And so agent-written `md` updates are showing up in PRs. So far it seems useful (our main repo is a 350k LOC TypeScript monolith).
Admittedly, this is way less sophisticated (& less complicated) than the "graph of edges in/out of every bounded context" in the OP, but that is probably again my "little-d" DDD perference, where I find some of "DDD at scale" patterns lead to, imo, over-engineering.
Why not put the content from the md file in the code as documentation?
Ideally, the code can actually help you structure that information. I've written a bunch of Haskell and OCaml like this, where the types in each module let me structure my documentation in a way that is actually easier for people—and maybe also LLMs—to track. As a bonus, it makes it more natural to keep the two in sync.
Technically the md file can have more free-form docs, but for the overview/fields headers that, by convention, map to the entity/field jsdocs, we do sync them over.
Something i read in the earlier paragraphs about llms being easier to work with in greenfield projects…
My experience has been the opposite. They work well on existing projects but are not so great at new ones (unless you are just vibe coding something simple).
I think it has something to do with it being able to rely on years of established structure/conventions on existing projects that makes them better IME
Agreed! I hate the start of a project with an LLM. The code it writes is always expletive, and wrangling it with rules, hooks, plans, and what have you is just a series of half measures. I'm basically in a fistfight with a cheerful drunk keyboard.
Once there's enough code, tests, and docs to appease the prefill gods and give it an understanding of what I expect in terms of code quality, style, and TDD etiquette, I can expect more consistent performance.
That said, I've never been comfortable enough with it to let it write more than a few of functions at a time without a diff/checking-in.
I've seen folks with 500K LOC branches written by an LLM and my heart just sinks...
If you start with architecture docs that outline the high level goals and what the core abstractions should be for a greenfield project and refine it as you fill in the actual code, you can get pretty good results.
That gives the first few agents the "north star" and then subsequent agents have the actual code, the runtime behavior, and the user experience expectations to work against after that, in addition to the architecture docs, that provide hard boundaries to vibe off of.
My experience with greenfield is that when you're just starting out, you have to be really strict and careful about overall architecture, establish a framework, provide structure, separate concerns well.
If the project is not throwaway, this first set of iterations is critical, without guidance LLMs will just bolt on things and produce special-cased spaghetti, even the very-frontier models.
Once the framework is established, it's smooth sailing.
But I haven't gotten to "very large" size yet, I expect it produces another kind of coherence problems.
I’ve had a good experience with a greenfield project.
The single thing that seems to have helped is that we all agreed to use OpenSpec early on, and to commit the specs alongside the code.
I have no affiliation with OpenSpec and I don’t suspect it’s doing anything unique here, but having the intent develop alongside the code in the repository seems to have ensured that agents have a more holistic view of the project.
It’s a night/day difference when I use an agent against this codebase that integrates its changes using OpenSpec and those that ignore it.
Depends on what the existing project looks like ;-)
One repo at my work is DDD-esque, clean architecture, modular monolith.
It suuuucks to work on. Even without AI, editing any feature requires navigating through a bunch of app, infra, domain layers instead of living in a vertical feature slice. Indirection from cargo cult design patterns we don't need, like Mediator. With AI, degradation in quality and accuracy is plainly obvious. Everything needs hand-holding. AI is dumber than our average human dev so the creaky architecture made things 10x worse
The greenfield one we started with AI a few months ago is kicking along nicely. Warts exist, there are some really large files and questionable bits, but it's way easier to work on because we didn't make the architecture some convoluted dwemer mine thingy
i'm sorry to say that besides like the first paragraph, this article is mostly ai slop. i thought that we mostly weren't allowing ai written content on hn? (@dang whats the policy here?)
aside from that, i think that this is overcomplicating a more basic idea. You should have agents that leave docs behind in your codebase, and those docs should be localized to the structure of your code, but you do not need any fancy format or manifest.
We've had a lot of success with having the agents write and maintain a docs.md file in each folder, and requiring that they both read and update that file whenever they make a change in that folder
When asked about DDD at a job interview I joked it was Debug Driven Development, when you start with empty directory and file the first bug "The app doesn't do anything. It should ..."
Non-ironically that's pretty much the way I work with AI agents now.
I've also been a little-d DDD fan, and we've had luck with per-entity `md` files to language-independent document domain behavior/quirks/usages.
I.e. an `Author.ts` has an `Author.md`, `Book.ts` has an `Book.md`.
For agents, we've given them a skill to read & write the `md` files:
https://github.com/joist-orm/joist-orm/blob/main/packages/co...
And so agent-written `md` updates are showing up in PRs. So far it seems useful (our main repo is a 350k LOC TypeScript monolith).
Admittedly, this is way less sophisticated (& less complicated) than the "graph of edges in/out of every bounded context" in the OP, but that is probably again my "little-d" DDD perference, where I find some of "DDD at scale" patterns lead to, imo, over-engineering.
Why not put the content from the md file in the code as documentation?
Ideally, the code can actually help you structure that information. I've written a bunch of Haskell and OCaml like this, where the types in each module let me structure my documentation in a way that is actually easier for people—and maybe also LLMs—to track. As a bonus, it makes it more natural to keep the two in sync.
> why not put the content from the md file in the code
We do that :-)
https://joist-orm.io/modeling/documentation/#two-way-syncing
Technically the md file can have more free-form docs, but for the overview/fields headers that, by convention, map to the entity/field jsdocs, we do sync them over.
Something i read in the earlier paragraphs about llms being easier to work with in greenfield projects…
My experience has been the opposite. They work well on existing projects but are not so great at new ones (unless you are just vibe coding something simple).
I think it has something to do with it being able to rely on years of established structure/conventions on existing projects that makes them better IME
Agreed! I hate the start of a project with an LLM. The code it writes is always expletive, and wrangling it with rules, hooks, plans, and what have you is just a series of half measures. I'm basically in a fistfight with a cheerful drunk keyboard.
Once there's enough code, tests, and docs to appease the prefill gods and give it an understanding of what I expect in terms of code quality, style, and TDD etiquette, I can expect more consistent performance.
That said, I've never been comfortable enough with it to let it write more than a few of functions at a time without a diff/checking-in.
I've seen folks with 500K LOC branches written by an LLM and my heart just sinks...
If you start with architecture docs that outline the high level goals and what the core abstractions should be for a greenfield project and refine it as you fill in the actual code, you can get pretty good results.
That gives the first few agents the "north star" and then subsequent agents have the actual code, the runtime behavior, and the user experience expectations to work against after that, in addition to the architecture docs, that provide hard boundaries to vibe off of.
My experience with greenfield is that when you're just starting out, you have to be really strict and careful about overall architecture, establish a framework, provide structure, separate concerns well.
If the project is not throwaway, this first set of iterations is critical, without guidance LLMs will just bolt on things and produce special-cased spaghetti, even the very-frontier models.
Once the framework is established, it's smooth sailing.
But I haven't gotten to "very large" size yet, I expect it produces another kind of coherence problems.
I’ve had a good experience with a greenfield project.
The single thing that seems to have helped is that we all agreed to use OpenSpec early on, and to commit the specs alongside the code.
I have no affiliation with OpenSpec and I don’t suspect it’s doing anything unique here, but having the intent develop alongside the code in the repository seems to have ensured that agents have a more holistic view of the project.
It’s a night/day difference when I use an agent against this codebase that integrates its changes using OpenSpec and those that ignore it.
Depends on what the existing project looks like ;-)
One repo at my work is DDD-esque, clean architecture, modular monolith.
It suuuucks to work on. Even without AI, editing any feature requires navigating through a bunch of app, infra, domain layers instead of living in a vertical feature slice. Indirection from cargo cult design patterns we don't need, like Mediator. With AI, degradation in quality and accuracy is plainly obvious. Everything needs hand-holding. AI is dumber than our average human dev so the creaky architecture made things 10x worse
The greenfield one we started with AI a few months ago is kicking along nicely. Warts exist, there are some really large files and questionable bits, but it's way easier to work on because we didn't make the architecture some convoluted dwemer mine thingy
i'm sorry to say that besides like the first paragraph, this article is mostly ai slop. i thought that we mostly weren't allowing ai written content on hn? (@dang whats the policy here?)
aside from that, i think that this is overcomplicating a more basic idea. You should have agents that leave docs behind in your codebase, and those docs should be localized to the structure of your code, but you do not need any fancy format or manifest.
We've had a lot of success with having the agents write and maintain a docs.md file in each folder, and requiring that they both read and update that file whenever they make a change in that folder
Just look at the companies YC invests in if you want clarification on HN’s AI stance.
When asked about DDD at a job interview I joked it was Debug Driven Development, when you start with empty directory and file the first bug "The app doesn't do anything. It should ..."
Non-ironically that's pretty much the way I work with AI agents now.