Author here, if you don't want to read all that, I'll post one excerpt that I think sums it up nicely:
> My point is, the spec must live somewhere, even if you don’t write it down. The spec is what you want the software to be. It often exists only in your head or in conversations. You and your team and your business will always care what the spec says, and that’s never going to change. So you’re better off writing it down now! And I think that a plain old list of acceptance criteria is a good place to start. (That’s really all that `feature.yaml` is.)
The traditional name for this spec is ‘source code’ — a canonical source of truth for the behaviour of a system that is as human-readable as we know how to make it, that will be processed by automated tools into a less-readable derived artefact for a computer to execute.
Checking the compiled artefact into the codebase without checking in its source code has always been a risky move!
I independently converged on something similar. I use two to three specification docs for my c++ work: a firmware manual (describes features and interfaces)) , an implementation plan (order of implementation, mechanisms where specified - new features go in here) and a product manual ( user story, external effects) I start with a user story, build an implementation plan, write the code, write the firmware manual, check the 3 documents +code for consistency and coherence. Either change the code or the documentation to reflect a coherent unified truth. (Implementation plan gradually becomes as-built) I also have the code comprehensively commented so that it is difficult to misinterpret. “Correct, coherent, consistent, commented”
We iterate features by feature through this process, and occasionally circle back on the original product manual to identify drift.
I also open each new context with an onboarding process that briefly describes the logos and the ethos of the project, why the agent should be deeply invested in the success of the project, as well as learnings.md which the agent writes as it comes across notable gotchas or strong preferences of mine.
Needless to say, I use one million context , and it’s a token fire… but the results are solid and my productivity is 5-10x
I wrote something similar recently about how agent-generated code lacks the institutional memory that human-written code has. There's nobody to ask why a decision was made (1).
“Specsmaxxing” is basically the right response to this. When you can't rely on authorial memory, you have to put the intent somewhere durable. Specs become the source of truth by default if we continue down the road of AI generated code.
I've been attaching to my commit messages a Git Trailer [1] of the Session UUID from the Claude Code conversation that created that commit.
It allows Claude to look back into the session where a change was made and see the decisions made, tradeoffs discussed and other history not captured by code, tests.
I had a similar experience refactoring a large codebase• The only thing that made it possible was that each commit message had a JIRA ticket number tying it to a requirement or task. I could find the people behind the business logic and ask them about it.
the recursive-mode workflow has full traceability, including why decisions were made, what the original requirement was, what the previous state was, etc. https://recursive-mode.dev/introduction
What's the difference between this and Jira. Your specs already live somewhere, it's where you defined them. That's why it's nice to put the Jira ticket number in your code / commit, so you can refer back to the spec when something breaks
A specification isn't a series of change requests! Using Jira as your source of truth is no different to just recording all your prompts. There's nothing you can easily review to spot contradictions or how things interact with one another.
I've been doing "specmaxxing" for a few months now. Unlike the author I don't use Yaml, I use a mix of Markdown and Gherkin. If you haven't encountered Gherkin before, it's not new and you might know it under the name Cucumber or BDD.
Gherkin is basically a structured form of English that can be fed into a unit testing framework to match against methods.
The nice thing about writing acceptance criteria this way is that they become executable and analyzable. You write some Gherkin and then ask the model to make the tests execute and pass. Now in a good IDE (IntelliJ has good support) you can run the acceptance criteria to ensure they pass, navigate from any specific acceptance criteria to the code which tests it (and from there to the code that implements it), you can generate reports, integrate it into CI and so on.
And when writing out acceptance tests that are quite similar, the IDE will help you with features like auto-complete. But if you need something that isn't implemented in the test-side code yet, no big deal. Just write it anyway and the model will write the mapping code.
There's a variant of Gherkin specifically designed for writing UI tests for web apps that also looks quite interesting. And because it's an old ecosystem there's lots of tooling around it.
Another thing I've found works well is asking the models to review every spec simultaneously and find contradictions. I've built myself a tool that does this and highlights the problems as errors in IntelliJ, like compiler errors. So I can click a button in the toolbar and then navigate between paragraphs that contradict each other. It's like a word processor but for writing specs.
Once you're doing spec driven development, you don't need to write prompts anymore. Every prompt can just be "Update the code and tests to match the changes to the specs."
The problem with gherkin is that it was a badly designed language.
The general idea of "readable specification language" was an inspired one but it failed on execution - it has gnarly syntax, no typing and bad abstractions.
This results in poor tests which are hard to maintain and diverge between being either too repetitive to be useful or too vague to be useful.
The ecosystem is big but it's built on crumbling foundations which is why when most people used it most of them got frustrated and gave up on it.
Annoyingly there's a certain amount of gaslighting around it too ("it didnt work for you coz you werent using it correctly") which is eleven different kinds of wrong.
Jira is only a set of changes though. What happens on a long (10+ year) and complex (10+) developer project with many changes and revisions? Eventually you need an explicit specification that itself has a "current state", and a change log. Theoretically you could generate this from Jira, but in my experience it eventually became a mess on any larger project that didn't have explicit and maintained writen requirements.
All through the agile era I wrote detailed specs for projects and then followed an agile process. The most successful parts of every project were the ones that we were able to spec best even when they diverged significantly from the original spec.
You don't plan to follow the plan. You plan in order to understand the whole problem space. Obviously no plan survives contact with reality.
Another point of view is that LLM:s perform to an extent on the same level as outsourcing does. This interface requires a bit more contract mass than doing everything within single team.
We never left waterfall in the end.
Working with and for dozens, collaborating with probably a hundred software companies in different scales, every single one said:
We do agile
Guess what?
Every single one of them was doing waterfall.
Their agile included preplanning and pre-specifying the full spec and each task, before the project kicked off. We'd have meetings where we'd drill down into tasks, folks would write them down so detailed that there would be no other way than doing that. Agile would be claimed, but the start date, end date, end spec and number of developers was always concrete.
Sometimes, the end date was too late, so a panic would ensue. Most of the time, the date was too late because developers had "unknowns" which then had to be "drilled down and specced so they wouldnt be unknowns". Sometimes, nearly 50% of the workweek was spent on meetings.
A few times, a project was running late - so to make sure we are _really_ doing it agile, we'd have morning standups, evening standups, weekly plannings, retrospectives, and backlog refinement. It would waste the time, and the "unknowns" aka "tickets to refine" were again, as always, dependant upon the PM/PO/CEO's wishes, which wouldn't get crystallized until it was _really last minute_.
One customer wanted us to do a 2 year agile plan on building their product. We had gigantic calls with 20+ people in them, out of which at least half had some kind of "Agile SCRUM Level 3 Black belt Jirajitsu" certificates.
To them, Agile was just a thing you say before you plan things. Agile was just an excuse to deal with project being late by pinning it on Agile. Agile was just a cop out of "PM didn't know what to do here so he didnt write anything down". Agile was a "we are modern and cool" sticker for a company.
And unfortunately, to most of them, agile was just a thing you say for the job, as their minds worked in waterfall mode, their obligations worked in waterfall mode, companies worked in waterfall mode, and if they failed their obligation to the waterfall, their job would go down one.
So while we were doing the Agile ceremonies, prancing around with our Scrum master hats, using the right words to fit into the Agile™ worldview - we were doing waterfall all along.
And after 15 years, I'm not even sure - did agile really ever exist?
> When rewriting the entire codebase is very quick and cheap, why bother iterating on small components?
We are nowhere near this scenario tbh. Token cost is very high and is currently heavily subsidized by VC money to gain market share. Also this realistically only applies to small projects, small codebases and mostly greenfield ones. No way you can rewrite the whole codebase quickly and cheaply in any mid-sized+ projects
But even assuming token cost plummets, any non-trivial piece of software that is valuable enough to generate income for the company is also big, complex, interconnected enough that cannot be rewritten quickly even by AI, also for business reasons too. If a piece of code works, is stable and is tested, then rewriting it will always bring a high degree of risk and uncertainty that in a lot of business critical applications is just not worth it. A stable system can stay untouched for years besides minor dependencies updates.
Nice! Your spec-maxxing is very resonant. I've been doing working with explicit requirements: elicit them from conversation with me or introspecting another piece of software; one-shot from them; and keep them up-to-date as I do the "old man shouts at Claude" iterations after whatever one-shotting came up with.
Unlike you, I wish for the LLM to do as much of the work as possible -- but "as possible" is doing a lot of work in that sentence. I'm still trying to get clear on exactly where I am needed and where Opus and iterations will get there eventually.
It has really challenged me to get clearer on what a requirement is vs a constraint (e.g., "you don't get to reinvent the database schema, we're building part of a larger system"). And I still battle with when and how to specify UI behaviours: so much UI is implicit, and it seems quite daunting to have to specify so much to get it working. I have new respect for whoever wrote the undoubtedly bajillion tests for Flutter and other UI toolkits.
Forgot to add: I get several benefits from doing this.
1. Specifications that live outside the code. We have a lot of code for which "what should this do?" is a subjective answer, because "what was this written to do?" is either oral legend or lost in time. As future Claude sessions add new features, this is how Claude can remember what was intentional in the existing code and what were accidents of implementation. And they're useful for documenters, support, etc.
2. Specifications that stay up to date as code is written. No spec survives first contact with the enemy (implementation in the real world). "Huh, there are TWO statuses for Missing orders, but we wrote this assuming just one. How do we display them? Which are we setting or is it configurable?" etc. Implementer finds things the specifier got wrong about reality, things the specifier missed that need to be specified/decided, and testing finds what they both missed.
I have a colleague working on saving architecture decisions, and his description of it feels like a higher-abstraction version of my saving and maintaining requirements.
Specifications doesn't tell you what to do, they say what the end state should be. In between that you need a codebase analysis step and an implementation plan.
I do (1) the same but (2) differently. In my workflow, (2) are AI generated specs using human written (1) as the input. It's an intermediate stage between (1) and the codebase, allowing for a gradual token expansion from 30k to 250k to the final code which is 2-3M. The benefit I've found with this approach is it gives the AI a way to iterate on the details of whole system in one context window, whereas fitting the whole codebase into one prompt is impossible. The code is then nothing more than a style transfer from (2).
So what I'm building is a github clone with epics/issues/kanban + specs/requirements/standards + CI/testing/coverage with the idea that all of those things connect so issues+requirements+testing all interact via code+webUI+CLI the point being that we can specify how a product is to function and the steps to get there and it's less a matter of telling a person or an LLM to read and implement the spec and more software actually keeping track at all times.
I actually read it all since it did not contain any hints of being AI generated (although I wouldn't be surprised to learn you did use AI to write it), so thank you for that. It's kind of crazy how I now have the default expectation that posts posted here are AI slop with little thought or care put in.
I am also stealing the idea of talking to LLMs as if it's an email. So funny, we need to be joymaxxing a bit more I think :)
:) Here is a crazy thought - what if we had some kind of a narrowed down, specific subset of normal language which would translate into specific computer-level instructions. So for example, instead of telling computer to read something from a file and transform it in a certain way, you actually had a specific instruction to open a file, which worked the same each time you used it and guaranteed to fail if you used it the wrong way? Wow, the possibilities are endless :)
Don’t be ridiculous, that would be extremely hard. Oppressive even, because it’s unattainable to an average person. And it is, otherwise there would be millions of programmers in the world. Was it unattainable or “we have to pay these suckers money, and they have rights and lives outside of work”? Bah! Just make sure to renew your subscription, agent will do the thinking and you bring the money.
But Paul Graham says that the guy from Replit whom he funded told him the source code is "object code" now, so we don't need to look at it all ? It must be utter wisdom since PG managed to get wealthy by selling some website during dotcom-mania so he must have insights we are missing?
Behaviour Driven Development or Spec Driven Development are, loosely, forms of Test Driven Development where you encode the specification into the code base. No impedance, full insight, formality through code.
I think people get really dogmatic about “test” projects, but with a touch of effort a unit test harness can be split up into integration tests, acceptance tests, and specification compliance tests. Pull the data out as human readable reports and you have a living, verifiable, specification.
Particularly using something comparable to back-ticks in F#, which let test names be defined with spaces and punctuation (ie “fulfills requirement 5.A.1, timeouts fail gracefully on mobile”), you can create specific layers of compiled, versioned, and verifiable specification baked into the codebase and available for PMs and testers and clients and approval committees.
> Nothing beats an organic, pasture-raised, hand-written spec.
Hah, I strongly empathize with the wording. I’ve been starting my design docs for fellow humans with “100% hand-written, organic content”, I might steal a part of yours.
Overall, cool idea. I don’t see myself using your SaaS, but the approach of tagging the requirements and constraints to make them easier to find sounds good.
One project you didn’t mention which I think is also, I think, a cool perspective on this is codespeak.dev , but I haven’t given it a go yet.
All in all, I feel like maintaining specs, and having agents translate spec diffs into code diffs is a promising area for the future. Good thing I enjoy writing!
Fair, I could have made that point clearer. It's a couple things. First is that I finally stopped experimenting with TUIs, harnesses, models, subagents, roles, skills, mcp, md libraries etc. and have mostly settled on this approach, and got back to building other things with it. I'm sure that won't last forever though.
Second is that I'm doing a lot less "seat of my pants prompting" and doing more engineering and ideating, which was a big goal of mine. So I'm feeling less psychotic there too.
And sort of tangentially to that, I think a significant subset of devs actually are willing to just prompt their way to nirvana, day in and day out. I'm not. I think the spec will carry a lot of weight for a long time. Maybe they will get further than I give them credit for? Maybe the whole digital world becomes a single chat box?
Some people seem to give very little thought to semantics and semiotics lately, to the point where people use words vaguely without even looking it up.
I guess I misappropriated the term then, woops. AI OCD? AI obsession? Whatever you call the behavior that I saw myself and others falling in to. Getting obnoxiously fixated on the tooling and the models to a counterproductive degree.
AI psychosis: (informal) A phenomenon wherein individuals reportedly develop or experience worsening psychosis, such as paranoia and delusions, in connection with their use of chatbots.
That’s the best part: you don’t. “You would extend the prompt to improve it”. They’ll just ask Claude to write an AI tool to overcome psychosis (the program will spam Anthropic servers with racial slurs which will promptly cause ban of the user, success).
So...is this just Cucumber cough cough behavior driven design again, but stored in YAML so that LLMs can read it easier by loading the AST instead of tokenizing the text?
It's not. And LLMs don't do well with YAML either. I've had the agent/model struggle with `sed` trying to count how many spaces are in there multiple times to get the file to pass. It's the worst format you can use for LLMs.
I use OpenSpec for my spec management, and I scrolled down to the comparison. The gripe seems to be with a semantic difference. Specs describing a current system is the basis for AS/IS Gap Analysis.
Also, I mainly pursue these tools so that I can have AI accelerate this process and broker an agreement after negotiating specs with the agent.
I'm also doing openspec for a few months now and it's really good if you invest enough in the specs (in the beginning I skimmed over much, now I pay attention to all details and fix anything that's wrong or where I see a gap).
The one thing I like that OP brings is to tie specs and code together. The openspec flow does help a lot in keeping code synced with specs, but when a spec changes, AI needs to find the relevant code to change it. It's pretty easy to miss something in large codebase (especially when there is lots of legacy stuff).
Being able to search for numbered spec tags to find relevant bits of code makes it much more likely to find what needs to be changed (and probably with less token use too).
I can see one benefit to a structured yaml for specs like the OP is doing: it gives you more control over what you include in the context window. But coming up with a good schema that doesn't handicap you or add cognitive burden, compared to the freeform flexibility of md/txt, is a challenge.
If the selling point is a new file format for spec management, it would be more interesting to provide an offering with org-mode. The author admits they were unaware of other pre-existing solutions before this project so I am providing context to their critique of OpenSpec.
Old ist new I guess. This is independent of whether A"I" or a human executes, the point is that you need this if specifying and execution lie apart, be it in time or space. This is basically the whole point of the V-Model and processes (if used correctly as a tool and not preferred as goals) and was already researched an formalized in the 60s and 70s.
But have you thought about “fun factor”? It’s where you sit like an addict in a casino for weeks and burn tokens in a hope of winning a software that you could’ve written?
Who doesn’t consider “fun” thinking about work crap all the time, writing to your agent, verifying walls of slop?
It’s like a yaml of an event model but less graphical. Right? I think I will prefer Event Modeling especially with Martin Dilger now building tooling very much with agents in mind. There is no one place to read about his most recent efforts except for his LinkedIn feed though I fear.l so won’t post any urls, but information is easy enough to find.
A full blown event model facilitates all communication, human (management, devs, ops) and agentic. But maybe I’m missing something, maybe the dashboard can have this function I didn’t dig into it too much.
I’m using Opus 4.6 and I’m so confused! Maybe I should try Opus 4.7, which is almost twice as expensive to get some clarity (but not too much, I need to save money for Opus 4.8)?
That's what the article is about - overcoming problems with AI cooding tools using specs in Yaml. If we've got that far, it might be better to write specs in a proper programming language instead and skip the AI layer altogether
Think the idea is to still get monumental acceleration between fancy YAML specs (bullet points with some indentation that an intelligent technical manager could write) and production ready code.
I just spent a week training up in spec driven development through bmad, which was awful, and speckit which was ok but not great. Both had what seemed like unnecessary ceremony around the specs, generating fields of spec documents which presumably fill up the context window quickly. I just kept thinking "this should be using something simpler, all this markdown is unnecessary"
What is yours agentic development experience with elixir? I used to like elixir a lot during a pre agentic era, but with coding agents it feels like the language isn't the best choice - slow compile time, weak type system (at least it was a year ago, I know there is work on that front), small ecosystem...
I’m building something similar with https://github.com/LabLeaks/special (apologies for the desultory slop-laden README, need to give that a lot more human attention) but I’ve gone in a slightly different direction: a “spec” is a product contract claim supported by attached tests that verify it. It’s a little Cucumber-y, if anyone remembers that, but a lot more flexible — you just write stuff like
@spec LINT_COMMAND.ORPHAN_VERIFIES
linter reports blocks that do not attach to a supported owned item.
Then
#[test]
// @verifies SPECIAL.LINT_COMMAND.ORPHAN_VERIFIES
fn rejects_orphan_verifies_blocks() {
let block = block_with_path("src/example.rs", &["@verifies EXPORT.ORPHAN"]);
let parsed = parse_current(&block);
assert!(parsed.verifies.is_empty());
assert_eq!(parsed.diagnostics.len(), 1);
assert!(
parsed.diagnostics[0]
.message
.contains("@verifies must attach to the next supported item")
);
}
And then the CLI command “special specs” pulls your specs and all attached verification + test code so you (or your LLM) to analyze whether the (hopefully passing!) test actually supports the product claim.
There’s also a bunch of other code quality commands and source annotations in there for architectural design & analysis, fuzzy-checking for DRY opportunities, and general codebase health. But on the overall principle, this article is dead-on: when developing with LLMs, your source of truth should be in your code, or at least co-located with it.
Completely subjective take, but I feel like 95% of these "tools" that are prompt-engineering inventions created by the authors with their bias and to suit their needs don't have anything supporting them besides the authors' subjective experience.
I have seen the same idea with processes, pipelines, lists, bullet points, jsons, yamls, trees, prioritization queues all for LLM context and instruction alignment. It's like the authors take the structure they are familiar with, and go 100% in on it until it provides value for them and then they think it's the best thing since sliced bread.
I would like, for once, to see some kind of exploration/abalation against other methods. Or even better, a tool that uses your data to figure out your personal bias and structure preference for writing specs, so that you can have a way of providing yourself value.
Dear Claude,
I hope this email finds you well.\
I am writing to ask if you could please do another task for me.\
Start by running \`npx @acai.sh/cli skill\`.\
This will teach you everything you need to know about our process for spec-driven development. Then, proceed to plan and implement the features specified in our spec files.
Love,\
\[your-name]
Honestly, I can no longer tell parody from reality. Whether in politics or AI.
> We are entering the post-slop era. My software is more robust, better tested, better integrated, and more observable than ever before. And my velocity keeps increasing!
Don't we just love the hard fact conclusions based on sample size N=1 and hand-waving arguments?
It's ok friend, all I did is put acceptance criteria in a list so I can parse it and quickly track cross-references. The rest is just Elixir/Phoenix and some creative writing.
I also have started numbering my Acceptance criteria and pushing that across the team(s). It’s going pretty well. Some note however are
1. Don’t write in yaml. It’s really hard for humans. Write in markdown and use a standard means to convert to lists / yaml.
2. Think beyond you writing your own specs - how does this expand into teams of tens or more. The ticketing system you have (jira? Bugzilla) is not designed for discussion of the acceptance criteria. I think we are heading into a world of waterfall again where we have discussions around the acceptance criteria. This is not a bad thing - is used to be called product management and they would write an upfront spec.
If this new world of a tech and a business user lead the writing of a new spec (like a PEP) and then then AI implements it and it’s put into a UAT harness for larger review and a daily cycle begins, we might have something.
YAML is one of the worst technologies ever invented, it has more warts than features. One of the benefits of LLMs is that they can write YAML for me, wherever I am forced to use it.
Otherwise, I like the idea of machine-readable specs.
the token usage isn’t sustainable. formal english is a barrier but requirement for specification. brevity is the language of money and that’s the premise of management using ai.
fyi language alone can’t define/describe requirements which is why UML existed.
Natural language is a fully general system and can define and describe everything.
You could deterministically process any UML diagram into a prose equivalent.
And in fact you couldn't do the other way around (any prose -> UML) because UML is less powerful than natural language and actually can't express everything that natural language can.
Can it also fully describe a composition by Bach or a Rembrandt's painting? In some weird, overly complex way it probably 'could', but it would be very painful. That's why we pick other forms of expression. We use other forms of expression to compact and optimise information delivery. Another benefit is that we cut out the noise. So yes UML cannot describe everything natural language can, but then again why should it - it was designed as a specific framework for designing relations between objects. Not more and not less. Similar for sequence diagrams or other forms of communicating ideas efficiently.
Could it be that slop PRs are less frequently rejected/commented due to (unfortunate) increased acceptance of it?
As it turns out when maxxing AI on leaf parts of a program, the quality of the code doesn't matter that much anymore when compared to building the fundament.
What is it with people and procrastinating with the most useless shit you can imagine?
First it was choice of editor: people were micro optimizing every aspect of their typing experience, editor wars where people would literally slaughter over suggesting another camp.
Editor wars v2: IDEs arrived and second editor war began.
Revenge of the note taking apps: Obsidian/Roam/Joplin/Apple Notes/Logseq. Just one plugin, just one more knowledge graph, bro, and I’ll have peak productivity. 10x is almost here.
AI: you’re witnessing it now.
Do people NOT have anything else in life? How are y’all finding time to do all of this shit? Are you doing it on company time? Do you have hobbies, do you learn foreign languages, travel, have kids or spouses, drive a car, other thousand “normie” things outside of staring at the freaking monitor or thinking about this shit 24/7? Did I miss the invention of a Time Machine?
Lmfao. Going to a site for computer geeks and complaining that they are computer geeks.
Also, a lot of folks don't write code anymore, and barely have the time to read the volume of code that AI produces. This may just be one of the most profound changes in an industry, and some folks are excited about it and want to get better at building with it.
I think the person who wrote this post made a good faith effort to share his learnings while promoting his tool.
It's fun how people brag of their agentmaxxing, but if you ask them what those agents are busy actually producing, it's invariably another agent harness so they can agentmaxx better. NFT/blockchain ecosystem was much the same.
>Do people NOT have anything else in life? How are y’all finding time to do all of this shit? Are you doing it on company time? Do you have hobbies, do you learn foreign languages, travel, have kids or spouses, drive a car, other thousand “normie” things outside of staring at the freaking monitor or thinking about this shit 24/7? Did I miss the invention of a Time Machine?
How are any of those things even remotely as interesting as arguing with people about an Emacs config?
The author is right but his message ain’t specsmaxing, because while somewhat understandable as a rationale what does it actually mean?
In other words: specs can be as detailed as it gets, and this is why developers have a hard time when they face as a senior an NDAed regulated environment. It ain’t software craftsmanship but data flow, hardware components, compliance on the lowest level including supply chains often times, information architecture - a simple app needs to comply to specs that amount to thousands of pages.
Context window: circular reference. A year ago? Specsmaxing by really weeding out any redundant words. Today? Yawn, like with 8mb RAM vs 512 Gigabytes.
AI wants to be easy on us so what is a spec anyway then?
To put it this way: the spec for the spec is constantly evolving.
Last year’s prompts lead to extremely different results today no matter how maxed out.
The author was on point with his introduction: AI is as junior in many ways when it comes to any sort of efficiency and optimization.
This is my revaluation after years of experimenting with AI. Beautiful code, sophisticated but performance wise and its architecture are laughable at best.
AI is not trained on optimization. Not the slightest and juniors have no clue about algorithms and Big O.
In fact Google used Big O as a basic entry level interview question for a very long time. They have to but the simple fact that in my experience 99% of devs never heard or consider it speaks volumes.
AI cannot compensate for that (yet).
I went the opposite and my specs focus heavily on architecture and the obvious dumb performance drains noobs do.
Google was mocked about Big O. And yes, failing to understand that Big O can be neglected thankfully in 99% of cases is part of its logic.
AI bloats your code. And a year long single dev project gets pumped out in hours. In short: a homerun for Big O because it looks on results that change depending on the variables. A function in mathematical terms.
So I think the author did a funny and great job of you focus on Big O if needed. Everything else is not that important because of being open to change and extension.
Big numbers need great architecture.
It screams loudly. And also think about leaks. Before AI I had virtually no memory leaks at all. Since AI NodeJS and React are worse leaking compared to IE 6 and 8.
I mean it.
Big O reduces them significantly, so don’t work around the Elephant in the room.
Architecture and optimization is brutally hard. Google blew my mind in this regard but this is another story of squeezing out even milliseconds out of a build tool used by all. A single dev laughs at it but failed the calculation as well as abstraction.
Should I apologize for being excited about something I built and use daily and for wanting people to try it, discuss it, critique it? Not sure by the tone of your message.
Don't apologize. Keep writing and trying things. Ignore the haters and non-curious, listen to the (even if salty) interested.
There's a fair amount of talk right now about the value being in the verification layer -- once there's a hard verification loop, the agents can do amazing things without getting (permanently) sidetracked. I think what you're working on is half way there -- in essence, you're probably relying on the LLMs notion of what a spec is and should be to the codebase.
What's not currently solved, and what I think is very interesting is how much automation can be added to the creation of verification. We all would unlock a lot more speed and productivity for even moderate gains on that side.
Read the room. What you "built" is neither exciting, nor something most people want to "try". Why? Because just like other AI boosters, you are still trying to somehow optimise the usage of natural language to make it work. But it will never "work" because the way the stochastic ML system is built, it has a failure built into the system.
Totally agree it's not exciting, even though I am personally excited by it, and I also agree it's not something most people want to try, even though some people do want to try it-- and I found a few of them right here on HN.
Disagree on the bit about it "never going to work" though.
Failure-prone stochastic ML systems produce testable, auditable code... just like failure-prone human brains can produce testable, auditable code. And in fact, in both cases, changes to our process can reduce the amount of failures that slip past testing and audit. Or can reap other rewards. Finding the a better process is what I'm interested in right now.
Why is the vibecoding crowd still holding onto the idea that markdown (or here yml) is a better spec then code?
Seriously, it's just not
Write your code like it's your spec and your software will be more stable, maintainable clearer to read.
Code is not transient, it is your friggin spec itself
And if your code isn't structured like it's a spec, then your code is garbage from the perspective of LLM driven development
Most 'programmers' cannot read or write code very well (or reason about structure or architecture) and so they want to 'program in english'.
Author here, if you don't want to read all that, I'll post one excerpt that I think sums it up nicely:
> My point is, the spec must live somewhere, even if you don’t write it down. The spec is what you want the software to be. It often exists only in your head or in conversations. You and your team and your business will always care what the spec says, and that’s never going to change. So you’re better off writing it down now! And I think that a plain old list of acceptance criteria is a good place to start. (That’s really all that `feature.yaml` is.)
The traditional name for this spec is ‘source code’ — a canonical source of truth for the behaviour of a system that is as human-readable as we know how to make it, that will be processed by automated tools into a less-readable derived artefact for a computer to execute.
Checking the compiled artefact into the codebase without checking in its source code has always been a risky move!
I independently converged on something similar. I use two to three specification docs for my c++ work: a firmware manual (describes features and interfaces)) , an implementation plan (order of implementation, mechanisms where specified - new features go in here) and a product manual ( user story, external effects) I start with a user story, build an implementation plan, write the code, write the firmware manual, check the 3 documents +code for consistency and coherence. Either change the code or the documentation to reflect a coherent unified truth. (Implementation plan gradually becomes as-built) I also have the code comprehensively commented so that it is difficult to misinterpret. “Correct, coherent, consistent, commented”
We iterate features by feature through this process, and occasionally circle back on the original product manual to identify drift.
I also open each new context with an onboarding process that briefly describes the logos and the ethos of the project, why the agent should be deeply invested in the success of the project, as well as learnings.md which the agent writes as it comes across notable gotchas or strong preferences of mine.
Needless to say, I use one million context , and it’s a token fire… but the results are solid and my productivity is 5-10x
I wrote something similar recently about how agent-generated code lacks the institutional memory that human-written code has. There's nobody to ask why a decision was made (1).
“Specsmaxxing” is basically the right response to this. When you can't rely on authorial memory, you have to put the intent somewhere durable. Specs become the source of truth by default if we continue down the road of AI generated code.
1: https://ossature.dev/blog/ai-generated-code-has-no-author/
I've been attaching to my commit messages a Git Trailer [1] of the Session UUID from the Claude Code conversation that created that commit.
It allows Claude to look back into the session where a change was made and see the decisions made, tradeoffs discussed and other history not captured by code, tests.
[1] https://git-scm.com/docs/git-interpret-trailers
I had a similar experience refactoring a large codebase• The only thing that made it possible was that each commit message had a JIRA ticket number tying it to a requirement or task. I could find the people behind the business logic and ask them about it.
the recursive-mode workflow has full traceability, including why decisions were made, what the original requirement was, what the previous state was, etc. https://recursive-mode.dev/introduction
[dead]
What's the difference between this and Jira. Your specs already live somewhere, it's where you defined them. That's why it's nice to put the Jira ticket number in your code / commit, so you can refer back to the spec when something breaks
A specification isn't a series of change requests! Using Jira as your source of truth is no different to just recording all your prompts. There's nothing you can easily review to spot contradictions or how things interact with one another.
I've been doing "specmaxxing" for a few months now. Unlike the author I don't use Yaml, I use a mix of Markdown and Gherkin. If you haven't encountered Gherkin before, it's not new and you might know it under the name Cucumber or BDD.
https://cucumber.io/docs/
Gherkin is basically a structured form of English that can be fed into a unit testing framework to match against methods.
The nice thing about writing acceptance criteria this way is that they become executable and analyzable. You write some Gherkin and then ask the model to make the tests execute and pass. Now in a good IDE (IntelliJ has good support) you can run the acceptance criteria to ensure they pass, navigate from any specific acceptance criteria to the code which tests it (and from there to the code that implements it), you can generate reports, integrate it into CI and so on.
And when writing out acceptance tests that are quite similar, the IDE will help you with features like auto-complete. But if you need something that isn't implemented in the test-side code yet, no big deal. Just write it anyway and the model will write the mapping code.
There's a variant of Gherkin specifically designed for writing UI tests for web apps that also looks quite interesting. And because it's an old ecosystem there's lots of tooling around it.
Another thing I've found works well is asking the models to review every spec simultaneously and find contradictions. I've built myself a tool that does this and highlights the problems as errors in IntelliJ, like compiler errors. So I can click a button in the toolbar and then navigate between paragraphs that contradict each other. It's like a word processor but for writing specs.
Once you're doing spec driven development, you don't need to write prompts anymore. Every prompt can just be "Update the code and tests to match the changes to the specs."
I agree, Cucumber works really well with LLMs.
> I use a mix of Markdown and Gherkin
Gherkin also has a Markdown based syntax that is not well known:
https://github.com/cucumber/gherkin/blob/main/MARKDOWN_WITH_...
I prefer that to the 'verbose' original syntax. MDG also renders nicely in code forges.
I solved this five months ago with recursive-mode: recursive-mode.dev/introduction
The problem with gherkin is that it was a badly designed language.
The general idea of "readable specification language" was an inspired one but it failed on execution - it has gnarly syntax, no typing and bad abstractions.
This results in poor tests which are hard to maintain and diverge between being either too repetitive to be useful or too vague to be useful.
The ecosystem is big but it's built on crumbling foundations which is why when most people used it most of them got frustrated and gave up on it.
Annoyingly there's a certain amount of gaslighting around it too ("it didnt work for you coz you werent using it correctly") which is eleven different kinds of wrong.
Jira is only a set of changes though. What happens on a long (10+ year) and complex (10+) developer project with many changes and revisions? Eventually you need an explicit specification that itself has a "current state", and a change log. Theoretically you could generate this from Jira, but in my experience it eventually became a mess on any larger project that didn't have explicit and maintained writen requirements.
Jira has current state and a change log. The proposal here is "use yaml instead of jira." Same damn thing, same damn mess.
What about when you migrate away from Jira, or when there’s a Cloudflare outage?
> will always care what the spec says, and that’s never going to change
Did I miss something or is everyone back in 1970s, working in waterfall processes now?
All through the agile era I wrote detailed specs for projects and then followed an agile process. The most successful parts of every project were the ones that we were able to spec best even when they diverged significantly from the original spec.
You don't plan to follow the plan. You plan in order to understand the whole problem space. Obviously no plan survives contact with reality.
Agree!
Another point of view is that LLM:s perform to an extent on the same level as outsourcing does. This interface requires a bit more contract mass than doing everything within single team.
We never left waterfall in the end. Working with and for dozens, collaborating with probably a hundred software companies in different scales, every single one said:
We do agile
Guess what? Every single one of them was doing waterfall.
Their agile included preplanning and pre-specifying the full spec and each task, before the project kicked off. We'd have meetings where we'd drill down into tasks, folks would write them down so detailed that there would be no other way than doing that. Agile would be claimed, but the start date, end date, end spec and number of developers was always concrete.
Sometimes, the end date was too late, so a panic would ensue. Most of the time, the date was too late because developers had "unknowns" which then had to be "drilled down and specced so they wouldnt be unknowns". Sometimes, nearly 50% of the workweek was spent on meetings.
A few times, a project was running late - so to make sure we are _really_ doing it agile, we'd have morning standups, evening standups, weekly plannings, retrospectives, and backlog refinement. It would waste the time, and the "unknowns" aka "tickets to refine" were again, as always, dependant upon the PM/PO/CEO's wishes, which wouldn't get crystallized until it was _really last minute_.
One customer wanted us to do a 2 year agile plan on building their product. We had gigantic calls with 20+ people in them, out of which at least half had some kind of "Agile SCRUM Level 3 Black belt Jirajitsu" certificates.
To them, Agile was just a thing you say before you plan things. Agile was just an excuse to deal with project being late by pinning it on Agile. Agile was just a cop out of "PM didn't know what to do here so he didnt write anything down". Agile was a "we are modern and cool" sticker for a company.
And unfortunately, to most of them, agile was just a thing you say for the job, as their minds worked in waterfall mode, their obligations worked in waterfall mode, companies worked in waterfall mode, and if they failed their obligation to the waterfall, their job would go down one.
So while we were doing the Agile ceremonies, prancing around with our Scrum master hats, using the right words to fit into the Agile™ worldview - we were doing waterfall all along.
And after 15 years, I'm not even sure - did agile really ever exist?
Sort of, but the downside of waterfall was you build the wrong thing and waste a shitload of time rewriting it.
When rewriting the entire codebase is very quick and cheap, why bother iterating on small components?
> When rewriting the entire codebase is very quick and cheap, why bother iterating on small components?
We are nowhere near this scenario tbh. Token cost is very high and is currently heavily subsidized by VC money to gain market share. Also this realistically only applies to small projects, small codebases and mostly greenfield ones. No way you can rewrite the whole codebase quickly and cheaply in any mid-sized+ projects
But even assuming token cost plummets, any non-trivial piece of software that is valuable enough to generate income for the company is also big, complex, interconnected enough that cannot be rewritten quickly even by AI, also for business reasons too. If a piece of code works, is stable and is tested, then rewriting it will always bring a high degree of risk and uncertainty that in a lot of business critical applications is just not worth it. A stable system can stay untouched for years besides minor dependencies updates.
waterfall is not the sole purveyor of written docs
distributed teams do well when proposals, decision, etc, are written down, and can be easily found and referenced
it doesn't mean docs are frozen in time and can't be patched like code
I read that as "the business caring about what the spec says will never change" rather than "the spec will never change".
waterfall doesn’t mean writing down decisions
[dead]
Nice! Your spec-maxxing is very resonant. I've been doing working with explicit requirements: elicit them from conversation with me or introspecting another piece of software; one-shot from them; and keep them up-to-date as I do the "old man shouts at Claude" iterations after whatever one-shotting came up with.
Unlike you, I wish for the LLM to do as much of the work as possible -- but "as possible" is doing a lot of work in that sentence. I'm still trying to get clear on exactly where I am needed and where Opus and iterations will get there eventually.
It has really challenged me to get clearer on what a requirement is vs a constraint (e.g., "you don't get to reinvent the database schema, we're building part of a larger system"). And I still battle with when and how to specify UI behaviours: so much UI is implicit, and it seems quite daunting to have to specify so much to get it working. I have new respect for whoever wrote the undoubtedly bajillion tests for Flutter and other UI toolkits.
Forgot to add: I get several benefits from doing this.
1. Specifications that live outside the code. We have a lot of code for which "what should this do?" is a subjective answer, because "what was this written to do?" is either oral legend or lost in time. As future Claude sessions add new features, this is how Claude can remember what was intentional in the existing code and what were accidents of implementation. And they're useful for documenters, support, etc.
2. Specifications that stay up to date as code is written. No spec survives first contact with the enemy (implementation in the real world). "Huh, there are TWO statuses for Missing orders, but we wrote this assuming just one. How do we display them? Which are we setting or is it configurable?" etc. Implementer finds things the specifier got wrong about reality, things the specifier missed that need to be specified/decided, and testing finds what they both missed.
I have a colleague working on saving architecture decisions, and his description of it feels like a higher-abstraction version of my saving and maintaining requirements.
Specifications doesn't tell you what to do, they say what the end state should be. In between that you need a codebase analysis step and an implementation plan.
My recursive-mode workflow handles all of that and more and gives you full traceability: https://recursive-mode.dev/introduction
I do (1) the same but (2) differently. In my workflow, (2) are AI generated specs using human written (1) as the input. It's an intermediate stage between (1) and the codebase, allowing for a gradual token expansion from 30k to 250k to the final code which is 2-3M. The benefit I've found with this approach is it gives the AI a way to iterate on the details of whole system in one context window, whereas fitting the whole codebase into one prompt is impossible. The code is then nothing more than a style transfer from (2).
Let's cut through the noise - what did you build with this very elaborate process and how much ARR is it generating ?
Asking the real questions. I would also really like know how much value AIs are bringing in terms of ARR or MRR.
Jfc
So what I'm building is a github clone with epics/issues/kanban + specs/requirements/standards + CI/testing/coverage with the idea that all of those things connect so issues+requirements+testing all interact via code+webUI+CLI the point being that we can specify how a product is to function and the steps to get there and it's less a matter of telling a person or an LLM to read and implement the spec and more software actually keeping track at all times.
I actually read it all since it did not contain any hints of being AI generated (although I wouldn't be surprised to learn you did use AI to write it), so thank you for that. It's kind of crazy how I now have the default expectation that posts posted here are AI slop with little thought or care put in.
I am also stealing the idea of talking to LLMs as if it's an email. So funny, we need to be joymaxxing a bit more I think :)
Great idea -- just one suggestion if you want it to catch on: perform some IncelCultureMinning on the nomenclature.
You probably don't want people associating your work with abusing crystal meth and hitting yourself in the face with a hammer.
For anyone missing the reference, SNL has a pretty good explainer:
https://www.youtube.com/watch?v=4XMPLdiXB1k
Wow - I love programming in YAML! You know what would make this really fun? Sprinkle in some Jinja. Then we'll be cooking with gas.
:) Here is a crazy thought - what if we had some kind of a narrowed down, specific subset of normal language which would translate into specific computer-level instructions. So for example, instead of telling computer to read something from a file and transform it in a certain way, you actually had a specific instruction to open a file, which worked the same each time you used it and guaranteed to fail if you used it the wrong way? Wow, the possibilities are endless :)
Don’t be ridiculous, that would be extremely hard. Oppressive even, because it’s unattainable to an average person. And it is, otherwise there would be millions of programmers in the world. Was it unattainable or “we have to pay these suckers money, and they have rights and lives outside of work”? Bah! Just make sure to renew your subscription, agent will do the thinking and you bring the money.
But Paul Graham says that the guy from Replit whom he funded told him the source code is "object code" now, so we don't need to look at it all ? It must be utter wisdom since PG managed to get wealthy by selling some website during dotcom-mania so he must have insights we are missing?
Exactly. At some point, the specification becomes so complex, it's easier to just write the code yourself.
It's why famously, programmers always say, the code is the documentation, because writing detailed docs is very tedious and nobody wants to do it.
There are middle-ways.
Behaviour Driven Development or Spec Driven Development are, loosely, forms of Test Driven Development where you encode the specification into the code base. No impedance, full insight, formality through code.
I think people get really dogmatic about “test” projects, but with a touch of effort a unit test harness can be split up into integration tests, acceptance tests, and specification compliance tests. Pull the data out as human readable reports and you have a living, verifiable, specification.
Particularly using something comparable to back-ticks in F#, which let test names be defined with spaces and punctuation (ie “fulfills requirement 5.A.1, timeouts fail gracefully on mobile”), you can create specific layers of compiled, versioned, and verifiable specification baked into the codebase and available for PMs and testers and clients and approval committees.
We might as well future proof this by writing specs in YAML-ified Ruby, this way it's more flexible, I've been told it's best practice!
Dreaming about ` | nindent 12` in my specs! :D
Crying in k8s templating
Love the writing style!
> Nothing beats an organic, pasture-raised, hand-written spec.
Hah, I strongly empathize with the wording. I’ve been starting my design docs for fellow humans with “100% hand-written, organic content”, I might steal a part of yours.
Overall, cool idea. I don’t see myself using your SaaS, but the approach of tagging the requirements and constraints to make them easier to find sounds good.
One project you didn’t mention which I think is also, I think, a cool perspective on this is codespeak.dev , but I haven’t given it a go yet.
All in all, I feel like maintaining specs, and having agents translate spec diffs into code diffs is a promising area for the future. Good thing I enjoy writing!
Unfortunate name collision between your ACID concept and the database principles (atomicity, consistency, isolation, durability).
Where is the part where the author overcomes ai psychosis? Reads like digging in deeper and deeper.
Fair, I could have made that point clearer. It's a couple things. First is that I finally stopped experimenting with TUIs, harnesses, models, subagents, roles, skills, mcp, md libraries etc. and have mostly settled on this approach, and got back to building other things with it. I'm sure that won't last forever though.
Second is that I'm doing a lot less "seat of my pants prompting" and doing more engineering and ideating, which was a big goal of mine. So I'm feeling less psychotic there too.
And sort of tangentially to that, I think a significant subset of devs actually are willing to just prompt their way to nirvana, day in and day out. I'm not. I think the spec will carry a lot of weight for a long time. Maybe they will get further than I give them credit for? Maybe the whole digital world becomes a single chat box?
I don’t understand how that relates to AI psychosis?
Some people seem to give very little thought to semantics and semiotics lately, to the point where people use words vaguely without even looking it up.
I guess I misappropriated the term then, woops. AI OCD? AI obsession? Whatever you call the behavior that I saw myself and others falling in to. Getting obnoxiously fixated on the tooling and the models to a counterproductive degree.
The thing is, devs with AI psychosis often work on memory systems and harnesses as part of their delusion, so i would not rule out you have it!
AI psychosis: (informal) A phenomenon wherein individuals reportedly develop or experience worsening psychosis, such as paranoia and delusions, in connection with their use of chatbots.
https://en.wiktionary.org/wiki/AI_psychosis
This is not psychosis.
That’s the best part: you don’t. “You would extend the prompt to improve it”. They’ll just ask Claude to write an AI tool to overcome psychosis (the program will spam Anthropic servers with racial slurs which will promptly cause ban of the user, success).
So...is this just Cucumber cough cough behavior driven design again, but stored in YAML so that LLMs can read it easier by loading the AST instead of tokenizing the text?
I didn’t quite understand why YAML is better than Markdown for such specifications.
If the specification is written in such a strict format as YAML, I would expect it to be executable, something like this https://blog.fooqux.com/blog/executable-specification/
But as far as I understood, for acai that is not the case.
It's not. And LLMs don't do well with YAML either. I've had the agent/model struggle with `sed` trying to count how many spaces are in there multiple times to get the file to pass. It's the worst format you can use for LLMs.
I use OpenSpec for my spec management, and I scrolled down to the comparison. The gripe seems to be with a semantic difference. Specs describing a current system is the basis for AS/IS Gap Analysis.
Also, I mainly pursue these tools so that I can have AI accelerate this process and broker an agreement after negotiating specs with the agent.
I'm also doing openspec for a few months now and it's really good if you invest enough in the specs (in the beginning I skimmed over much, now I pay attention to all details and fix anything that's wrong or where I see a gap).
The one thing I like that OP brings is to tie specs and code together. The openspec flow does help a lot in keeping code synced with specs, but when a spec changes, AI needs to find the relevant code to change it. It's pretty easy to miss something in large codebase (especially when there is lots of legacy stuff).
Being able to search for numbered spec tags to find relevant bits of code makes it much more likely to find what needs to be changed (and probably with less token use too).
I can see one benefit to a structured yaml for specs like the OP is doing: it gives you more control over what you include in the context window. But coming up with a good schema that doesn't handicap you or add cognitive burden, compared to the freeform flexibility of md/txt, is a challenge.
If the selling point is a new file format for spec management, it would be more interesting to provide an offering with org-mode. The author admits they were unaware of other pre-existing solutions before this project so I am providing context to their critique of OpenSpec.
Old ist new I guess. This is independent of whether A"I" or a human executes, the point is that you need this if specifying and execution lie apart, be it in time or space. This is basically the whole point of the V-Model and processes (if used correctly as a tool and not preferred as goals) and was already researched an formalized in the 60s and 70s.
And once you’ve written all these specs you realize it became so slow that it’s faster to do it yourself in editor
People don’t actually track wall clock time, I’ve noticed.
at which point you realize you never had a plan written down and you are using the code as a spec
Which takes us back to this:
https://haskellforall.com/2026/03/a-sufficiently-detailed-sp...
But have you thought about “fun factor”? It’s where you sit like an addict in a casino for weeks and burn tokens in a hope of winning a software that you could’ve written? Who doesn’t consider “fun” thinking about work crap all the time, writing to your agent, verifying walls of slop?
It’s like a yaml of an event model but less graphical. Right? I think I will prefer Event Modeling especially with Martin Dilger now building tooling very much with agents in mind. There is no one place to read about his most recent efforts except for his LinkedIn feed though I fear.l so won’t post any urls, but information is easy enough to find.
A full blown event model facilitates all communication, human (management, devs, ops) and agentic. But maybe I’m missing something, maybe the dashboard can have this function I didn’t dig into it too much.
I'm still confused as to why folks don't just write executable specs.
Some of us do! That's called Gherkin.
Ambiguity is the grease that keeps everything turning.
So basically tests?
Yes, except a test can be turing complete - i.e. code.
An executable spec like gherkin or hitchstory is config - it has no loops or conditionals. There are a number of rarely recognized benefits to this.
Could you expand on this?
code
Literate programming would provide specs and code instead of working backwards from hard coded functions to figure out specs.
> working backwards from hard coded functions to figure out specs.
People do that? Actual professionals?
If you're confused, and have tried Opus for coding, I'm keen to hear what problems or workflows it's not good at.
If you're genuinely confused, and haven't tried Opus for coding, then it's not surprising you're confused!
It is also okay for you to just not like the idea of LLMs for coding (but say that!).
I’m using Opus 4.6 and I’m so confused! Maybe I should try Opus 4.7, which is almost twice as expensive to get some clarity (but not too much, I need to save money for Opus 4.8)?
That's what the article is about - overcoming problems with AI cooding tools using specs in Yaml. If we've got that far, it might be better to write specs in a proper programming language instead and skip the AI layer altogether
Think the idea is to still get monumental acceleration between fancy YAML specs (bullet points with some indentation that an intelligent technical manager could write) and production ready code.
I just spent a week training up in spec driven development through bmad, which was awful, and speckit which was ok but not great. Both had what seemed like unnecessary ceremony around the specs, generating fields of spec documents which presumably fill up the context window quickly. I just kept thinking "this should be using something simpler, all this markdown is unnecessary"
This seems like the answer to that thought!
What is yours agentic development experience with elixir? I used to like elixir a lot during a pre agentic era, but with coding agents it feels like the language isn't the best choice - slow compile time, weak type system (at least it was a year ago, I know there is work on that front), small ecosystem...
Stopped at "Specsmaxxing".
I’m building something similar with https://github.com/LabLeaks/special (apologies for the desultory slop-laden README, need to give that a lot more human attention) but I’ve gone in a slightly different direction: a “spec” is a product contract claim supported by attached tests that verify it. It’s a little Cucumber-y, if anyone remembers that, but a lot more flexible — you just write stuff like
Then }And then the CLI command “special specs” pulls your specs and all attached verification + test code so you (or your LLM) to analyze whether the (hopefully passing!) test actually supports the product claim.
There’s also a bunch of other code quality commands and source annotations in there for architectural design & analysis, fuzzy-checking for DRY opportunities, and general codebase health. But on the overall principle, this article is dead-on: when developing with LLMs, your source of truth should be in your code, or at least co-located with it.
Completely subjective take, but I feel like 95% of these "tools" that are prompt-engineering inventions created by the authors with their bias and to suit their needs don't have anything supporting them besides the authors' subjective experience.
I have seen the same idea with processes, pipelines, lists, bullet points, jsons, yamls, trees, prioritization queues all for LLM context and instruction alignment. It's like the authors take the structure they are familiar with, and go 100% in on it until it provides value for them and then they think it's the best thing since sliced bread.
I would like, for once, to see some kind of exploration/abalation against other methods. Or even better, a tool that uses your data to figure out your personal bias and structure preference for writing specs, so that you can have a way of providing yourself value.
It's Vibesmaxxing
It's like horoscopes for the entirely-too-AI-pilled. Founded in nothing but vibes.
"Don't write prompts like that, do it like this! I swear it's better. Claude says so!"
Try openspec: https://github.com/Fission-AI/OpenSpec/blob/main/docs/gettin...
That was my initial thought when reading the headline but the author states they didn't know it existed before doing this project and critiques it.
Indeed I have a lot of catch up to do, will spend some time with the popular tools before I go too much further down this road.
Yesterday I heard about lat.md [1] which seems to have similar ideas about annotating code with spec refs. I now need to try them both.
[1] https://www.lat.md/
Oh cool, thanks for sharing. LLMs tend to adopt these patterns on their own, so I'm sure my ideas are far from unique.
Agents.md actually scales and outperforms alternatives so I would be skeptical about overcomplicating the proces https://vercel.com/blog/agents-md-outperforms-skills-in-our-...
Small advice - make one repo “main” and link to it from the website instead of an organisation.
I wanted to star the project to track the progress but it feels a bit weird.. Which repo shall I track? Server? Cli? Sounds like a misc repos.
Ooh thought I did that, thanks.
There are also Architectural Decision Records (ADRs), which might be something similar. https://adr.github.io/
some folks also refer to these as PRODUCT decision records
> We are entering the post-slop era. My software is more robust, better tested, better integrated, and more observable than ever before. And my velocity keeps increasing!
Don't we just love the hard fact conclusions based on sample size N=1 and hand-waving arguments?
They’re at the forefront of the industry. Catch up, slowpoke!
Yeah. Waiting to be left behind...since 2020...
I'm tired, boss.
This industry has become a parody of itself, and people are celebrating.
It's ok friend, all I did is put acceptance criteria in a list so I can parse it and quickly track cross-references. The rest is just Elixir/Phoenix and some creative writing.
I also have started numbering my Acceptance criteria and pushing that across the team(s). It’s going pretty well. Some note however are
1. Don’t write in yaml. It’s really hard for humans. Write in markdown and use a standard means to convert to lists / yaml.
2. Think beyond you writing your own specs - how does this expand into teams of tens or more. The ticketing system you have (jira? Bugzilla) is not designed for discussion of the acceptance criteria. I think we are heading into a world of waterfall again where we have discussions around the acceptance criteria. This is not a bad thing - is used to be called product management and they would write an upfront spec.
If this new world of a tech and a business user lead the writing of a new spec (like a PEP) and then then AI implements it and it’s put into a UAT harness for larger review and a daily cycle begins, we might have something.
Good luck
YAML is one of the worst technologies ever invented, it has more warts than features. One of the benefits of LLMs is that they can write YAML for me, wherever I am forced to use it.
Otherwise, I like the idea of machine-readable specs.
the token usage isn’t sustainable. formal english is a barrier but requirement for specification. brevity is the language of money and that’s the premise of management using ai.
fyi language alone can’t define/describe requirements which is why UML existed.
Natural language is a fully general system and can define and describe everything.
You could deterministically process any UML diagram into a prose equivalent.
And in fact you couldn't do the other way around (any prose -> UML) because UML is less powerful than natural language and actually can't express everything that natural language can.
> can define and describe everything.
Can it also fully describe a composition by Bach or a Rembrandt's painting? In some weird, overly complex way it probably 'could', but it would be very painful. That's why we pick other forms of expression. We use other forms of expression to compact and optimise information delivery. Another benefit is that we cut out the noise. So yes UML cannot describe everything natural language can, but then again why should it - it was designed as a specific framework for designing relations between objects. Not more and not less. Similar for sequence diagrams or other forms of communicating ideas efficiently.
There are also diagram notation languages and LLMs are happy to both consume and produce e.g. Mermaid.
I think uml exists to help humans understand and communicate specifications, not because language alone is insufficient.
I mean, if you can't agree on what UML is, then what hope do you have to agree on what the spec says?
ugh with the "maxxing" everything
Could it be that slop PRs are less frequently rejected/commented due to (unfortunate) increased acceptance of it? As it turns out when maxxing AI on leaf parts of a program, the quality of the code doesn't matter that much anymore when compared to building the fundament.
Anything [prefix]maxxing just sounds so bad. It just feels so Andrew Tate...
More nonsense buzzword soup de jour. Can I play along at home? How about Vibewatermaxxing? Surely in the new age it should catch on.
This industry is just getting more and more bonkers.
What is it with people and procrastinating with the most useless shit you can imagine?
First it was choice of editor: people were micro optimizing every aspect of their typing experience, editor wars where people would literally slaughter over suggesting another camp.
Editor wars v2: IDEs arrived and second editor war began.
Revenge of the note taking apps: Obsidian/Roam/Joplin/Apple Notes/Logseq. Just one plugin, just one more knowledge graph, bro, and I’ll have peak productivity. 10x is almost here.
AI: you’re witnessing it now.
Do people NOT have anything else in life? How are y’all finding time to do all of this shit? Are you doing it on company time? Do you have hobbies, do you learn foreign languages, travel, have kids or spouses, drive a car, other thousand “normie” things outside of staring at the freaking monitor or thinking about this shit 24/7? Did I miss the invention of a Time Machine?
Lmfao. Going to a site for computer geeks and complaining that they are computer geeks.
Also, a lot of folks don't write code anymore, and barely have the time to read the volume of code that AI produces. This may just be one of the most profound changes in an industry, and some folks are excited about it and want to get better at building with it.
I think the person who wrote this post made a good faith effort to share his learnings while promoting his tool.
A lot of people sadly, nowadays don't have anything resembling to a social life and family... So here we are now, specmaxxing and shit :)
It's fun how people brag of their agentmaxxing, but if you ask them what those agents are busy actually producing, it's invariably another agent harness so they can agentmaxx better. NFT/blockchain ecosystem was much the same.
I think people find joy in trying to optimise (maxxxxxx) their setup be it editor AI note taking etc. They make time for it
>Do people NOT have anything else in life? How are y’all finding time to do all of this shit? Are you doing it on company time? Do you have hobbies, do you learn foreign languages, travel, have kids or spouses, drive a car, other thousand “normie” things outside of staring at the freaking monitor or thinking about this shit 24/7? Did I miss the invention of a Time Machine?
How are any of those things even remotely as interesting as arguing with people about an Emacs config?
If you have ever been to car forums, it's quite the same there.
People are people.
The author is right but his message ain’t specsmaxing, because while somewhat understandable as a rationale what does it actually mean?
In other words: specs can be as detailed as it gets, and this is why developers have a hard time when they face as a senior an NDAed regulated environment. It ain’t software craftsmanship but data flow, hardware components, compliance on the lowest level including supply chains often times, information architecture - a simple app needs to comply to specs that amount to thousands of pages.
Context window: circular reference. A year ago? Specsmaxing by really weeding out any redundant words. Today? Yawn, like with 8mb RAM vs 512 Gigabytes.
AI wants to be easy on us so what is a spec anyway then?
To put it this way: the spec for the spec is constantly evolving.
Last year’s prompts lead to extremely different results today no matter how maxed out.
The author was on point with his introduction: AI is as junior in many ways when it comes to any sort of efficiency and optimization.
This is my revaluation after years of experimenting with AI. Beautiful code, sophisticated but performance wise and its architecture are laughable at best.
AI is not trained on optimization. Not the slightest and juniors have no clue about algorithms and Big O.
In fact Google used Big O as a basic entry level interview question for a very long time. They have to but the simple fact that in my experience 99% of devs never heard or consider it speaks volumes.
AI cannot compensate for that (yet).
I went the opposite and my specs focus heavily on architecture and the obvious dumb performance drains noobs do.
Google was mocked about Big O. And yes, failing to understand that Big O can be neglected thankfully in 99% of cases is part of its logic.
AI bloats your code. And a year long single dev project gets pumped out in hours. In short: a homerun for Big O because it looks on results that change depending on the variables. A function in mathematical terms.
So I think the author did a funny and great job of you focus on Big O if needed. Everything else is not that important because of being open to change and extension.
Big numbers need great architecture.
It screams loudly. And also think about leaks. Before AI I had virtually no memory leaks at all. Since AI NodeJS and React are worse leaking compared to IE 6 and 8. I mean it.
Big O reduces them significantly, so don’t work around the Elephant in the room.
Architecture and optimization is brutally hard. Google blew my mind in this regard but this is another story of squeezing out even milliseconds out of a build tool used by all. A single dev laughs at it but failed the calculation as well as abstraction.
Grindmaxxing, a long form blog post that is actually just an advertisement for his website.
A tried and true content marketing strategy. The 100+ upvotes suggest he's doing something right.
Should I apologize for being excited about something I built and use daily and for wanting people to try it, discuss it, critique it? Not sure by the tone of your message.
Don't apologize. Keep writing and trying things. Ignore the haters and non-curious, listen to the (even if salty) interested.
There's a fair amount of talk right now about the value being in the verification layer -- once there's a hard verification loop, the agents can do amazing things without getting (permanently) sidetracked. I think what you're working on is half way there -- in essence, you're probably relying on the LLMs notion of what a spec is and should be to the codebase.
What's not currently solved, and what I think is very interesting is how much automation can be added to the creation of verification. We all would unlock a lot more speed and productivity for even moderate gains on that side.
Read the room. What you "built" is neither exciting, nor something most people want to "try". Why? Because just like other AI boosters, you are still trying to somehow optimise the usage of natural language to make it work. But it will never "work" because the way the stochastic ML system is built, it has a failure built into the system.
Totally agree it's not exciting, even though I am personally excited by it, and I also agree it's not something most people want to try, even though some people do want to try it-- and I found a few of them right here on HN.
Disagree on the bit about it "never going to work" though.
Failure-prone stochastic ML systems produce testable, auditable code... just like failure-prone human brains can produce testable, auditable code. And in fact, in both cases, changes to our process can reduce the amount of failures that slip past testing and audit. Or can reap other rewards. Finding the a better process is what I'm interested in right now.
No need to apologize, just don’t act surprised when people call you out.
[dead]
[dead]