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).
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?
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.
BTW, you can avoid your comments being flagged and killed by writing them yourself! I know it’s tempting to offshore it to AI (especially after you’ve vibe-coded a whole project) but some genuine human communication goes a long way.
The only thing is that these issues seem like human code problems and IME LLMs don’t really write code like this anymore. It’s almost the opposite in python, actually, where Claude leans on writing lots of 2-3 liner private utils which is a separate kind of complexity and organization problem.
I still find it useful specifically for React where it’s frustratingly normalized to write many branches in your JSX though.
There is an implication that all AI changes add complexity and reduce quality, but it is obvious that the quality and complexity is a property of the code not the writer, so all equivalent changes should be equal no matter what the shource.
How do clear improvements to make something simpler yet more functional score under this system?
The problem with cohesion is understanding what is "single purpose". Nothing can determine this without interpreting the code and making a judgement call on whether it needs to be refactored into two or more classes.
So I'm looking to approximate this. Instead of looking for cohesion, Change Impact formula approximates this by looking at existing complexity in the file.
If the change adds more complex functions to an already complex class, it scores high. Smell of it doing too many things.
If the change adds a complex function to an empty file, it's likely just a complex single problem.
If the change adds simple change to complex file, then if this happens too many times, we flag it for concern. Especially if it changes a lot of other files too.
So it's not mean to measure your code exactly. It's meant to highlight smell that needs investigation. And the formula points out the classes that need looking at.
Plus I'm not sure that all changes are equivalent. I'm not looking at single one off changes. I'm looking at 60 changes in a row. The first 20 might go through fine. However, after the changes pile on, there needs to be some refactoring to keep the code clean. This is trying to catch it before it becomes a mess.
How do you determine that this is correct = adding complexity to a single file vs adding less complexity to multiple files that this one file then orchestrates.
The novelty of Change Impact is it looks at the context of the complexity. Adding complexity to something already complex should be flagged for review.
I've not come across any static code analysis so far that has provided this.
And in my own research of architectures, I came to this formula.
Claude is pretty good at adding focused changes and if a fix is already present, then it correctly points it out rather than adding unnecessary refactors.
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.
Instead of focusing solely on the complexity of the thing you change. Include the context around what you are changing also.
This allows the Change Impact formula to differentiate between two different files of the same length. One is a complex god class with high CC. The other is an entity module that is simple but just has a lot of methods due to properties. Adding to the god class really should warrant a refactor. Adding to the entity class is just another field to store for the entity. The Change Impact formula can differentiate and point to the god classes.
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)
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
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
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.
However, I'd bring in Brooks discussion on essential and accidental complexity. In other words, there being No Silver Bullet https://www.cs.unc.edu/techreports/86-020.pdf
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.
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?
It's looking at the nature of on going changes.
A single complex file written once and never touched is likely a good cohesive file about a single problem solved.
However, if we start making many changes to that file increasing it's complexity, it's likely not the problem getting more complex. It's likely we're adding more purposes to the file, eroding it's cohesion. This is the signal I'm trying to catch to avoid a mess starting.
The research I'm actually doing is comparing:
- mutative architecture (Spring) which as single method handling the request. It grows in complexity as changes are added
- additive architecture (OfficeFloor) which does function injection. A pipeline of functions, where new functionality is an injected function as a step in the pipeline (i.e. adding functions, rather than mutating a single handler function)
What I've found is the complexity of the overall system is what it is. Between both architectures, the overall CC is the same. However, the distribution is very different. Mutative architectures pile changes into god handler method. Additive architectures spread the complexity into smaller manageable files.
This, however, does not come with perfect wins. Spread also means AI doesn't read the dispersed files. AI is good with complexity and tends to fix things when it sees them (so does fix bugs in god classes). But this is only to a certain point when the complexity then comes to overwhelm the AI.
The formula is trying to catch the mess of on going changes at a point before the expensive refactoring or rewriting becomes necessary.