OK.
Ooh, hello. Hey, everyone. Thanks for joining in. Today, we will be talking about how to kill code reviews, everyone's favorite topic. I'm Ankit, co-founder of Aviator at Aviator VR, building an AI code verification platform. So we'll bring in some of the ideas and concepts that we build in our product. But first, let's dive in a little bit. A few months ago, I wrote a post on Laying Space about how to kill code review, creating a framework of a five-layers trust model. This model was focused around how do we actually, layer by layer, build trust into the code that can then be merged without needing line-by-line review.
And I got some things right, and I got some things wrong. So this talk will be about diving a bit more into it. I'm not going to talk about specific layers, but we will talk about some of the concepts that emerged from this session. So let's talk about the problem. We are looking today at the volume of code increasing every day, and we are struggling to keep up. So when we think about how long it will take us to actually stop reading code line by line, the reality is we've already stopped reviewing it.
There is 861% code churn. That means we are producing more code. The incidence-to-PR ratio is increasing. That means even if you are doing reviews today, they are not effective. So the median time of review is increasing. We have just increased the bottleneck from coding, now coding is solved, to now reviewing, where everything just gets stuck there. You're spending 4x the time that you were spending before just waiting for the reviews. And today, over 30% of changes are actually getting merged without a review at all. So let's think about it a little bit. Think about AI reviews.
Everyone is probably using some form of AI reviews today. When AI writes the code and AI reviews the code, why are we doing it in a UI? Right? We open GitHub, there are maybe two or three AI coding agents who are doing the reviews. The review goes back past the code, and you're talking back to the user or the agent, and it gets resolved. And you're doing back and forth with the agent. Where is the human here in the loop? You're eventually just looking at, OK, if AI has reviewed it, most of the things probably will be found. Let's do a skimming of it and merge it. So when AI reviews and nobody reads, we have configured the wrong thing. So let's take a step back.
Code reviews are not very old. They are maybe 15 to 20 years old. In 2006 was when Google launched Montrean internally, and they made formal code review a thing. If you think about it, Windows, back in the day, the first versions were actually built without reviews. But if we look carefully, code review is not just about code reviews. Obviously, we are looking at it as catching bugs, understanding conventions, identifying security issues. But code review is also about alignment. And this was one piece that was missing from my five-layers model that I talked about a few months ago.
So a big part of code reviews is knowledge sharing, mentorship, architectural feedback, onboarding, being able to collaborate. Again, if you're doing white coding, you're working on a solo project, this is not a talk for you. If you are working in teams, which I believe most of you folks are, collaborating in teams, you're not likely using completely dark factories, orchestrators where nobody looks at the code. You're actually collaborating in teams. You need to do knowledge sharing, which is the alignment part. And that is the most important aspect of reviews. So for semantic accuracy, we can build better tooling, but alignment must survive.
So let's dive into alignment. What does it mean? In today's world, can we actually think of a better model than aligning just based on reading line-by-line code? So most folks have probably heard about spec-driven development by now. Spec-driven development is, OK, we write a spec, it covers all the details, we pass it to an agent, it generates code, and then we verify. So what's wrong here? If you look back in 1970, this is what the waterfall model was. You have requirements, you have specification, you implement, and then you verify. But there's no feedback loop. The spec is written before we identify everything else. Right?
That's why today everyone still wants to use coding sessions, whether it's cloud code, codex, cursor, whatever you're using, you want to interact with the agents. And the reason you're interacting with agents is because there were certain things that are not clear in the spec, and we still need to capture that. And second, as you implement, you identify more issues, and you never go back and update the spec. Because if you're doing spec-driven development, it's already done. Once the spec is done, you expect the code will come deterministically. But guess what? LLMs are not deterministic. They're going to make decisions themselves.
So that's why spec-driven development is a great methodology, but it falls short in day-to-day software development. But there are some interesting aspects of this that we should carry forward. The most important part is the intent. And intent doesn't only live in the spec. Intent lives in your Jira ticket. That's the goal, right? It's where you express what we want to do. It lives in your PRDs. It's details. It's a plan. But most importantly, it lives in your PRDs today. This is where the real decisions are being made.
You start with, OK, this is a Jira ticket I'm going to look at. But you're going back and forth with the agent, and this is where all the user decisions are being made. But what we do today is we create a change, we create a pull request, and then we throw away the PRDs. And this is one of the things that we need to change. Let's also talk about semantic accuracy, because you're saying, hey, OK, I understand the alignment part, but there are still bugs in the code. Who's going to look at that? LLMs are also not great at this. And we already talked about how AI agents, reviewers, may not also be perfect. So this is where I introduce you to the concept of AI Slop Registry.
So think about this. If you're reviewing code today manually, and I expect everyone should be doing some degree of this, we are essentially possibly identifying the same issues over and over again. Can we actually capture these concepts and codify them so that we don't have to always create that review feedback one by one? You also have all of those things automatically identified.
The beauty of this is if you do it a few times, you now build a system that actually learns over time. So think of this as you're doing more training on top of the standard LLM that you have actually extracted, built on top of. So AI Slop Registers now can create better results because they're trained, they're learning from the review experience that you as humans are providing. Every recurring comment is now a guardrail that you don't have to review again. OK. So let's try to put both of these, alignment and semantic accuracy, together. It is two halves of the same problem.
We are trying to understand what are the core mechanics of review. How do we actually break it down into two parts, which is the alignment and the semantic accuracy, and bring them together into a single loop? So first thing is you take your session and you capture the user responses, and that essentially forms your acceptance criteria. The acceptance criteria then tied with your AI Slop Register that you are now constantly maintaining finally creates a test plan.
And this is a test plan that then gets verified. This is part of the system that we are building, the verification system, where it spins up a preview, takes your test plan, and makes sure it actually works end to end. Even if the code looks right, does it actually work? So this verification part has become interesting. And now the kicker is this is now your review surface. You're not reviewing code line by line, but rather you're looking at the evidence of what was the intent, did the user actually implement the capability that was defined in their intent, and did the behavior actually meet the requirements that we had in the acceptance criteria.
So you're still having the architectural decisions, you're still having these arguments, but the review surface changes.
So just giving a walkthrough of how we have built our system, the session becomes the criteria. So all these decisions that you are doing here with the agent, you're asking, you're providing this feedback. Even if there is a simple task, many times you're going back and forth, the agent will stop to ask questions. These are the decisions that we need to capture. This is the intent. This is what makes your review, makes your collaboration, more valuable. This is how you teach your junior engineers how to improve over time. These are the decisions that make a software engineer valuable today.
We convert those into acceptance criteria. This is where you can also leverage LLMs to do so. So you capture these user decisions and make sure you can actually create a test plan based on this. I know test plan creation is always painful, which is why I would always recommend people use LLMs for this purpose. And finally, the criteria plus invariant is what makes the test plan. And we build the verification systems to actually capture the test plan, run your previews, and be able to test based on this test plan. So even imagine if you're building a new feature, you don't have to maintain tests at all. This is creating tests in real time.
And this is where you can leverage the power of LLMs because test plan maintenance and creation can be very painful. But the value of the human in the loop here is the governance and the review part, and the part where you're reviewing the test plan and not the code. Right?
I know it's been over 20 years since we came up with test-driven development. This, in some ways, is closer to behavior-driven development, where the test plan is now something that even you can share with your product managers, your designers, everyone can participate in the test plan. And then the test plan is now in English. At the same time, we have deterministic verification that actually verifies whether the particular test criteria has been measured. Let's move on. So this is where I would say the system is not supposed to be perfect. It's deterministic where it can be, but LLM where you must.
Not everything can be deterministic. Not every system can be built in a way that is 100% built on deterministic systems. This is where you use LLM as a fallback. Let me give an example. If you are making a change in your web application, what you can do is it creates a test plan of what the behavior changes. Let's say you introduce a new payment form. So it creates a new payment form. The verification here is does the payment system change. An AI agent can go and browse through your application to fill out a form and capture screenshots as evidence, and then take those screenshots, as well as your database snapshots, to identify whether the criteria was met.
So the screenshot testing or sampling can then still be done by agents, but at the same time you're creating more solid evidence that now a reviewer can look at and build more confidence that this actually works. So now reviewers are reviewing the intent, not the diff. You're reviewing the intent decisions, what we set out to build, what we tried and rejected, and capture all of these things from the sessions.
Remember, capturing it from the sessions is the key. If we try to build it from the code, you end up in the same situation that we were talking about. I think Dex was talking about yesterday, which is if your code is built by the same agent that is actually building a test plan, it's not going to build a test plan that will actually catch issues. So that's why it's important to actually use the session information to build out a test plan.
You can discuss architectural decisions, how you're creating the data models, how these services interact with each other. So you've moved one level above. So instead of reviewing line by line, you're actually having discussions on architecture, which are very critical for any kind of collaboration. And then you look at the evidence, everything that was collected from the verification. So here's homework for everyone. Go home and mine your last 1,000 review comments and build out an AI Slop Register for the things that are repeatable.
So a vast majority of the comments that we're providing in your code review are something that we repeat over and over again. This compounds with every merged PR. Every time you capture something as a register, you don't have to capture that comment again. And this is where you can actually codify some of the best practices of doing code, maintain semantic accuracy, and at the same time, do not lose the collaboration part of the review. It does follow a J-curve. So the pain is real. You will have to spend some time to actually make it pay off because initially creating a registry can take some time.
And this is where I would recommend you folks come and try out our product. So code review is not just about code review. It is about really getting the alignment. And where we can build better tools is creating the semantic accuracy and defining your AI Slop Register. So if you remember one thing from today, remember code review is not just about code review. It is about getting the alignment.
And yes, we are piloting our new product called Verify. Please join and be our early design partners. We are working with a few companies to pilot out a new verification system. This combines both the alignment side of things, as well as building tools and capabilities for detecting semantic accuracy using the AI Slop Register. Thank you, everyone. Thanks for joining. Thank you, everyone. Thanks for joining. that I talked about a few months ago. So a big part of code reviews is knowledge sharing, mentorship, architectural feedback, onboarding, being able to collaborate. This is, again, if you're doing white coding, you're working on a solo project,
this is not a talk for you. If you are working in teams, which I believe most of you folks are, collaborating in teams, you're not likely using completely dark factories, orchestrators where nobody looks at the code. You're actually collaborating in teams. You need to do knowledge sharing, which is the alignment part. And that is the part which is the most important aspect of reviews. So for semantic accuracy, we can build better tooling, but alignment must survive. So let's just kind of like dive into alignment. What does it mean? Like in today's world, can we actually think of better model than aligning just based on reading line-by-line code?
So most folks have probably heard about spec-driven development by now. So spec-driven development is like, okay, we write a spec, you know, it covers all the details, we pass it to an agent, it generates a code, and then we verify. So what's wrong here? If you look back in 1970, this is what waterfall model was. You know, you have requirements, you have specification, you implement and then you verify. But there's no feedback loop. You know, the spec is written before we identified everything else. Right? That's why today everyone still wants to use your coding sessions, whether it's cloud code, codex, cursor, whatever you're using, you want to interact with the agents.
And the reason you're interacting with agents is because there were certain things which are not clear in the spec, and we still need to capture that. And second is, as you implement, you identify more issues, and you never go back and update the spec. Because like, if you're doing a spec-driven development, it's already done. Once the spec is done, you expect, like, you know, the code will come deterministically. But guess what? LLM is not deterministic. It's going to make decisions itself. So that's why spec-driven development is a great methodology, but it falls short in day-to-day software development. But there are some interesting aspects of this
which we should carry forward. The most important part is the intent. And intent doesn't only live in the spec. Intent lives in your Jira ticket. That's the goal, right? Like, it's kind of like where you express what we want to do. It lives in your PRDs. It's kind of like details. Like, it's a plan. But most importantly, it lives in your PRDs today. This is where the real decisions are being made. You start with like, okay, this is a Jira ticket I'm going to look at. But you're going back and forth with the agent, and this is where all the user decisions are being made. But what we do today is we create a change, we create a pull request, and then we throw away the PRDs.
And this is one of the things that we need to change. Let's just first talk also about semantic accuracy, because like, you know, you're saying, hey, okay, I understand the alignment part, but like there are still bugs in the code. Who's going to look at that? LLMs are also not great at this. And we already talked about how AI agents may, reviewers may not also be perfect. So this is where I introduced you to the concept of AI Slop Registry. So think about this. We are reviewing, if you're reviewing code today manually, and I expect everyone should be doing some degree of this. We are essentially possibly identifying the same issues over and over again.
Can we actually capture these concepts and codify them so that we don't have to always create those reviewing feedback one by one? You actually also have all of those things automatically identified. The beauty of this is if you do it a few times, you now build sort of like a system which actually learns over time. So think of this as sort of like, you know, you're doing more training on top of the standard LLM that you have actually extracted, like built on top of. So AI Slop Registers now can create better results because it's trained, it's learning from the review experience that you as humans are providing. Every recurring comment is now a guardrail
that you don't have to review again. Okay. So let's kind of like try to put both of these alignment and the semantic accuracy together. It is two halves of the same problem. We are trying to understand what are the core mechanics of review. How do we actually break it down into two parts, which is the alignment and the semantic accuracy, and bring them together into a single loop. So first thing is you go take your session and you capture the user responses and that essentially forms your acceptance criteria. The acceptance criteria then tied with your AI Slop Register that you are now constantly maintaining, finally creates a test plan.
And this is a test plan which then gets verified. This is part of the system that we are building, is the verification system where it spins up a preview, takes your test plan, and makes sure it actually works end-to-end. Even if the code looks right, does it actually work? So this verification part has become interesting. And now the kicker is this is now your review surface. You're not reviewing code line by line, but rather you're looking at the evidence of what was the intent, did the user actually implement the capability that was defined in their intent, and did the behavior actually meet the requirements that we had in the as acceptance criteria.
So you're still having the architectural decisions, you're still having these arguments, but the review surface changes.
So just giving a walkthrough of like how we have built our system, the session becomes a criteria. So all these decisions that you are doing here with the agent, you're asking, you know, you're providing this feedback. Like even if there is a simple task, many times you're going back and forth, agent will stop to ask questions. These are the decisions that we need to capture. This is the intent. This is what makes your review, makes your collaboration more valuable. This is how you teach your junior engineers on how to improve over time. These are the decisions which make a software engineer valuable today. We convert those into an acceptance criteria.
This is where you can also leverage LLM to do so. So you capture these user decisions and make sure you can actually create a test plan based on this. I know test plan creation is always painful, which is where I would always recommend people to use LLM for this purpose. And finally, the criteria plus invariant is what makes the test plan. And we build the verification systems to actually capture the test plan, run your previews, and be able to test based on this test plan. So even imagine if you're building a new feature, you don't have to maintain test at all. This is creating test in real time. And this is where you can leverage the power of LLM
because the test plan maintenance and creation can be very painful. But the value of human in the loop here is the governance and the review part and the part where you're reviewing the test plan and not the code. Right? I know it's been like over 20 years we came up with test-driven development. This in some ways is closer to behavior-driven development where the test plan is now something which even you can share with your product managers, your designers, everyone can participate in the test plan. And then the test plan is now in English. At the same time, we have now deterministic verification which actually verifies whether the particular test criteria
has been measured. Let's move on. So this is where I would say the system is not supposed to be perfect. It's deterministic where it can be, but LLM where you must. Not everything can be deterministic. Not every system can be built in a way which is 100% built on deterministic systems. This is where you use LLM as a fallback. Let me give an example. Like if you are making a change in your web application, what you can do is it creates a test plan of what the behavior changes. Let's say you introduce a new payment form. So it creates a new payment form. The verification here is does the payment system changes. An AI agent can go and browse through your application
to fill out a form and capture screenshots as evidence and then take those screenshots as well as your database snapshots to identify whether the criteria was met. So the screenshot testing or sampling can then still be done by agents, but at the same time you're creating more solid evidence which now a reviewer can look at and build more confidence that this actually works.
So now reviewers are reviewing the intent, not the diff. You're reviewing the intent decisions, what we set to build out, what we tried and rejected, and capture all of these things from the sessions. Remember, capturing it from the sessions is the key. If we try to build it from the code, you end up in the same situation that we were talking about. I think Dex was talking about yesterday, which is if your code is built by the same agent which is actually building a test plan, it's not going to build a test plan which will actually catch issues. So that's why it's important to actually use the session information to build out a test plan.
You can discuss architectural decisions, so how you're creating the data models, how these services interact with each other. So you're moved one level above. So you're essentially, instead of reviewing line by line, you're actually having discussions on architecture, which are very critical for any kind of collaboration. And then you look at the evidence, everything that was collected from the verification.
So here's a homework for everyone. Go home and mine your last 1,000 review comments and build out an AI slot register for the things which are repeatable. So a vast majority of the comments that we're providing in your code review are something that we repeat over and over again. This compounds with every merged PR. Every time you capture something as a register, you don't have to capture that comment again. And this is where you can actually codify some of the best practices of doing code, maintain semantic accuracy, at the same time, do not lose the collaboration part of the review.
It does follow a J-curve. So pain is real. You will have to spend some time to actually make it pay off because initially creating a registry can take some time. And this is where I would recommend you folks can come and try out our product. So code review is not just about code review. It is about really getting the alignment. And where we can build better tools is creating the semantic accuracy and defining your AI slot register. So if you remember one thing from today, remember code review is not just about code review. It is about getting the alignment. And yes, we are piloting our new product called Verify. Please join and be our early design partners.
We are working with a few companies to pilot out a new verification system. This combines both the alignment side of things, as well as building tools and capabilities for detecting semantic accuracy using the AI slot register. Thank you everyone. Thanks for joining.
Thank you everyone. Thanks for joining.