Maybe I missed it, but it look like this has just a single metric. Maybe instead of making a new project, you could try to get this metric added to a existing tool like https://dekobon.github.io/big-code-analysis/index.html which already has dozens of metrics.
Yes, I'm doing my own research on AI augmented pipelines https://blog.officefloor.net . I actually found most code quality tools look for bugs and complexity, but nothing much about cohesive erosion. The nice thing about this metric, is that it determine the files where the erosion is occurring. I turned it into a GitHub action to make it easier to access to get wider feedback on the metric. The GitHub action triggers on your merge request and tells you the files where erosion is occurring to refactor. This stops erosion before it gets too expensive to change (big refactors or rewrite). Yes, happy to work with others to get the metric into other tools.
I just dump all user messages from all sessions in a project into a flat .md file and have agents synthesize the user's intent. Then, using that extracted intent, the agents review code and tests. I call this a retro/reflection pass. It checks whether the code matches the intent and whether the tests match the code.
Compactly formatted user messages are something an agent can ingest in a few minutes, even if they are thousands of lines long. And the quality of those messages is great: they don't track what the agent does well, only what changes and what breaks.
Having this top-down view helps a lot. Usually, within a session and deep into a task, the agent loses the global perspective and optimizes for local success. I find it weird there is no harness that treats user messages as high value signal (except my own, of course, I have it, https://github.com/horiacristescu/playbook-harness).
The problem with specification and user discussion is they still have errors that code has. But unlike code, there are no tests to confirm correctness.
So now we have a definition of the system in a non-exact language with no ability to test to confirm it's correctness. The code holds the essential complexity and now we are adding accidental complexity on top to manage.
Again agree the specifications and user discussion provides context for the AI. However, a well written test suite provides similar context that can actually confirm correctness of the system.
However, saying all the above. Focus of ImpactGate ( https://impactgate.officefloor.net ) is about erosion of the code, not correctness.
Yes, agree. It's a learning process. I tend to find when I build systems that at some point you need to stop analysis and just start building things to explore the problem. As you do, you prototype, refactor and possibly throw out ideas in favour of understanding the problem and discovery the real solution.
The code becomes a reflection of that.
I'm interested in your experiences of capturing specifications and user discussion on whether this captures the end intentions? Or whether it keeps you focused on earlier dead end directions?
I think agents are pretty capable of reading a log when given the explicit task of extracting the latest version of what the user wants. In general, they work well for direct tasks like this. They don't forget and do something else the way they do when they are deep into development work or debugging.
Besides intent, I also mine signs of "user friction," which I use as input for the agent to come up with new tests. What I complain about is one of the signals driving testing.
I've never encountered that term before (cohesive erosion) but I like it, if I'm interpreting it correctly.
Do you mean like the hyper focus an LLM puts on the task in front of it so you end up with drift (duplicated concepts/multiple ways of doing things, terminology drift (e.g., now we have "customer" and "client"). That sort of thing?
When you think about a god class or god method, it occurs over time by adding more than a single responsibility.
Yes, there are generally complex algorithms but they usually are not things developers write (imported from libraries).
What is usually going on in the god class/method is that things keep getting added to it. These things should be separated out. So the cohesiveness of the class/method erodes into doing too many things.
The idea of the Change Impact formula is to catch this early so you start refactoring to separate out into classes with single cohesive purposes.
The problem with AI is it handles complexity really well and will happily keep piling changes into god classes/methods reaching ridiculous CC levels (have see over 200). Previously developers would get annoyed and do the refactor. But with AI these days, changes are happening faster. So Change Impact is to try to monitor the cohesive erosion.
Comes from the basic Computer Science principals of High Cohesion and Low Coupling.
High cohesion means the functionality of a component are closely related and focused on performing a single well defined task. Basically single classes for single purposes.
Erosion of this is when classes start doing to many things, in the case of God classes.
The Change Impact formula looks at a way of detecting when the cohesion is eroding and flagging it on a change (as the pull/merge request itself should generally be single focus cohesive change)
That project in itself looks very interesting. How are people using it, any examples of how people get this into an actual report / CI test / benchmark / whatever ?
Code metrics in general aren’t that widely used. I’ve only ever worked at one place (a bank) that tracked it, and that was only because sonarcube had it built in.
While a lot of metrics make intuitive sense, we don’t have that much hard evidence to prove or disprove their value. Part of it is the whole “if a metric becomes a target, it ceases to be a good metric” thing. Adding the checks to a large existing project probably has negative value. But I think it’s worth doing for greenfield projects.
For humans, these should just be advisory. But for LLMs I’m happy enough to make it a blocking check.
I keep thinking of doing an experiment where I give the same LLM the same problem, and only change which metric is enforced. And then see if any of them have a noticeable effect on correctness/maintainability.
> any examples of how people get this into an actual report / CI test / benchmark / whatever ?
Yeah they have examples of adding it to CI, or local checks, generate html reports, etc in their docs.
I think the future of development will be a lot of automated quality checks like these on AI-drafted code that humans review and ensures that it doesn't muck up the business logic and actually fulfills the acceptance criteria. that said, I don't think existing toolsets are really great at actually measuring code quality
I've been in the process of reviewing and validating a lot of tools like this (qlty, Sonarqube, fallow, etc) and the false positive rate is anywhere from 20% to 80% for a lot of our sniff tests (zizmor produces an overwhelming majority of false positives here for what feels like arbitrary and very context-dependent GHA requirements)
the last thing I want to do is to annoy the hell out of our devs by requiring checks like these to pass especially since it's only a small percentage of them who vibe code everything and then also vibe response to code reviews. I feel like that's the anti-pattern that we'd push people towards by requiring checks like these to pass
another avenue of exploration has been requiring test coverage but also good test quality metrics (eg are there negative tests? mutation testing? empty asserts?) something that seems quite easy to spin up into a skill and pair with a deterministic harness. trash-tests is a neat little project that incorporates some of this: https://github.com/frangelbarrera/trash-tests (disclosure: I am not the repo owner or even a contributor, just a quality nerd who loves underdogs lol)
all in all, it really does feel like we'll need a revamp of the SDLC with our current expected velocities
Yep, agree on annoying developers. So just to cover usability it can run in warn and block mode to address this.
Though to the bigger point of your comment, yes SDLC are becoming faster. We can churn out code at a ridiculous rate. However, doesn't mean it's good code. And hence, there are studies showing things actually slowing down because reviews pile up.
I guess I look at the Change Impact formula (and https://impactgate.officefloor.net implementation of it) as threshold tool. Small changes that aren't contributing to god classes, just let through. When things start to smell, the files involved get marked for review.
Ideally this then can cut down on review time and allow overall increased velocity.
But yes relies on trusting the AI to do "simple" things
Yep, this all actually started because of experimenting with my own open source project https://officefloor.net (giving full disclosure)
I was testing the additive pipeline style of OfficeFloor against the mutative handler style of Spring. I was looking to see what factors could be used to allow AI to make long on going changes (experiment is 60 changes to an end point, where all add functionality and every 4th change is mutative on existing rules). Then I watch how AI manages to make the 60 changes in each architecture.
I've done many runs and you are quite right about Goodhart effect in giving it the metric. Never knew Spring code could be written so badly.
I've tried runs with better prompting also and I'm starting to find the key factor is actually the architecture itself.
From my initial findings, it's seeming that additive pipeline architectures hold up much better against AI slop than our typically single method web handler architectures.
I wonder why they didn’t throw a real chess engine in there for a baseline. There are engines where you can set the elo in the settings, so it should possible to see these LLMs relative to a human 1500 rather than just relative to each other.
This is true, but I'm not sure it matters? I was poking around at the lichess database recently and those elo calibrated bots are
remarkably well calibrated, their rating variance sticks out like a sore thumb compared to human players even at similar game volumes. So it should still be a decent predictor of how good a human at that level is, even if the playstyle seems alien.
This is simply blatant misinformation. If you play a game online on lichess and go to the analysis board you can find when your game becomes novel. It will be within 20 turns unless you are intentionally following a known opening. In fact it will likely become unique within 10-15 turns.
It's not my experience at all. If you find yourself in a novel position within 10-15 moves it's likely a resignable one.
Edit: maybe you don't understand what i mean by novel position. I mean any position that has never been reached in the billions of lichess games, including bullet games among beginners.
Also yes I will concede that it's possible to make "quiet" moves. pawn nudges that barely affect anything. If you're doing those you're not 1500 ELO. You're intentionally trying to throw wrenches and I just don't see why an ELO bot even needs to bother with nonsense like that. This is supposed to be for fun / training!
I think your prediction is a bit early. Maybe a decade early. 2033 is 7 years away. The transition is happening really fast, faster than most people (or political leaders) know, but not that fast.
We will hit 1TW per year of new solar soon, but to get to 100% electricity by the end of 2033 I think we would need closer to 3TW per year.
How does parallelism come into play for this conversation? Async/await is a concurrency construct. And Java’s virtual threads are also a concurrency construct. Neither of them have anything to do with parallelism. Or am I misunderstanding something?
I'd say Java's virtual threads are also a parallelism construct (at least in the performance sense, not logical guarantees), since they're scheduled on a pool.
I was bit disappointed how few metrics the article mentioned. There are tons of code quality metrics that have been thought of over the last 40 years. We don’t have a good idea which ones are worth enforcing though. And we don’t know if the metrics that are good for humans are also good for LLMs
> You and the other poster just need to find some agreeable benchmark so you can decide who won.
That is the hard part. I'm not sure the usage numbers we'd need to decide the outcome are public now, and I can't predict ~4yr out whether any currently available metric will continue to be available. Stack Overflow's popular language thing has been going for a while, that will probably still exist (if SO does) but do we have any reliable metrics to tell us what fraction of code is AI generated?
I think the outcome will be very obvious when the the time comes to settle the bet, but I agree that it's not easy to write those terms today.
But speaking of Stack Overflow, that's a good point to examine more closely. Simply looking at Stack Overflow's usage trend [1] should have been a strong clue that a seismic shift was under way. A couple of things could be responsible for it, though:
1) Stack Overflow's movers and shakers have finally assholed themselves into irrelevance. That's possible, and it could potentially explain the secular decline that began around 2015. But it doesn't explain the rate of descent since 2023, given that the core rules haven't changed for years. It also doesn't explain the meteoric rise between 2008-2014. What they were doing clearly worked, right up until it didn't.
2) AI is now answering questions that would previously have been posted to SO. That's the conventional argument. Hard to dispute it. And it leads to...
3) Stack Overflow has not failed its mission, but fulfilled it: most of the code that will ever need to be written already has been. That's an argument I've never seen anyone else propose, likely because it's a really stupid argument that's been made many times before. I think it's true this time, though, at least until genuinely-new hardware paradigms come along. I think it's a big reason why AI will be doing our jobs for us going forward. Programming is a robot's job now.
3a) As a corollary not related to the SO question, I think the math community is starting to wake up to a similar realization: we have all the math we need. Or, rather, we have all the math we can comprehend. The low-hanging fruit is all gone. Mochizuki's work, which requires a large part of multiple peoples' careers to prove or refute, is an example of that phenomenon. Leading mathematicians are coming to recognize that their best shot at contributing to progress in math is to work on AI.
The groundwork has been laid, and now we need to find better ways to reuse and recycle what we have. That's how AI will help us reach the next level. It's just crazy to think that our industry will be recognizable in 3-5 more years.
As a joke. Back when 2 came out, they promised there would never be a version 3. I.e., no breaking changes. Then they realized they did need to make a breaking change, so the only way to be true to their promise was to skip version 3.
No, you misunderstand what is going on. htmx is an exercising in learning web development by someone who didn’t follow 2 decades of web development progress.
He is catching up though, now approaching the early 2010s jQuery (moxi) and backbonejs era with fixiproject.org
Htmx is a direct descendant of Intercooler.js, which has been around since circa 2013, more than a decade. Intercooler still exists and is used in production, and you can clearly see that it works almost exactly the same way as htmx does today, it just bundles jQuery together.
The creator has been working on these ideas for a long time, but the core idea–swapping HTML from the server into the DOM–has been constant throughout.
It's exactly because he followed the past two decades of web dev that htmx avoids almost everything about it. Fixi and htmx extensions follow the 80/20 principle and can work by just being dropped in with a script tag. You don't need an elaborate npm setup, same as everything else in the htmx ecosystem.
The creator is stuck in 2012 and so on as I said. If you look at project fixie, the reason htmx 4 exists, it shows how the 80/20 is an incomplete idea for something as general as a web framework. 80/20 might work for your specific use case, but as a general notion, it either gets built on reasonable constructs or grows into a Frankenstein. Htmx approach is a mirror of yaml in my ways. In denial of what it is trying to be and so turns into complete monster.
Correction: they didn’t ‘realize they needed to make a breaking change’, they got excited by the improvements they could unlock by using the Fetch API and wanted to make a breaking change.
The upgrade is completely voluntary: it won’t even be set as the default version in npm till next year, and there are no known security issues that would force anyone to upgrade. People happy with v2 can just stay on it for the foreseeable future.
I wish they would have done what Taalas did with chatjimmy.ai and just directly host a model for us to view, rather than just claiming it’s 50x faster than Nvidia/groq. Their claim is specifically for a 1 trillion param model. So they could have just grabbed GLM 5.2, or similar, and hosted it.
Joke’s on us, all of their pages are LLM pages! LLM generated, that is.
Btw, can guarantee that they are not ready to demonstrate that yet. They’re using 2D FLASH with 30M weights per die [1], so to get to 1T they will need… 33,333 dies. Interesting scaling problem to say the least
But they also declare having a "Mead" technology that stores at least 175b NNs in a single chip through 3D stacking - see https://www.mythic.ai/mead and other posts in this page.
A confusing thing is that the goal is tackled through a number of proposals... Why Vanguard if they have Mead? If Mead, how to get the memory integration that are explicit on Vanguard?
100x seems like an underestimate. Even with no model improvements, we should see that sort of reduction. Looking at TSMC’s margins, Nvidia’s margins, and OAI/Anth (alleged) margins on inference, there is a room for a 100x reduction.
Right now all three of those are at abnormally high levels. Competition will come for all three.
reply