OpenAI like Daybreak thing, which is kind of like how Claude has Mythos, which are like these gated things that do security stuff now, so they're starting to refuse prompts in just regular Codex to just do a basic— it's just a prompt saying audit my site for security. It's not even asking it to go hack it. It's something totally benign. Hopefully that'll just be a bug that they'll work out. But it makes me worry that we're creating this environment where we're giving an asymmetric advantage to the offense instead of the defense because we're hobbling the defense. Because I tried to sign up for Daybreak and I needed a YubiKey. I have a YubiKey. It's like 3 years old and I plugged it in and it says this YubiKey is not compatible. You click this link and buy a YubiKey, and it takes like 3 days to get here. And Thibault was like tweeting, it takes 15 minutes to set up Daybreak. I'm like, BS, that's a lie. That's a straight up lie.

[00:32:46] - Santosh Yadav

Exactly. That's crazy. So, uh, and then, uh, we are actually also going to probably work with some of the maintainers so, you know, they also get access to this. Um, we are, we are going to announce that maybe in a few days, uh, once we go live with this thing. So keep a watch. I mean, we will be working with maintainers as well to make sure like they also get the access to these models because, you know, of course we need to help those maintainers. Who else will be doing that?

[00:33:14] - Anthony Campolo

If we don't protect open source, we're not even protecting ourselves because we're using these open source libraries. So we are incentivized ourselves to make sure that the open source libraries are secured.

[00:33:25] - Santosh Yadav

Exactly. Yeah, I think everyone needs to support open source. So, and we have seen how critical is that, right? I mean, TanStack got hacked in first, Nx got hacked in the past, and it directly impacts the enterprises. So if enterprises think they have the best system in the world and they are safe, I think, uh, we are learning that lesson, lesson that no one is safe now. Uh, everyone, like, you know, these hackers have access to AI and they are also using it for not helping, uh, or not making the world better, but they are making it to, you know, abuse the system and then hack the data. And that, that might be your, your data. So, um, yeah, I mean, you can run— just run the deep scan. Um, as I said, like, you can also schedule it, uh, and just go ahead and run your first scan for free. I mean, if you have a smaller repository, the cost shouldn't be more than $50 for sure. But for bigger repositories it may be more. And if you are an open source maintainer, you are handling a critical project, we will—

[00:34:31] - Anthony Campolo

it's called carrots.

[00:34:33] - Santosh Yadav

Yeah, I mean, uh, sorry, which one?

[00:34:36] - Anthony Campolo

Sorry, so click the credit card and cost details, it says estimated carrots, available carrots. Is that the term for like credits?

[00:34:44] - Santosh Yadav

Ah, uh, yeah, I think, uh, where was that? Sorry.

[00:34:48] - Anthony Campolo

So, so click credit and cost details again. OK, and so you see how it says all these units, they're all in carets. So the estimated carets, 9.92 carets, available carets, 100 carets. Yeah, carets applied to estimate, 9.92 carets. So like I'm assuming caret is a credit.

[00:35:05] - Santosh Yadav

Yeah, yeah, it's credit actually. So, but yeah, as I mentioned, we will be, we will be working with open source maintainers soon and we are already talking to a few of the maintainers we know. Uh, to see like, uh, what it— what the feedback is. And we'll be, uh, opening it for more, more, more and more open source projects as well.

[00:35:26] - Anthony Campolo

Are you in touch with the Solid team?

[00:35:28] - Santosh Yadav

Yeah, uh, already, already messaged, uh, yesterday. So I'm going to work with them as well. So they, they agreed to actually use it. Uh, so I'm— we are just— yeah, we are figuring out like how we can actually give it to, uh, you know, the real maintainers and not someone who can hack your data or hack your repository. So it's, it's like we, we are just right now working with maintainers we know, and, uh, we already know SolidJS, so we have a good relationship with them. So David, uh, is already in our Slack, so we messaged him saying, you know, uh, let's, let's try this out, uh, and we will be working with them. So we are going to work with Solid, and we are also talking to some other maintainers like you know, Nuxt team. Uh, we're talking to PNPM as well. So yeah, let's see. Uh, but yeah, we, we are coming after like, you know, all this critical issues. Uh, and, uh, and I think this is, uh, also like, you know, you have, uh, you can also see all of, of your open findings on the screen once you run the deep scan. Of course, you need to know like how many open how many issues are open, how many closed. So you can get a single look at the entire thing. And, uh, there will be like, you know, right now CodeRabbit does some level of review, uh, which is to find mostly bugs. Uh, it can also find the security vulnerabilities once, uh, you know, you enable this on your repository. So it will be able to add extra, let's say for example, if you have exposed secrets or anything, it can do that as well. So for example, here I think we have repositories which you, which you can connect, and you will be able to see like if you have any number of findings available on this repository or not. So it can give you all the information. So this is like mostly dashboard around the same data. Security reviews, this is again the dashboard, and we can also show you the scan history because deep scan is a little bit I mean, it's, it's very, uh, it's going to cost money, uh, so it's not like something which you have to run daily, most probably, because I mean, once you got like a day zero vulnerability, it's not like it's going to happen in few days. At least it will take few, maybe weeks, until unless you merge more code. And you can schedule it like maybe running monthly or maybe half a month. It's up to you, like like what you think, like how important the security is and how often you get those security issues. So you can actually schedule this and then you can see all the information here. So yeah, so this is, this is the security thing.

[00:38:16] - Anthony Campolo

Yeah, it's really interesting because like you're first saying like we're for code review, that's how you started, then you built kind of this other stuff. And when I think of your code review, you can do a broad code review or you could do certain kinds of code review like a security review review in particular would be a really big one to do. So I'm thinking of like other things because now you have this whole infrastructure, we have all the code review stuff in, so you could have like an accessibility review that would do a lot of like stuff with, with like Playwright or something, or, uh, like a performance, you know, review or things like that. So are those things that you could see also being like features, or do you want to just basically let the review itself like kind of allow the user to find those types of different reviews themselves?

[00:39:01] - Santosh Yadav

We already do actually accessibility, uh, like issues. We do find accessibility issues if you, if you have, like, if anyone who has used CodeRabbit, uh, I have, I have used it for my projects and it already does that. So accessibility is something which is built in, but you can also bring— I think this is one, one thing which we missed talking about, but we have something known as connections. You know, you can also bring your MCPs So let's say if any, any, there is any MCPs you want to use, uh, you can actually bring them. For example, E18E, uh, which is, uh, it's, it's an ecosystem performance effort which is driven by one of my friends James. So e18e.dev. So we recently added this. This is very good in case you are, you are a developer and you want to make sure you are using like the dependencies which are real because, you know, After this AI thing, somehow AI has actually downloaded, started downloading the packages which was built 5 years ago, 6 years ago. It is not maintained, but somehow your AI loves it for some reason.

[00:40:07] - Anthony Campolo

Yeah, sure. We'll just pick really old things that are probably like the training data or something. Um, do you use MCPs that you plug into CodeRabbit when you're like using the platform yourself?

[00:40:19] - Santosh Yadav

I do. I mean, uh, so right now I think I, I still need to do this connection again. I was using E18E, which is, which is nice. And internally we use, of course, like a bunch of MCPs, which is of course hidden because it's, it's our organization. But we do use Linear, we do have, you know, BigQuery or other, other MCPs, which— and even GCP. So for example, I think, yeah, BigQuery is here, right? So you can actually connect and get some data, ask for some data. So yeah, GCP is also here. For, for looking at logs. So, you know, I can, I can do all this stuff using the MCPs as well. Google Analytics. So you can actually just ask CodeRabbit if there is any information available. But yeah, that, and you can bring your own. So it's like, this is, these are some available ones, but you know, you can actually just add a new connector. And I think this should be setup origin. Okay, this is, this is cursor connection, but yeah.

[00:41:19] - Anthony Campolo

Does the CodeRabbit team own all of these or are these community MCPs?

[00:41:25] - Santosh Yadav

Like, so these are all community MCPs and we just have provided it by default.

[00:41:30] - Anthony Campolo

So yes, you're just basically providing a nice interface to kind of plug together because it's protocol, so it makes it easy to just get all the MCPs going. Huh, that's really cool.

[00:41:42] - Santosh Yadav

And what else? So connections, scopes. But yeah, I mean, we also have this now. I mean, a lot of, a lot of companies are using it, uh, already. So

[00:41:56] - Anthony Campolo

what do you think— let's go high level— when people do code review with AI, what are just some of like the first things that people do wrong or that they just don't anticipate? That they like instantly hit, hit like a big issue with?

[00:42:11] - Santosh Yadav

Yeah, I think a lot of companies, a lot of developers, when they start doing code reviews, uh, the biggest problem is the lack of context, you know. So when, when you're doing code reviews, try to bring as much information possible near to your code base. So for example, we have something known as, um, multi-repo analysis. Because not everyone is using monorepos, right? Even though I would love if people use more and more monorepos, and, uh, they start there. So let's say they, they have multiple repositories, and then they give half context, half baked context to your agents, even, even code review agents. So, uh, if you have some system like this, you can actually bring— let's go and see the configuration where Okay, back to work. Settings, I can just see. Yeah, so auto-linked repositories. So you can actually go ahead and, you know, link multiple repositories. This is auto-link, but you can also bring multiple repositories. So for example, let's say if you have a polyrepo with your frontend somewhere else, backend somewhere else, do that configuration before you go ahead and start using CodeRabbit. Bring your MCPs. So let's say if you have— sorry, if you have a Jira or Linear or any other system where you have more information about your product, try to bring it to the system. It will— because it will be able to do the thorough review. One thing which we do really well is context engineering. We try to get all this information and then do the review, so which gives you in-depth and, you know, real bugs. It helps you to find real bugs, not just some nitpicks which we generally do, used to do it in past. And also, like, you should go ahead and revisit the configuration you have because there are so many things. For example, with the— we have premerge checks. You can actually enable some premerge checks inside your repository, inside your organization, and just control all these things from here. Just go ahead and— I mean, so if, if there are enterprises, of course we go ahead and work with enterprises to figure out like what kind and what pre-merge checks are useful for them, which are not, and we help them onboard everything. And then we have a fun section. I think everyone knows about it. Uh, you can set a tone. Uh, this is— and then you have poems, uh, which was, as, as we were talking about it, like, it was like a really popular feature back in the day. Uh, but it depends, like, the choice, like, uh, of choice of the teams, if even they— if they want to have fun or they don't want to have fun. So you can decide all this thing. We also generate emojis like, you know, ASCII or emojis. So you can decide all those things. But important things, uh, MCPs, and as well as link your repositories, right? So you— your code base gets more context.

[00:45:14] - Anthony Campolo

Yeah, I just created an account. I'm linking, linking my stuff now while you're talking. So I'm curious to run the, the security scan Um, are there any other high-level features or things that are kind of upcoming that you're working on and excited about?

[00:45:31] - Santosh Yadav

Uh, for now, I think, uh, security is the major focus, so we are excited to announce it soon. Uh, there will be, uh, there will be blog posts coming out, um, we'll share it. But one thing, one more thing which I wanted to talk about is like CodeRabbit can talk, like, you know, it can reply to your questions. So a lot of people think, oh, you know what, it is giving me wrong previews. But if you, if you try to go in detail, probably there is some actual issues. But in case, in case, just in case, like CodeRabbit gives you some suggestions which you don't like, because, uh, we, uh, we all, uh, you know, who is using AI right now and have been doing engineering for years, we have been talking about one thing which is, you know, test. We are telling people to not lose the test, the test about like how they write code or how they solve the issues. But AI doesn't know your test and you have to teach AI. So, you know, this is where, uh, you— CodeRabbit also has a learning feature where it can learn from you. Simple example, let's say if I go back to any PR comments here, I can actually reply to CodeRabbit saying CodeRabbit, uh, this is not right. We do it in X way. That's it. Now CodeRabbit goes ahead, adds it as a learning, and you will never get that, that review again. It will work as you expected it to. So a lot of people don't use this enough, I would say. So if you see something like this, teach CodeRabbit, teach AI, and it will be behaving much better after that.

[00:47:21] - Anthony Campolo

Awesome. Um, I'd be curious to hear about, um, like the conferences you've been going to. And, um, I know you're actually, um, you're, um, traveling right now, so be super curious to hear about that.

[00:47:36] - Santosh Yadav

Yeah, so I was at All Things Open last week. Um, we, uh, the entire team was there. It was fun. I mean, I, I, it was my first time at All Things Open and a lot of— met a lot of people. A lot of open source maintainers were there. Uh, huge community of students, I would say, who also came to the conference. I think All Things Open is known to support the local students as well, so they gave them free tickets. Which was nice. I mean, nice gesture. And now I'm back in SF. So I went to— I went yesterday to another conference called, uh, AgentCon and MCPCon, which is 2 days conference actually. So they are doing one, uh, there was day 1 yesterday and they— there is a day 2 today, but I decided to skip, uh, which is again, I, I saw some interesting talks, like interesting discussion happening about AI And the MCPs as well. Uh, there was a discuss— there was a talk around WebMCP, which, which sounds really interesting. I was like blown away by the use case, uh, of WebMCPs. I hope like more and more browsers start supporting it. I think it's experimental and still supported by Chrome. Uh, so Sarah Drasner was there talking about it. And, um, next week I'm— I'll be at few events. So I'm going to speak at one of the conferences which is happening here called ODSC West, Open Data Science Summit. So this is where I'm doing a workshop.

[00:49:06] - Anthony Campolo

And I'm not familiar with this one.

[00:49:08] - Santosh Yadav

Yeah, I mean, so some, some friends are doing it. So somehow they reached out to me and says, if I'm— they asked like if I'm interested in doing a remote workshop. And I was like, oh, I mean, I can do in person because I'm anyways attending. And so I'm doing an in-person workshop now. And we are doing some events. So if you're in city, if you're in town, uh, hit me up, uh, send me a message on maybe LinkedIn or X, wherever you are, and I can invite you to our event. Uh, so, and then GitHub Universe, of course, like that's why most of the people are here, uh, next week. Uh, so if you are coming to GitHub Universe, let me know. I mean, I'll be at the event as well.

[00:49:53] - Anthony Campolo

Very cool. So it's not like GitHub Universe is probably like your favorite conference?

[00:49:58] - Santosh Yadav

I mean, uh, I love to be there because it's— I think GitHub has helped me a lot personally in my career. I think everyone, right? I mean, most of our, uh, most of open source maintainers somehow own their career because GitHub existed, or most of the people— I mean, we know each other because GitHub existed, I think.

[00:50:16] - Anthony Campolo

Sure.

[00:50:17] - Santosh Yadav

Yeah. So yeah, I mean, and then I get to meet the friends from the community. A lot of GitHub stars will be there, um, and a lot of friends actually travel just to attend the Universe. So, uh, I love it. I love the vibes.

[00:50:33] - Anthony Campolo

Very cool. Awesome. Uh, are there any like community things that people should know about for CodeRabbit? Is it like a Discord or like do you do events, anything like that?

[00:50:46] - Santosh Yadav

Yeah, I mean, uh, you can actually— uh, we are on Luma. So if you, if you live in San Francisco, uh, you can just search for Luma Code Rabbit. I mean, we do a lot of events here, um, and you can just come, uh, to any of the events. But I think next one is November 11th in Berlin, but there will be more events, of course, uh, in the, in the, in the town. So if you are in SF, I think we do a lot of in-person meetups at our office. And yeah, I mean, you can, you can join there. We have a Discord. We have an open Discord for everyone. You can actually join our Discord as well. Most of our DevRel are very active on social media. So you can, if you have any questions, of course you can send me a message or anyone, anybody in the team, and they will be happy to help you.

[00:51:37] - Anthony Campolo

Very cool. Yeah. Awesome. We can probably start wrapping it up here. I'm really impressed with the platform and I'm excited to try it out myself. I did get the deep scan running, so fair, that would probably take a little while, but I'll let you know how it goes. And yeah, thanks so much for going over all of that. It definitely has grown a lot more than when I first heard about it because I remember when Aiden from Million.js, he had a big tweet where he like, I'm gonna try every single code review thing. And then they all kind of like, battle it out. And at the time, I remember already being aware of CodeRabbit as kind of like one of the big kind of offerings for that. And I find that usually that momentum will catch you, get you pretty far. And now, like, you have a lot of, a lot of people using this. Yeah, really good flywheel, it seems like. And it's very cool, the security feature especially. That's thing I'm most excited about. So, um, any last kind of a call to actions or links you want to leave our listeners with?

[00:52:44] - Santosh Yadav

Yeah, I mean, I'm actually building some, uh, things on the side, uh, so if, uh, they're— of course they are open source. So one of them is actually we are building an Angular dev tool for human and agents both. So if you go to my repository, it's in pinned repository, it's called, uh, Angular Inspector DevTools. So it's, uh, it has like MCP, it has a CLI, and it works with all the ecosystem. Like, you know, you can work, you can use it with Analog.js, you can use it with like SSR, you can use it with NgRx, NativeScript, Angular Native, and it works everywhere. So Single Dev Tool works everywhere, and we are still building it. So we are still, uh, you know, we want to do 1.0 release soon. But we are trying to fix a little bit issues here and there. But if you are using Angular, I mean, just give it a try. Help us, like, you know, improve this so more and more people will be able to use it. And maybe try it out on your own repository if you have one. It's, it's not going to— I'm not going to read your code or anything. It's just open source. So yeah, I mean, this is one. And Another thing which I'm working on is, um, something known as Monorepo Doctor.

[00:54:02] - Anthony Campolo

Because I saw this in, um, when you're showing the example for the different stages, I saw there's a skill that said monorepo skill. So that was— yeah, really it.

[00:54:14] - Santosh Yadav

Yeah, because, uh, I, I saw— I've seen like, of course I've been doing monorepos for close to 7 years and I keep seeing the similar patterns in every monorepo I So I was like, you know what, I've been— I think I've spent enough time fixing those issues. And I was like, okay, let's build this thing and make it open source. So this is like, uh, what I wanted to share with the world. So go ahead and give it a try if you have a monorepo. And you know, I just recently added like a cleanup skill as well, uh, which can actually work in a loop. Uh, I recently was working in our code base where we had like close to 800 issues. And you know, this is again like, this is not something which I just made up of on my own. This is something which I have tried, uh, used it, and then actually making it open source. So go ahead and give it a try if you have monorepos already available, and give me the feedback if it works or not and how we can improve, what more tools we can build. So I'm going to actually probably build more tools around monorepos And yeah, that's it. So this is the last thing.

[00:55:21] - Anthony Campolo

Yeah, that actually, that's something I'd like to try because I'm in that thing, you know, where you find yourself reverse engineering something from first principles. You're like, why am I like reinventing the idea of a monorepo? I had all these Git repos that each I would have a specialized kind of, you know, agents file and all these bespoke rules. And I started building up like 4 or 5 of them and then I was like, okay, there's things I want to share between them. Like skills and things like that, but it's not, it didn't start as a monorepo, so I'm kind of gluing them all together. So yeah, that's a, that's a problem that I'm running into right now with my own stuff that I'm trying to simplify. 'Cause it's like, I, I ended up, when I, for, there was a while, especially like 2023, even before we had like these Codex and Claude Code, I was just vibe coding through like the ChatGPT app. I was just like writing code and I was using this tool RepoMix that would like dump a whole bunch of files into context along with docs. So you could basically have it like one-shot specific file changes. And then I would just kind of copy and paste the thing in. And it just was so fascinating that I kept like starting new projects to like see how it would start from the total beginning. And then I found different patterns that worked. So I'm like, okay, so this is actually how I want to structure my apps now. But now I have like 5 of them that are all kind of really close but different in certain like obnoxious ways. In terms of different versionings and did they use this database or this thing. Having a monorepo that would actually standardize all those things would solve a whole bunch of headaches right there. Cool. Thank you so much for joining us. We'll end it here. Just stay on for a bit after we end the stream. Thank you everyone who watched and hopefully Dev is still alive. Um, I'm not sure what happened to him, but, uh, yeah,

[00:57:11] - Santosh Yadav

so

[00:57:12] - Anthony Campolo

yeah, um, so thanks for coming on. And, um, yeah, I'd love to come back on your stream if you're, if you're looking for guests, or me and Dev could both come on and chat about whatever. So, um, do you still regularly stream?

[00:57:24] - Santosh Yadav

Not, uh, not, not right now, but yeah, I'll be back. I'm planning something, um, so I'll get back to you. I mean, I'm, I'm thinking about something, so I'll end

[00:57:33] - Anthony Campolo

up— pressure, but I just, I just enjoy streaming. This is like what I do for fun on a Friday night. So perfect.

[00:57:39] - Santosh Yadav

Yeah, but, uh, have a good day ahead and enjoy your weekend, and I hope to see you later. All

[00:57:45] - Anthony Campolo

right, and we'll catch you next time, everyone.

[00:57:48] - Santosh Yadav

Yep.

On this pageJump to section