This is from March 2026, but it does match what I've encountered professionally. Code review was always a significant bottleneck and is even more so with agentic coding. We've had discussions about leaning more on automated review, but it hasn't translated into any practical improvements so far. I think senior engineers find it very difficult to give up on line-by-line review, no matter how much better the agents have gotten.
I think if there is a major change in code review culture, it's not going to happen first in established engineering teams. It'll happen in startups that adopt a no-human-review policy from jump and then maintain it as they mature. I'm sure that's happening, but it's likely invisible to most studies and surveys at the moment. If it proves successful, and such companies start shipping significantly faster, the rest of the industry will be forced to adopt.
> I think senior engineers find it very difficult to give up on line-by-line review, no matter how much better the agents have gotten.
Why do you feel like that? Agents have got better but better how? Looking at the frontier model output, I’d say the same myopic approach that causes long term issues without frequent course correction is still there.
I agree. The latest models are getting quite good at validation, as long as they are clear on what the requirements are. Recently I got frustrated that while we have a huge test suite I noticed the LLM was hardly ever breaking a test with its changes (PRs only added tests, never changed old ones) which seemed suspicious.
I had it do a full audit. There were about 100k lines of tests, of which it determined about 60k were near useless. Many were overly mocked, some mocked to the point that they really weren't testing anything at all. It deleted those and wrote more expansive scenario tests to replace them, caught a lot of bugs on the way and fixed those too. Total time was about 2 hours.
I think humans need to review requirements, specifications, and test scenarios. And for some industries and uses they're still going to have to review every detail. But for a typical CRUD app, I think we're at a point where we can already set the process and quality bar higher than ever, in addition to moving faster.
I chuckle a little bit at this specification thing. We used to do this. We used to write big binders of specifications, and then we'd set about implementing them. We stopped doing it because it didn't really work. We'd learn things in the process of building which negated the value of the big specs we wrote. That's why we stopped doing it. This is where continuous improvement, CI/CD, tighter release cycles, etc. were all born.
Glad to see someone else notice this. I dread looking at or dealing with AI written unit tests because of the incredible amount of whiteboxing that goes on, though usually for the opposite reason because AI testing is almost entirely whitebox testing which fails if some internal details change
Just have AI write requirements then another AI write specs based on the requirements then another AI write tests based on the specs and another AI write the implementation based on the tests then another AI write the PR based on the implementation then another AI approve the PR and another AI deploy approved PRs to production!
> I think senior engineers find it very difficult to give up on line-by-line review, no matter how much better the agents have gotten.
I actually dislike traditional code review. In part this is because most of our code review tools are very cumbersome to use (GitHub's and GitLab's web interfaces in particular).
But also because humans aren't all that great at code review when it comes to correctness and security. Part of the problem is that you don't have access to the author's mental model for non-trivial changes and have to reconstruct it, otherwise your understanding of the program logic will be inadequate.
The problem is that purely agentic code review has orthogonal problems, and typically remains subpar when it comes to judging architecture and design considerations, compared to human reviewers. Both false positives and false negatives also remain a concern.
My best results so far have been with agent-assisted code review in a proper IDE, helping you to quickly construct a mental model and to analyze the underlying logic (in addition to having agents do a first pass for correctness issues).
> But also because humans aren't all that great at code review when it comes to correctness and security. Part of the problem is that you don't have access to the author's mental model for non-trivial changes and have to reconstruct it, otherwise your understanding of the program logic will be inadequate.
Code review is not really about correctness and security. We have testing and other verification process for these. Code review is about idea discussion mostly and correcting a few stuff that may have slipped through the author's initial review.
Your PR wants to change the software in some way. My review is to comment on how you've implemented that change, like did you miss some areas where you needed to change, does your description is actually what is happening (either a cursory reading or through testing), or are you doing some unnecessary change.
IMO, the initial reading shouldn't take more than 10 minutes. More than that (so a big diff) it should have been discussed upfront, and the design agreed. If it's an important area, a more careful reading should be done about the tests and relevant stuff.
But generally, if a review takes more than 10 minutes, there's some failure of communication going on (reviewer not apt to understand the change, big diffs that should have been discussed outside of the PR, obtuse code,...)
We shouldn't give up on line-by-line review for serious applications.
My company's leadership is oscillating every few weeks on AI goals. Their most recent story is "we're expecting developer velocity to approximately double over the new 2-3 years as people get better at using AI". But of course that's completely non sequitur. People on my team spend about 10% of their time writing code and 90% doing design and analysis. So if AI doubles our code writing velocity, then it will make us 5% faster overall.
The point is that while good to remember, and specifically Fred Brook’s “No Silver Bullet” article about it, there are second order effects.
If coding becomes 100x cheaper, what’s going to happen next? At least to some degree, design and planning was there since coding is expensive. (Not just coding, debugging also got squashed by AI and was an expensive part of the job)
This is why the original comment seems pretty spot on, organizations will learn how to skip more phases, and startups are probably a place where this will be happening faster. (Not clearly though, a lot of big tech are happy tokenmaxxing to try and not stay behind those startups when they come)
Brooks' point was more about accidental vs. essential complexity. You seem to be implying that design was only ever an artifact of "expensive" typing, but design exists to prevent systems from collapsing under different failure states.
If the cost of emitting syntax collapses without adequately resolving the essential complexity then you have not eliminated the design phase and you're just externalizing your failure modes onto downstream teams and systems.
...and they are not using LLMs to help speed up design and analysis?
I'm no booster but this is most of what I currently spend tokens on so it's maybe not as absurd as it sounds. Maybe those gains are possible and without having to sacrifice serious code reviews too.
As a senior engineer that is struggling to give up line-by-line reviews, I agree with your take.
I've spent literal decades where one of the most valuable practices has been code reviews, for many dimensions. It is one of the best ways to ensure quality production releases. It is one of the best ways to transfer skill to junior engineers. It increases tribal knowledge throughout the engineering org by exposing others to systems they may not be working in. The number of benefits to a traditional human-centric engineering team is hard to overstate.
But I'm starting to feel that instead of starting at one extreme (heavy handed reviews) and moving towards less review, it probably makes more sense to start at the other extreme (no review at all) and move towards a brand new practice tailored for the new agentic reality.
It is looking increasingly likely that we don't need to modify code review for the agentic era, we need some completely new practice that hasn't yet been invented.
It's really about hamstringing the AIs so they can't write crap. Writing code is not a problem anymore, so anyone with an idea can write a script to do what they want. But if they try to connect two ideas together, now they have a data modeling problem. This is where everything seems to fail with AI coding (largely because the original author had the idea piecemeal).
You don't give up on line-by-line review just because the human is getting better either, and senior engineers want other people to do line by line review on their own code. It's not a matter of how good or bad the person or program generating the code is.
> I think if there is a major change in code review culture, it's not going to happen first in established engineering teams. It'll happen in startups that adopt a no-human-review policy from jump and then maintain it as they mature.
I've been hearing this for more than a year. But if these startups are all moving at superhuman speed, a month instead of a year from idea to a shipping product, where are the products? Every time I ask that, people talk about the amazing velocity gains they experience for the thing they are about to ship. It never does, unless we count throwaway hobby projects at GitHub - by that metric, AI is adding a lot of software. And I guess never-downloaded slop-games on mobile app stores.
But if you actually go shopping for consumer or enterprise software that's camera-ready, the landscape is basically the same as in 2022.
I lean towards a hybrid review process where humans accept their limitations, focus what they're good on(higher level concepts and architecture of the proposed change), and let the LLMs do the rest.
Doing a line by line review as a human is silly now.
It's not code review that's slowing things down. It takes time to get the thing you want, and then you have to be reasonably sure that it's right anyway, or you risk shipping a piece of crap to production and tanking your own reputation. Even on teams that rubber stamp PRs I see people pushing drafts and waiting to merge until they feel comfortable with what they've vibed out.
If you're right (and you probably are), we're in for a wild ride further down the quality death spiral.
The era of user telemetry and conversion tracking already decoupled product design from customer experience. No need to wait for customer's to articulate their experience or desires or for discerning analysts to amalgamate and ponder their opinions. The numbers tell you everything you need to know! Just throw 3000 pounds of spaghetti at the wall every month and wait to see what sticks.
But now you really were bottlenecked on engineering.
Only 3000 pounds of spaghetti per month? And throwing more bodies at it barely even makes a difference? Oof.
Of course the solution is to just flush pureed spaghetti-slop through as hose as fast as you can. Brilliant!
What could wrong? The users are sure to love it. We'll know because they keep clicking. That's how they tell us the truth, even as every forum fills with louder and louder complaints!
You have to classify each code review based on what it changes and decide how to handle it, agent review and/or human review. You can have an agent do that.
My experience with prototyping games is that you need very strong human driven guidance and review in the beginning, then you can increasingly take your hands off the wheel.
If you take it off too early you end up with shit and will spend weeks fixing all these baked bad assumptions your project is built on top of. I spent weeks fixing some of this shit then the LLM revived a not insignificant portion of it in a later change because it got "confused" (its words) because some of the comments it previously made were left around.
I think non-senior engineers don’t realize just how much of a burden is barely reviewed ai generated code.
I’ve seen a few projects that heavily rely on ai with not very good (though senior) devs driving it - and its a mess. Not the code. The resulting product. Code is actually sound as ai is pretty good now.
All because ai does code reviews and dump decisions go into prod
> I think senior engineers find it very difficult to give up on line-by-line review
i am pretty senior. I hate code reviews so much. They are so boring, easily the most tedious part of my job. Its such a chore. Who is clinging to line-by-line review. Ppl dont really give that much about their employers.
Code review isn’t a quality gate, it’s a way to maintain shared understanding of a codebase.
Throwing out code review is equivalent to throwing your hands up and saying you don’t need to understand anything about what your code is doing in its black box.
I don’t think anyone on this forum is half as smart as they think they are, especially in terms of systems engineering a black box solution. Myself included. Someone needs to understand what’s going on inside the Widget Factory until we have full Artificial General Intelligence.
I initially pivoted to fully automated code reviews, sometimes with a brief skim of things. I've been shifting back to mostly manual review. In the background I'll also run some agentic scans, and once I'm done I'll look to see what the agents had to say. And then I go back in by hand armed with that knowledge and verify it & come to my own conclusion.
Why? Because I started to run into way too many AI generated messes in codebases. The agents are really good at some things and complete ass at others.
It really depends on the context. A react button background color - meh - the critical part of the backend where a new 300ms delay will cost the company millions then yeah.
> the critical part of the backend where a new 300ms delay will cost the company millions then yeah.
If it is so critical, why don't we have tests and benchmarks for it?
Oh right, because "Greg the super-coder" never wrote any of this down and it lives in his head so he can be the arbiter of change.
It was toxic before agentic coding, it's still toxic now.
Document your systems. Add tests and benchmarks for what is important.
We've always known this, we have a dozen or more textbooks written about it. But we always gave it a pass because "Greg the super-coder" was producing more value than he was hurting.
That's no longer the case. Your best coder is now Opus or Astra
I hate code reviews, but I owe it to my coworkers that I know will read my PRs to review my own generated slop line by line. Most of these agents are still overly verbose and I'm in a situation where no, not every single edge case is relevant enough to warrant the additional loc or complexity introduced.
As is standard for academic papers, this is an attempt to describe the state of the world as of when they did the research. But then it takes time to write it up and get it out, after which newspapers report it as current events.
In this case, we're looking at a picture from 2025. The models involved in the main agent results were primarily Claude 3.5 through Sonnet 4.5, GPT-4.1 through GPT-5, and Gemini 2.5. And the paper's design means you're mostly looking at the very initial effects of deployment, before companies developed good review processes.
They've updated it for the agentic code era, and have found basically the same thing. Code can be created faster but that has moved the bottleneck to review.
Man, I'm probably going to expose my own inexperience here, but like, did anybody actually expect code reviews to catch and prevent bugs on a line by line basis? Like ever?
I never have operated this way. Code reviews were a basic sanity check, a chance to implement best practices, and a chance to teach these things to juniors.
If I was concerned about something breaking, I'd duscuss test coverage.
I’ve caught bugs in code reviews before, but I would agree that it’s probably not the most crucial thing you should strive to do with code reviews.
That said, I think I’ve done a tremendous amount of PREVENTING bugs from being introduced via code review. Caught all kinds of weird code patterns, duplication, wrong levels of abstraction, etc. Essentially, the kinds of places that bugs like to hide and grow in.
I still think that’s valuable, and it’s the kind of thing that models generally aren’t great at, to this day. It generally relies on having a large amount of context about the project and about what’s being solved, so if you wanted to do this via automated review, you maybe COULD, but it would probably get pretty expensive, especially on larger projects.
It was more important for giant C or fortran code bases. You will legit catch pointer errors or memory mgmt stuff, going line by line. Once Sun busted in with Java is when I really saw the shift to skim reviews. So much stuff is caught by compilers now, or SAST, that using dev time to find code smells is mostly a waste of their time. And it's not like the old way caught everything, it's just that the stuff we take for granted being caught by compilers or Sonarqube now, just used to get shipped to prod and cause abends and production support calls at 3am.
I read this paper several days ago, when my director shared it. Several things jumped out at me:
- the data is mostly from 2025 and earlier
- the baseline mean lines of code per month was 5017, the estimated increase was 1494 lines per worker-month
- the baseline is 5.34 PRs per month, and increased by 1.22
- the baseline mean for code review is 7.03 days
The second and third bullets are confirmation that this is old data. In late 2026, people aren't delivering 30% more lines of code. If people were just producing 30% more lines, they'd still be reading every line. My point isn't that "more lines is good". I'm worried about a few too many lines of code that no one reads. My point is that this shows the data is old.
For the latter, I'm a bit annoyed by using mean, median would be more valuable, but I think it's safe to say that the engineering culture was already pretty bad. If your average pull request sits for 7 days, you evidently don't care about shipping anything, and agents aren't going to change that.
I'm noticing AI catching many bugs in code that previously went unnoticed. Some of them are important, and it's great to catch them early.
The other 11 items in the list of 12 would probably never be hit in practice, but now that you know, you have to sort out how to handle each (even if that just means documenting, filing, or ignoring: you have to understand each of the disparate 12 issues first.)
Before it was all coding on its own, I asked a model for bugs in a new project, saved me the 15 minutes of new web project I waste every time trying to remember obscure first time setup things you dont touch for years in under a minute.
I suspect theres a lot of hours saved that are otherwise wasted by devs when some simple thing throws off their implementation completely. Basically we cant measure the saved developer hours by LLMs because it could be literally anything.
I’d strongly disagree with one aspect. Homemade personal use software has Definately skyrocketed. Projects I never thought I’d even have the time for are now possible.
Yes, I am literally inundated by cool new tools I've built that I never would have before.
I'm literally building a black-box client + server emulator for one of my favorite old MMOs, remotely, from a coffe shop in SF as I type this, on my laptop at home via Claude remote-control.
Any study that wants to say anything about the impact of coding agents on whether more software is being created or not can't limit itself to what firms are doing that create a footprint of "work events — including GitHub coding activity, Jira issues, and Google Calendar events".
This is software that would not existed without coding agents: https://fluximp.org
Sure its a coding agent itself, but I would never have tried to create it manually as I would have expected it to be too much work to see it through as solo dev. There were no issues created anywhere, no calendar events, and there is very little github activity (I primarily used Mercurial on a private server; git is just for distribution).
I started it with Kilo Code and after a couple of months it was purely self-modified and has been since then. I was a bit worried that it wouldn't work on Linux and I'd either have to use another agent or use it running on my Windows box to modify the code on my Linux box to get it working, but it just worked so the streak of being self-modified has continue.
I want to make an indie game and I'm working towards that with the project I'm using to benchmark FluxImp. My first job was as game dev and I took a shot at indie game dev 20 years ago (failed; not fun). Its something I had long written off as too time consuming to even start, but with coding agents I no longer feel that's true.
All of the stories about non-software engineers making small software that is useful to them also seems unlikely to generate the kind of footprint that lends itself to being easily studied.
I don't see how AI coding agents are going to create 'more software' until the PO's and PM's of the world step back and let developer's communicate directly with the stakeholders. Or the PO's and PM's could cross-train on AI coding agents.
We are still being stressed to create all stories with such fidelity that 'any' developer can pick it up and implement it in the timeframe allowed for the size the story was given. When the story is written in that detail, we may as well copy it into an AI agent and have it implement it. But we can't do that, so we have people who have SME write up story's that sit in the backlog for 2 weeks until a developer can bring it in, so that they can spend a day implementing it (really 2 hours with 2 hours of meetings and 4 hours of video games) because the smallest pointing size we can give is 1 point and people aren't allowed to take on more than 8 points a sprint.
How can we expect things to get done faster when the team that reports work up to senior management put's such strict limitations on our throughput?
There are many ways to say the same thing, that code throughput wasn't the only bottleneck, communication is by far the hardest, most brittle and time consuming part of software production.
We need to start becoming better writers, capable of expressing complex thoughts concisely and with technical precision. Software production has become a sibling field to linguistics, more than it already was.
I've spent most of my time working on tools to keep my agents on track, and keep code legible. Where I'm probably most strict in the tooling is on documentation, which makes the code review much more manageable.
I'm completely uninterested in the opinion of a video game journalist on what does and doesn't work well in writing software. The author knows about as much about writing code as he knows about performing heart surgery
Ars Technica in general has also been catering to their knee-jerk anti-AI audience. I used to find their comments section somewhat interesting but now it's the same old tired, outdated dunks on AI.
Yes, instead of reviewing 1,000 lines a day I now "review" 10,000 lines a day. My brain is not 10x faster, so either I spend 10x more time reviewing code, or I just opt to accept that AI is probably right based on the complexity of the problem. However I realize that if we just let the agents loose on the codebase, everything would speed up like 100x.
Invariably once certain AI models in certain domains reach "full automation" quality this will happen, until now the human eyeball bottleneck is holding a leash on millions of tireless superhuman coders who have just a few narrow but egregious contextual awareness issues.
Given what I've been able to accomplish with even just local models at home, and my experience at my day job, I imagine this has almost everything to do with the organizations themselves and less to do with using AI to build software.
My experience is that agents are decent in coding a feature, solving a bug - doing specific well defined task. They are very bad at judging UI/UX and user flows through app. Like really, really bad. They can quickly put together vibe coded app but make it usable takes a lot of extra manual work.
Did anyone solve this problem? Any good approaches?
I confirm this diagnosis. The UIs are always clunky, disorganized, and kinda ugly. But they work well enough. I guess UIs are opinionated enough that we probably all customize them anyway. Maybe the agents know this so don't bother putting too much effort in? (Deep thought of the day: acquired)
Wrong thing to index on. Always bet on the slope, not on points in the slope. Things will only get better and better and factor in that downward slope thought while taking actions and pick ones that are favorable. The folks who focus on a point in time will be left behind, period.
But the longer the codebase survives the more hair and cruft it accumulates. My once beautiful logical paths get carved up by compounding requirements and edge cases.
That's what the AI was trained on. The end result - the hairy ball of wax that noone can look at sideways.
I often think of that whenever AI comes up with what I consider to be an irrational but plausible edge case.
Aside from other comments mentioning flaws and cutoff of study, anyone in the industry or even working on hobby products can say this conclusion is wrong.
They generate more software too. It's just organizations haven't adapted to remove the bottlenecks yet. I replaced so much software for my self and its better than all the alternatives I used to use. And there are tons of other people doing the same thing.
Along these lines: agents will write mountains of super repetitive test code, seemingly uninterested in refactoring it (without explicit instruction) to reduce redundancies.
In any non-trivial integration test scenario, it's easy to end up several times the LoC in tests compared to actual shipped code.
Second problem: different models, agents, and harnesses seem to have different tendencies and they tend to avoid touching code that is not their own unless explicitly instructed. What I mean by this: it will happily duplicate some helper method from another test file rather than pulling it out and reusing it, seemingly avoiding coupling (good kind in this case since model level changes should be propagated atomically).
Tests are where I see agents just spew absolute mountains of code.
When people have gotten excited about having so many tests, I've asked them what their plan is when 500 of their 1000 test suddenly stop passing. I've always favored 50 tests I know are meaningful, where any failure is something to respond to, over 500 tests with a bunch of overlap and edge cases I'm not sure matter. But now people aren't even thinking about that, they're just letting the agents make all those decisions.
I think this approach will continue to work for some people and orgs, but others are going to hit walls where their agents can't manage the mountains anymore. It's going to be interesting to see how people and orgs respond when that happens.
> I've asked them what their plan is when 500 of their 1000 test suddenly stop passing
That's not a terrible thing; would indicate that some shared behavior is broken and -- assuming well structured code -- hopefully easy to isolate the behavior.
Bigger problems to me are more practical: 1) large tests waste context both in and out, 2) large volume of tests waste CI cycles which are becoming the bottleneck.
Every senior I've ever worked with at some point embraced the idea of minimalism. Produce minimal changes with a minimal surface. Worked well at startups and especially well on large spaghetti legacy projects. Seeing all of a sudden now how suddenly number of kLocs, PRs, token maxxing, number of tasks in progress is again a metric with which non-engineers brag about on engineering conferences feels like the industry has reverted to some braindead pre-infant state. I guess all mistakes need to be relearned at every hype cycle.
This doesn't solve the issue with "too many tests", but one of my teams addressed this to some extent by writing "blueprints" for test scaffolding that specifically aims to surface the scenario under test.
Was initially hand written as part of our codebase before AI, just extracted it as MIT, OSS since there's no secret sauce here. But really cuts down on test verbosity and makes it much, much easier to read, review, and evaluate tests.
This headline being in the present tense, while reporting on a data from 2025, makes it basically dishonest. 2026 AI coding agents bear little resemblance to 2025 AI coding agents, and early-2026 coding agents are not like like-2026 coding agents.
Maybe but I’d argue the frontier corps will self destruct via green annd product management enough to make this impractical. So far here’s how I see that going
1) can’t easily autopilot due to security. Ganna keep on prompting you into running weird shell commands
2) it’ll launch your code on weird Anthropic or ChatGPT sites behind a paywall
3) between the two of them it’ll
Misalign the solution so it’s unusable
Way back in the day, everyone wrote assembly. Then things like C came around, which had to be compiled to assembly, and most people objected: "I don't want to be that far from what the CPU is doing!".
Each time we move up an abstraction layer, people object, saying that nothing will work right if we move up that high in the stack. Give it time. I'll bet in 5-10 years, almost no one will look at code anymore, except in special cases.
We still have people who write in assembly, we still have people who write in C, mostly for performance reasons. Because you still need to be close to the CPU to get that last bit of performance. But in most cases, it doesn't matter. It doesn't matter if the LLM makes bad code if it works and passes all the tests.
Granted, this is the first time we're going from deterministic to non-deterministic, but is it really? Large scale distributed systems are already non-deterministic. They create inputs and situations that programs aren't designed to handle all the time. What's a little more chaos? :)
This is from March 2026, but it does match what I've encountered professionally. Code review was always a significant bottleneck and is even more so with agentic coding. We've had discussions about leaning more on automated review, but it hasn't translated into any practical improvements so far. I think senior engineers find it very difficult to give up on line-by-line review, no matter how much better the agents have gotten.
I think if there is a major change in code review culture, it's not going to happen first in established engineering teams. It'll happen in startups that adopt a no-human-review policy from jump and then maintain it as they mature. I'm sure that's happening, but it's likely invisible to most studies and surveys at the moment. If it proves successful, and such companies start shipping significantly faster, the rest of the industry will be forced to adopt.
I would reframe this as: Quality was always a significant bottleneck and is even more so with agentic coding.
Code review is where many of the design & code quality problems are observed/highlighted.
> I think senior engineers find it very difficult to give up on line-by-line review, no matter how much better the agents have gotten.
Why do you feel like that? Agents have got better but better how? Looking at the frontier model output, I’d say the same myopic approach that causes long term issues without frequent course correction is still there.
I agree. The latest models are getting quite good at validation, as long as they are clear on what the requirements are. Recently I got frustrated that while we have a huge test suite I noticed the LLM was hardly ever breaking a test with its changes (PRs only added tests, never changed old ones) which seemed suspicious.
I had it do a full audit. There were about 100k lines of tests, of which it determined about 60k were near useless. Many were overly mocked, some mocked to the point that they really weren't testing anything at all. It deleted those and wrote more expansive scenario tests to replace them, caught a lot of bugs on the way and fixed those too. Total time was about 2 hours.
I think humans need to review requirements, specifications, and test scenarios. And for some industries and uses they're still going to have to review every detail. But for a typical CRUD app, I think we're at a point where we can already set the process and quality bar higher than ever, in addition to moving faster.
I chuckle a little bit at this specification thing. We used to do this. We used to write big binders of specifications, and then we'd set about implementing them. We stopped doing it because it didn't really work. We'd learn things in the process of building which negated the value of the big specs we wrote. That's why we stopped doing it. This is where continuous improvement, CI/CD, tighter release cycles, etc. were all born.
It is funny. At some point the code is the specification.
Glad to see someone else notice this. I dread looking at or dealing with AI written unit tests because of the incredible amount of whiteboxing that goes on, though usually for the opposite reason because AI testing is almost entirely whitebox testing which fails if some internal details change
Yes, unfortunately a lot of requirements and specifications are now AI slop, thus developers stop reading them, and agent produce even more slop.
We need to take back writing requirements and specifications to humans.
Adopting high level BDD style testing of human written user stories (cucumber) is probably a good idea.
Just have AI write requirements then another AI write specs based on the requirements then another AI write tests based on the specs and another AI write the implementation based on the tests then another AI write the PR based on the implementation then another AI approve the PR and another AI deploy approved PRs to production!
> I think senior engineers find it very difficult to give up on line-by-line review, no matter how much better the agents have gotten.
I actually dislike traditional code review. In part this is because most of our code review tools are very cumbersome to use (GitHub's and GitLab's web interfaces in particular).
But also because humans aren't all that great at code review when it comes to correctness and security. Part of the problem is that you don't have access to the author's mental model for non-trivial changes and have to reconstruct it, otherwise your understanding of the program logic will be inadequate.
The problem is that purely agentic code review has orthogonal problems, and typically remains subpar when it comes to judging architecture and design considerations, compared to human reviewers. Both false positives and false negatives also remain a concern.
My best results so far have been with agent-assisted code review in a proper IDE, helping you to quickly construct a mental model and to analyze the underlying logic (in addition to having agents do a first pass for correctness issues).
> But also because humans aren't all that great at code review when it comes to correctness and security. Part of the problem is that you don't have access to the author's mental model for non-trivial changes and have to reconstruct it, otherwise your understanding of the program logic will be inadequate.
Code review is not really about correctness and security. We have testing and other verification process for these. Code review is about idea discussion mostly and correcting a few stuff that may have slipped through the author's initial review.
Your PR wants to change the software in some way. My review is to comment on how you've implemented that change, like did you miss some areas where you needed to change, does your description is actually what is happening (either a cursory reading or through testing), or are you doing some unnecessary change.
IMO, the initial reading shouldn't take more than 10 minutes. More than that (so a big diff) it should have been discussed upfront, and the design agreed. If it's an important area, a more careful reading should be done about the tests and relevant stuff.
But generally, if a review takes more than 10 minutes, there's some failure of communication going on (reviewer not apt to understand the change, big diffs that should have been discussed outside of the PR, obtuse code,...)
We shouldn't give up on line-by-line review for serious applications.
My company's leadership is oscillating every few weeks on AI goals. Their most recent story is "we're expecting developer velocity to approximately double over the new 2-3 years as people get better at using AI". But of course that's completely non sequitur. People on my team spend about 10% of their time writing code and 90% doing design and analysis. So if AI doubles our code writing velocity, then it will make us 5% faster overall.
I wish they taught Amdahl's law in business school. Or at the very least the simple concept of "bottleneck".
I’m sure they learn it by other names.
The point is that while good to remember, and specifically Fred Brook’s “No Silver Bullet” article about it, there are second order effects.
If coding becomes 100x cheaper, what’s going to happen next? At least to some degree, design and planning was there since coding is expensive. (Not just coding, debugging also got squashed by AI and was an expensive part of the job)
This is why the original comment seems pretty spot on, organizations will learn how to skip more phases, and startups are probably a place where this will be happening faster. (Not clearly though, a lot of big tech are happy tokenmaxxing to try and not stay behind those startups when they come)
Brooks' point was more about accidental vs. essential complexity. You seem to be implying that design was only ever an artifact of "expensive" typing, but design exists to prevent systems from collapsing under different failure states.
If the cost of emitting syntax collapses without adequately resolving the essential complexity then you have not eliminated the design phase and you're just externalizing your failure modes onto downstream teams and systems.
"The Goal" is a key business book for a reason, but leadership seems to not get the lesson. Or care?
Which is basically what this study is showing.
Is your company striving to automate design and analysis, then?
> We shouldn't give up on line-by-line review for serious applications.
But how many changes are actually serious?
I can have an agent mass-produce security bug fixes, zip through our low-priority backlog, etc.
IMO - we need to get out of our own way.
Building is cheap now. Agents are also really good at following checklists and running tests.
So instead of having individual engineers reviewing code, we need to
* Leave human review for the most impactful changes
* Get better at defining the quality bar
* Continually codify human concerns into rules or systems that can be checked against
A lot of this was touched on a few months ago with Spec-driven development becoming a hot topic when agentic coding started.
We should continue that with spec-driven review, testing, etc.
...and they are not using LLMs to help speed up design and analysis?
I'm no booster but this is most of what I currently spend tokens on so it's maybe not as absurd as it sounds. Maybe those gains are possible and without having to sacrifice serious code reviews too.
As a senior engineer that is struggling to give up line-by-line reviews, I agree with your take.
I've spent literal decades where one of the most valuable practices has been code reviews, for many dimensions. It is one of the best ways to ensure quality production releases. It is one of the best ways to transfer skill to junior engineers. It increases tribal knowledge throughout the engineering org by exposing others to systems they may not be working in. The number of benefits to a traditional human-centric engineering team is hard to overstate.
But I'm starting to feel that instead of starting at one extreme (heavy handed reviews) and moving towards less review, it probably makes more sense to start at the other extreme (no review at all) and move towards a brand new practice tailored for the new agentic reality.
It is looking increasingly likely that we don't need to modify code review for the agentic era, we need some completely new practice that hasn't yet been invented.
It's really about hamstringing the AIs so they can't write crap. Writing code is not a problem anymore, so anyone with an idea can write a script to do what they want. But if they try to connect two ideas together, now they have a data modeling problem. This is where everything seems to fail with AI coding (largely because the original author had the idea piecemeal).
You don't give up on line-by-line review just because the human is getting better either, and senior engineers want other people to do line by line review on their own code. It's not a matter of how good or bad the person or program generating the code is.
> I think if there is a major change in code review culture, it's not going to happen first in established engineering teams. It'll happen in startups that adopt a no-human-review policy from jump and then maintain it as they mature.
I've been hearing this for more than a year. But if these startups are all moving at superhuman speed, a month instead of a year from idea to a shipping product, where are the products? Every time I ask that, people talk about the amazing velocity gains they experience for the thing they are about to ship. It never does, unless we count throwaway hobby projects at GitHub - by that metric, AI is adding a lot of software. And I guess never-downloaded slop-games on mobile app stores.
But if you actually go shopping for consumer or enterprise software that's camera-ready, the landscape is basically the same as in 2022.
I lean towards a hybrid review process where humans accept their limitations, focus what they're good on(higher level concepts and architecture of the proposed change), and let the LLMs do the rest.
Doing a line by line review as a human is silly now.
It's not code review that's slowing things down. It takes time to get the thing you want, and then you have to be reasonably sure that it's right anyway, or you risk shipping a piece of crap to production and tanking your own reputation. Even on teams that rubber stamp PRs I see people pushing drafts and waiting to merge until they feel comfortable with what they've vibed out.
> shipping significantly faster
If you're right (and you probably are), we're in for a wild ride further down the quality death spiral.
The era of user telemetry and conversion tracking already decoupled product design from customer experience. No need to wait for customer's to articulate their experience or desires or for discerning analysts to amalgamate and ponder their opinions. The numbers tell you everything you need to know! Just throw 3000 pounds of spaghetti at the wall every month and wait to see what sticks.
But now you really were bottlenecked on engineering.
Only 3000 pounds of spaghetti per month? And throwing more bodies at it barely even makes a difference? Oof.
Of course the solution is to just flush pureed spaghetti-slop through as hose as fast as you can. Brilliant!
What could wrong? The users are sure to love it. We'll know because they keep clicking. That's how they tell us the truth, even as every forum fills with louder and louder complaints!
What could possibly go wrong?
The study was updated in August 2026.
You have to classify each code review based on what it changes and decide how to handle it, agent review and/or human review. You can have an agent do that.
My experience with prototyping games is that you need very strong human driven guidance and review in the beginning, then you can increasingly take your hands off the wheel.
If you take it off too early you end up with shit and will spend weeks fixing all these baked bad assumptions your project is built on top of. I spent weeks fixing some of this shit then the LLM revived a not insignificant portion of it in a later change because it got "confused" (its words) because some of the comments it previously made were left around.
I think non-senior engineers don’t realize just how much of a burden is barely reviewed ai generated code.
I’ve seen a few projects that heavily rely on ai with not very good (though senior) devs driving it - and its a mess. Not the code. The resulting product. Code is actually sound as ai is pretty good now.
All because ai does code reviews and dump decisions go into prod
can you be more specific. If the code is not a mess, how is the resulting product a mess?
> I think senior engineers find it very difficult to give up on line-by-line review
i am pretty senior. I hate code reviews so much. They are so boring, easily the most tedious part of my job. Its such a chore. Who is clinging to line-by-line review. Ppl dont really give that much about their employers.
So you’re not writing the code, and you’re not reviewing the code…
What exactly would you say it is you do around here?
You don’t sound very senior.
Code review isn’t a quality gate, it’s a way to maintain shared understanding of a codebase.
Throwing out code review is equivalent to throwing your hands up and saying you don’t need to understand anything about what your code is doing in its black box.
I don’t think anyone on this forum is half as smart as they think they are, especially in terms of systems engineering a black box solution. Myself included. Someone needs to understand what’s going on inside the Widget Factory until we have full Artificial General Intelligence.
I initially pivoted to fully automated code reviews, sometimes with a brief skim of things. I've been shifting back to mostly manual review. In the background I'll also run some agentic scans, and once I'm done I'll look to see what the agents had to say. And then I go back in by hand armed with that knowledge and verify it & come to my own conclusion.
Why? Because I started to run into way too many AI generated messes in codebases. The agents are really good at some things and complete ass at others.
Absolutely. And also the idea that this is somehow the last bastion against introducing bugs? Maybe run a test
> Who is clinging to line-by-line review.
It really depends on the context. A react button background color - meh - the critical part of the backend where a new 300ms delay will cost the company millions then yeah.
> the critical part of the backend where a new 300ms delay will cost the company millions then yeah.
If it is so critical, why don't we have tests and benchmarks for it?
Oh right, because "Greg the super-coder" never wrote any of this down and it lives in his head so he can be the arbiter of change.
It was toxic before agentic coding, it's still toxic now.
Document your systems. Add tests and benchmarks for what is important.
We've always known this, we have a dozen or more textbooks written about it. But we always gave it a pass because "Greg the super-coder" was producing more value than he was hurting.
That's no longer the case. Your best coder is now Opus or Astra
Good thing you reviewed the PR so you could catch the gap in Greg’s documentation and force him to write it before approving his PR!
> he critical part of the backend where a new 300ms delay will cost the company millions then yeah.
If a performance regression like this isn't caught by an automated testing suite, you have made a horrible mistake.
> new 300ms delay will cost the company millions
these might be 1% typically
I hate code reviews, but I owe it to my coworkers that I know will read my PRs to review my own generated slop line by line. Most of these agents are still overly verbose and I'm in a situation where no, not every single edge case is relevant enough to warrant the additional loc or complexity introduced.
[dead]
As is standard for academic papers, this is an attempt to describe the state of the world as of when they did the research. But then it takes time to write it up and get it out, after which newspapers report it as current events.
In this case, we're looking at a picture from 2025. The models involved in the main agent results were primarily Claude 3.5 through Sonnet 4.5, GPT-4.1 through GPT-5, and Gemini 2.5. And the paper's design means you're mostly looking at the very initial effects of deployment, before companies developed good review processes.
The study was updated twice with the last time on August 2026.
They've updated it for the agentic code era, and have found basically the same thing. Code can be created faster but that has moved the bottleneck to review.
Man, I'm probably going to expose my own inexperience here, but like, did anybody actually expect code reviews to catch and prevent bugs on a line by line basis? Like ever?
I never have operated this way. Code reviews were a basic sanity check, a chance to implement best practices, and a chance to teach these things to juniors.
If I was concerned about something breaking, I'd duscuss test coverage.
Was I doing it wrong this whole time??
I’ve caught bugs in code reviews before, but I would agree that it’s probably not the most crucial thing you should strive to do with code reviews.
That said, I think I’ve done a tremendous amount of PREVENTING bugs from being introduced via code review. Caught all kinds of weird code patterns, duplication, wrong levels of abstraction, etc. Essentially, the kinds of places that bugs like to hide and grow in.
I still think that’s valuable, and it’s the kind of thing that models generally aren’t great at, to this day. It generally relies on having a large amount of context about the project and about what’s being solved, so if you wanted to do this via automated review, you maybe COULD, but it would probably get pretty expensive, especially on larger projects.
It was more important for giant C or fortran code bases. You will legit catch pointer errors or memory mgmt stuff, going line by line. Once Sun busted in with Java is when I really saw the shift to skim reviews. So much stuff is caught by compilers now, or SAST, that using dev time to find code smells is mostly a waste of their time. And it's not like the old way caught everything, it's just that the stuff we take for granted being caught by compilers or Sonarqube now, just used to get shipped to prod and cause abends and production support calls at 3am.
I read this paper several days ago, when my director shared it. Several things jumped out at me:
- the data is mostly from 2025 and earlier
- the baseline mean lines of code per month was 5017, the estimated increase was 1494 lines per worker-month
- the baseline is 5.34 PRs per month, and increased by 1.22
- the baseline mean for code review is 7.03 days
The second and third bullets are confirmation that this is old data. In late 2026, people aren't delivering 30% more lines of code. If people were just producing 30% more lines, they'd still be reading every line. My point isn't that "more lines is good". I'm worried about a few too many lines of code that no one reads. My point is that this shows the data is old.
For the latter, I'm a bit annoyed by using mean, median would be more valuable, but I think it's safe to say that the engineering culture was already pretty bad. If your average pull request sits for 7 days, you evidently don't care about shipping anything, and agents aren't going to change that.
You read the paper, saw that the stated collection dates ranged from January 2021 to March 2026, and are claiming the authors are lying?
If you’d asked reasonably, I might have noted the slight correction (I misremembered that data collection ended January 2026, not March).
But since you’re such a charmer, why did you read the word “mostly” then lie about me lying?
It’s unambiguously true that most of the data (including that on coding agents) is from 2025 and earlier.
I'm noticing AI catching many bugs in code that previously went unnoticed. Some of them are important, and it's great to catch them early.
The other 11 items in the list of 12 would probably never be hit in practice, but now that you know, you have to sort out how to handle each (even if that just means documenting, filing, or ignoring: you have to understand each of the disparate 12 issues first.)
Before it was all coding on its own, I asked a model for bugs in a new project, saved me the 15 minutes of new web project I waste every time trying to remember obscure first time setup things you dont touch for years in under a minute.
I suspect theres a lot of hours saved that are otherwise wasted by devs when some simple thing throws off their implementation completely. Basically we cant measure the saved developer hours by LLMs because it could be literally anything.
I’d strongly disagree with one aspect. Homemade personal use software has Definately skyrocketed. Projects I never thought I’d even have the time for are now possible.
Yes, I am literally inundated by cool new tools I've built that I never would have before.
I'm literally building a black-box client + server emulator for one of my favorite old MMOs, remotely, from a coffe shop in SF as I type this, on my laptop at home via Claude remote-control.
Insane.
Organizations that use Jira have not reaped efficiency gains from AI, news at 11
Any study that wants to say anything about the impact of coding agents on whether more software is being created or not can't limit itself to what firms are doing that create a footprint of "work events — including GitHub coding activity, Jira issues, and Google Calendar events".
This is software that would not existed without coding agents: https://fluximp.org
Sure its a coding agent itself, but I would never have tried to create it manually as I would have expected it to be too much work to see it through as solo dev. There were no issues created anywhere, no calendar events, and there is very little github activity (I primarily used Mercurial on a private server; git is just for distribution).
I started it with Kilo Code and after a couple of months it was purely self-modified and has been since then. I was a bit worried that it wouldn't work on Linux and I'd either have to use another agent or use it running on my Windows box to modify the code on my Linux box to get it working, but it just worked so the streak of being self-modified has continue.
I want to make an indie game and I'm working towards that with the project I'm using to benchmark FluxImp. My first job was as game dev and I took a shot at indie game dev 20 years ago (failed; not fun). Its something I had long written off as too time consuming to even start, but with coding agents I no longer feel that's true.
All of the stories about non-software engineers making small software that is useful to them also seems unlikely to generate the kind of footprint that lends itself to being easily studied.
I don't see how AI coding agents are going to create 'more software' until the PO's and PM's of the world step back and let developer's communicate directly with the stakeholders. Or the PO's and PM's could cross-train on AI coding agents.
We are still being stressed to create all stories with such fidelity that 'any' developer can pick it up and implement it in the timeframe allowed for the size the story was given. When the story is written in that detail, we may as well copy it into an AI agent and have it implement it. But we can't do that, so we have people who have SME write up story's that sit in the backlog for 2 weeks until a developer can bring it in, so that they can spend a day implementing it (really 2 hours with 2 hours of meetings and 4 hours of video games) because the smallest pointing size we can give is 1 point and people aren't allowed to take on more than 8 points a sprint.
How can we expect things to get done faster when the team that reports work up to senior management put's such strict limitations on our throughput?
Listen man. Let me give you a tip.
Don't take it seriously.
Efficiency is not your job, until you can 1) solve it, 2) present the solved and implemented problem to senior leadership in one fell swoop.
Corporate America is a giant slush fund and this is how it keeps going.
Be grateful!
I would be more concerned with playing video games half the day than whether the PMs are slowing you down by figuring out the requirements
There are many ways to say the same thing, that code throughput wasn't the only bottleneck, communication is by far the hardest, most brittle and time consuming part of software production.
We need to start becoming better writers, capable of expressing complex thoughts concisely and with technical precision. Software production has become a sibling field to linguistics, more than it already was.
I've spent most of my time working on tools to keep my agents on track, and keep code legible. Where I'm probably most strict in the tooling is on documentation, which makes the code review much more manageable.
I'm completely uninterested in the opinion of a video game journalist on what does and doesn't work well in writing software. The author knows about as much about writing code as he knows about performing heart surgery
Ars Technica in general has also been catering to their knee-jerk anti-AI audience. I used to find their comments section somewhat interesting but now it's the same old tired, outdated dunks on AI.
Yes, instead of reviewing 1,000 lines a day I now "review" 10,000 lines a day. My brain is not 10x faster, so either I spend 10x more time reviewing code, or I just opt to accept that AI is probably right based on the complexity of the problem. However I realize that if we just let the agents loose on the codebase, everything would speed up like 100x.
Invariably once certain AI models in certain domains reach "full automation" quality this will happen, until now the human eyeball bottleneck is holding a leash on millions of tireless superhuman coders who have just a few narrow but egregious contextual awareness issues.
Given what I've been able to accomplish with even just local models at home, and my experience at my day job, I imagine this has almost everything to do with the organizations themselves and less to do with using AI to build software.
My experience is that agents are decent in coding a feature, solving a bug - doing specific well defined task. They are very bad at judging UI/UX and user flows through app. Like really, really bad. They can quickly put together vibe coded app but make it usable takes a lot of extra manual work.
Did anyone solve this problem? Any good approaches?
I confirm this diagnosis. The UIs are always clunky, disorganized, and kinda ugly. But they work well enough. I guess UIs are opinionated enough that we probably all customize them anyway. Maybe the agents know this so don't bother putting too much effort in? (Deep thought of the day: acquired)
Wrong thing to index on. Always bet on the slope, not on points in the slope. Things will only get better and better and factor in that downward slope thought while taking actions and pick ones that are favorable. The folks who focus on a point in time will be left behind, period.
LLMs generate more code yes.
I generate more beautiful code, yes.
But the longer the codebase survives the more hair and cruft it accumulates. My once beautiful logical paths get carved up by compounding requirements and edge cases.
That's what the AI was trained on. The end result - the hairy ball of wax that noone can look at sideways.
I often think of that whenever AI comes up with what I consider to be an irrational but plausible edge case.
Aside from other comments mentioning flaws and cutoff of study, anyone in the industry or even working on hobby products can say this conclusion is wrong.
They generate more software too. It's just organizations haven't adapted to remove the bottlenecks yet. I replaced so much software for my self and its better than all the alternatives I used to use. And there are tons of other people doing the same thing.
How come? For example, "Brandon Thomas released Artcraft, a free, open-source suite of apps ... alternatives to the entire Adobe Creative Suite".
I checked and it looks and works great!
That's pretty sick.
Maybe for some orgs, I know theres orgs out there where there is no such review only QA and unit testing.
We got rid of end to end QA as head of engineering decided to "shift left" on automation.
The paper is from January, so the data is probably from 2025. Very unlikely anything in the paper is relevant now (if it ever was).
The paper was updated in August this year.
Along these lines: agents will write mountains of super repetitive test code, seemingly uninterested in refactoring it (without explicit instruction) to reduce redundancies.
In any non-trivial integration test scenario, it's easy to end up several times the LoC in tests compared to actual shipped code.
Second problem: different models, agents, and harnesses seem to have different tendencies and they tend to avoid touching code that is not their own unless explicitly instructed. What I mean by this: it will happily duplicate some helper method from another test file rather than pulling it out and reusing it, seemingly avoiding coupling (good kind in this case since model level changes should be propagated atomically).
Tests are where I see agents just spew absolute mountains of code.
When people have gotten excited about having so many tests, I've asked them what their plan is when 500 of their 1000 test suddenly stop passing. I've always favored 50 tests I know are meaningful, where any failure is something to respond to, over 500 tests with a bunch of overlap and edge cases I'm not sure matter. But now people aren't even thinking about that, they're just letting the agents make all those decisions.
I think this approach will continue to work for some people and orgs, but others are going to hit walls where their agents can't manage the mountains anymore. It's going to be interesting to see how people and orgs respond when that happens.
Bigger problems to me are more practical: 1) large tests waste context both in and out, 2) large volume of tests waste CI cycles which are becoming the bottleneck.
Back up to the point where all the tests were passing, of course? Then debug.
or git bisect until you get the exact commit that introduced it
Every senior I've ever worked with at some point embraced the idea of minimalism. Produce minimal changes with a minimal surface. Worked well at startups and especially well on large spaghetti legacy projects. Seeing all of a sudden now how suddenly number of kLocs, PRs, token maxxing, number of tasks in progress is again a metric with which non-engineers brag about on engineering conferences feels like the industry has reverted to some braindead pre-infant state. I guess all mistakes need to be relearned at every hype cycle.
This doesn't solve the issue with "too many tests", but one of my teams addressed this to some extent by writing "blueprints" for test scaffolding that specifically aims to surface the scenario under test.
https://github.com/zeeq-ai/zeeq-biome
Was initially hand written as part of our codebase before AI, just extracted it as MIT, OSS since there's no secret sauce here. But really cuts down on test verbosity and makes it much, much easier to read, review, and evaluate tests.
This headline being in the present tense, while reporting on a data from 2025, makes it basically dishonest. 2026 AI coding agents bear little resemblance to 2025 AI coding agents, and early-2026 coding agents are not like like-2026 coding agents.
I take both approaches. just /yolo or /goal or measure twice and cut once.
yolo works when I need to focus on higher priority, and make sure things just work (though not "great").
For the latter, I use it for more like spec-driven fashion with HITL. Basically everytime it ffinishes a task, I review and then move on.
Both appraoches work. just make sure your yolo slop isn't bad and passes stuff.
LoC don't mean jack shit.
Maybe but I’d argue the frontier corps will self destruct via green annd product management enough to make this impractical. So far here’s how I see that going 1) can’t easily autopilot due to security. Ganna keep on prompting you into running weird shell commands 2) it’ll launch your code on weird Anthropic or ChatGPT sites behind a paywall 3) between the two of them it’ll Misalign the solution so it’s unusable
[flagged]
[dead]
Way back in the day, everyone wrote assembly. Then things like C came around, which had to be compiled to assembly, and most people objected: "I don't want to be that far from what the CPU is doing!".
Each time we move up an abstraction layer, people object, saying that nothing will work right if we move up that high in the stack. Give it time. I'll bet in 5-10 years, almost no one will look at code anymore, except in special cases.
We still have people who write in assembly, we still have people who write in C, mostly for performance reasons. Because you still need to be close to the CPU to get that last bit of performance. But in most cases, it doesn't matter. It doesn't matter if the LLM makes bad code if it works and passes all the tests.
Granted, this is the first time we're going from deterministic to non-deterministic, but is it really? Large scale distributed systems are already non-deterministic. They create inputs and situations that programs aren't designed to handle all the time. What's a little more chaos? :)