Oracle, the law firm with a tech business attached, probably wants to retain the option to sue other people for AI-washing their proprietary code, and that doesn't work if they're also publicly accepting AI contributions to their code with no apparent concern for the provenance of it. The tech business would probably tell them this isn't going to be a needle they are going to be able to successfully thread, but the law firm is in the driver's seat.
"Do not fall into the trap of anthropomorphizing Larry Ellison. You need to think of Larry Ellison the way you think of a lawnmower. You don't anthropomorphize your lawnmower, the lawnmower just mows the lawn - you stick your hand in there and it'll chop it off, the end. You don't think "oh, the lawnmower hates me" - lawnmower doesn't give a * about you, the lawnmower can't hate you. Don't anthropomorphize the lawnmower. Don't fall into that trap about Oracle." --Bryan Cantrill
The flaw in the analogy is that oracle is made of people. It's not a lawnmower, it's a group of people with human motivations and human thoughts. It's true they don't HATE people, but it's also true they absolutely have no conscience in negotiations and have zero concern about relationships.
Maybe so, but perhaps the most successful part of the AI marketing pitch has been exploiting management's hatred of labor (and vice versa).
A significant motivator for a lot of the nonsensical AI layoffs and initiatives over the last three years has been that it's a great opportunity for management to get ill on their slaves.
Tech labor got a little too big for their britches during the hiring spree of 2021, and AI was a great opportunity to take them down a peg, even if it offered little in the way of ROI.
How many bureaucracies in history have behaved in deeply misanthropic ways, especially when led by misanthropic leaders? Nazis/Gestapo ("I was just obeying orders!"), the KGB, East German Stasi...
What about a guy with a lawnmower intentionally chopping off your foot? And what about a CEO making decisions on behalf of the corporate entity called Oracle?
It’s been established that LLM-generated code is not copyrighted so I can fully understand the company living from copyrighted data to not accept LLM-generated contributions.
> It’s been established that LLM-generated code is not copyrighted
If that's a reference to Thaler v. Perlmutter, the only thing that's been established is that an LLM can't be considered an author under the Copyright Act, only a human being can. It says nothing about the consequences of a human claiming authorship of LLM-generated code, which would be relevant here.
I highly doubt that that's the last word on that matter, but even if: Even before LLMs you could combine individual non-copyrighted components into something copyrighted.
This is a false narrative based on a (IMO often intentional) misunderstanding. It has by no means been established by any court that LLM-generated code is not copyrightable.
Thaler v. Perlmutter stands for a much narrower proposition and at any rate is not binding nationally, SCOTUS having denied certiorari.
I didn't know openjdk was developed by Oracle. I always thought it was a community implementation of java as opposed to Oracle's proprietary implementation.
I kind of get their overall point, even if it is ironic that Oracle is all in on AI. They don’t want a ton of contributions to review that weren’t made with care, adding burden to, as they put it: “already limited time of human reviewers”.
Especially for a project that runs to many major businesses, this could pose a massive risk.
I understand the intent of it. It's really annoying/rude when someone gives you a pile of AI generated crap and says "here, you figure it out." Anyone can type these prompts now, the hard part is figuring out if the output makes sense.
Seems like a better way to do this than ban all AI generated code, but it's probably a matter where they're far more concerned about their own inconvenience than someone else's.
Ghostty has what I’d consider to be a reasonable policy. You submit what you want to change and how, explain it in your own words. A trusted member can vouch for you. You can then make a PR, but you need to explain the solution and how it works in your own words.
I’m sure someone could still just prompt their way through but it’s a decent soft gate IMO.
Really, I think the long term solution is people's attitudes need to catch up with the technology.
Humans still need to review code for taste, making sure AI is writing sensible, well organized output. That's a core "AI skill" for now, at least until it gets better...
If ai rapidly gets better, then instead of submitting prs you can ask your agent to fork and improve the project and bypass the policy that way. if it goes slower they have time to reconsider policies in the future. if it doesn't improve then we can probably find out some middle ground approach like the ghostty one
in any of those cases there's no reason to get swamped by cheap PRs for now
> the hard part is figuring out if the output makes sense.
I think for Java specifically, it's also very much: which prompts make sense to write. Java has had a storied history of features that made future development hard. That's resulted in a culture where they like to think quite deeply about how to add a feature BEFORE they sanction any sort of programming.
This is the firm whose debt is one level above junk and whose survival as a single unbroken corporate entity is fully predicated on OpenAI being able to make money from the data centres they (Oracle) are building for them, let alone simply paying for them.
I accept @jerf's explanation is the right one but this is an absolutely almighty signal that AI-first developers should heed. And someone should ask Altman about it on the record.
So contributions to OpenJDK have to be hand-written, while internally, Oracle uses AI generated code... I guess the logic is that employees who use AI generated code (theoretically) review their code better? The whole thing still seems a little strange.
No, their internal PRs have the same problem, but they are pushing an AI narrative from the c-suite and can hide the reality of their private code bases.
The "limited ROI on AI investment" articles will continue to percolate slowly into the brains of the LinkedIn hive-mind until we hit a tipping point, and then we'll finally shut up about how a handy dev tool with some decent use-cases is the dawning of the singularity that will replace all white collar labor and get back to actually building business value.
Let me guess: AGI is 2-6 weeks away, software development is solved, and anyone who disagrees with your enlightened perspective is a backwards-looking luddite.
You can get your coworkers to stop sending you slop (I've done it already on my team) but it's probably much harder to stop an onslaught of open source contributors.
Nonsense, it's easy: you just need a social process like Linux has. Patches are ignored unless you've proven yourself to somebody. Start by participating in technical discussions on IRC or the smaller mailing lists; do that for a while before you submit patches. For smaller projects: "if I don't know you, I don't merge you".
Linux's LLM policy allows AI-generated code because the project was never vulnerable to this DoS attack in the first place. They kept the social defense they've had for decades.
The killzone here is Github culture, where people decided it was normal to accept code from anonymous randoms with anime avatars. They're doomed.
I believe thats the real reason, its not just Oracle, it would eventually be everyone in this kind of boat. No business would publicly use (or claim) AI to write code with recent verdict over the IP rights of AI generated code.
It would be interesting to see how this effects the copyleft Licences with contributors are using AI for the PRs
> AI to write code with recent verdict over the IP rights of AI generated code.
Which recent verdict?
> It would be interesting to see how this effects the copyleft Licences with contributors are using AI for the PRs
I know Linux and GCC have been diligent about tagging and tracking LLM based contributions. In the worst case they can chuck it all out and handwrite it back.
There is this old one around contents like images and all, the ripple affect is pretty much everywhere.
Its sort of coupled with the recent penalty of $1.5B on Anthropic. The catch 22 is that some portion of the LLM training data can be classified as IP theft. Even though Anthropic has been fined the data still exists and can be used by LLM to generate code for you. So if you are claiming something as an IP and it has stolen part in it then it leaves you in a hard place.
Language translations may save you in some cases, though the whole definition of cleanroom has been in debate recently too where people are trying to rewrite opensource/famous libraries in different language and claiming IP rights over them.
Anthropic will have judged the benefits of the settlement, not just the headline cost. It could have been a strategic move by Anthropic: we can't know without information we don't have. https://news.ycombinator.com/item?id=49014389 1.5B looks like ~2% of funding/income.
This constitutes an admission on the part of one of the largest AI investors that AI generated/"assisted" code is inadequate (in part) because it is produced in flagrant violation of license rights.
Using AI is vanity? I've never heard that before. Why is that?
Refusing to use AI honestly seems more like vanity to me. Like "no machine could ever do what I do".
That said I would not necessarily accuse people who don't want to use AI of vanity - there are good reasons not to want to use it - but if someone else did it I'd find it easier to understand.
yeah, oracle is literally betting the future of the company on OpenAI owning the majority of the AI market, so it's pretty strange to see such a ban. Their debt is already pretty much junk (and I'm not exaggerating, it is rated BBB- by s&p)
The Oracle antipathy is less about the policy itself and more about the apparent hypocrisy of hyping and funding these tools as "solving programming" and then forbidding their use on their own open source project.
> I feel like that was the epitome of AI assisted but quality code
I totally agree with you on this! The engineer is still actively engaging while getting the performance gains of not having to type out functions. The Engineer is still in charge vs agentic "engineering" a model + harness can spit out whatever and the engineer is left to review tons and tons of code
How does all this AI generated code blocking gappen? Does a human decide that or an AI is tasked with sequencing that PR's DNA to ascertain its possible inhuman origins?
What if someone generates code with AI and then goes into the IDE and then bathes it, dresses it (including adding comments) etc, in a way a human would? What then? No, here I am not exploring a way to fool the code DNA checking, but rather trying to find out what the real problem is with the AI generated code? (Other than license issues, too many PRs etc)
They're just so information-dense and sprawl in scope.
Too many PRs wouldn't be a problem if they were easily digestible: But they often include these giant refactors with confusing changes that aren't elaborated.
The real underlying problem is with authors who do not understand the code they submit. If you cannot defend the PR, you shouldn't be submitting it.
The "blocking" happens on the side of the AI users. You like using AI. You are aware of this policy.
Will you submit an AI-generated pull request to OpenJDK? No? Voilà, you have successfully self-enforced the OpenJDK policy.
Or do you go to the trouble of finding an interesting issue to work on, get your agent to code it up, manually polish it to make it less AI-looking in case there are doubts, and then submit it? Why would you? To prove some kind of point, to yourself, that you can never disclose publicly? Most people have better things to do. Voilà, the policy is, again, self-enforced. Enjoy your day on the beach instead of trying to trick a project that is politely asking you not to trick it!
Policies like this will result in the death of the branded software.
As we move forward it will be easier than ever to just maintain and keep your fork of software with the changes you want or need. No more approval, bureaucracy, or arguing. Just tell the AI agent want you want changed and you have it.
This will be used for huge things too. Like maybe you want a specific fork of Java that only supports for each iterators, goodby linters, hello compile time error.
You truly envision a future where every program is written in a custom programming language, for a custom operating system, for a single user who will now be in charge of understanding and maintaining it forever? That user being your grandma, your baker, your CEO?
Software as a list of requirements and that's it. The local LLM appliance everybody has taking in a document specifying hardware, interfaces, and requirements and spitting out software changeable locally via conversation with its users.
In the same way you have a cookbook with recipes to make dinner instead of ordering out.
> Software as a list of requirements and that's it. [...] a document specifying hardware, interfaces, and requirements [...]
For that, you'd want the list of requirements and hardware documentation to be written in a precise, formal language. That's no different than writing them in a programming language (though a declarative one, instead of the more common imperative ones).
I've in the past (way before LLMs existed) thought about automatically generating device drivers from hardware documentation. But besides the need for very precise documentation, hardware never works exactly as documented; a human-written device driver can avoid problematic areas (perhaps even by accident), while a computer-written device driver would end up exploiting every corner case of the documentation.
Most people don't bake their own bread even if it's more approachable, and cheaper then vibecoding. Bread is just a recipe one may say. But in modern times have we ever witnessed disappearance of specialization? I don't think so.
I have my doubts about this. Brands are a about reputation, and reputation strongly affects responsibility when business decisions are made. If a manager decides to vibecode a solution which eventually fails, said manager will be punished. If instead a mainstream software is bought, and it fails - well, everybody has the same trouble around, right? It's not a personal mistake anymore. I suppose there are niches where branded software may give way, but I don't think it's gonna be a general trend.
> easier than ever to just maintain and keep your fork of software with the changes you want or need
I don't imagine this is practical or desirable for all situations. Good software is built from being battle tested by many users in many environments. Even with the advancements in AI tools, I don't imagine they'll become omnipotent anytime soon.
> No more approval, bureaucracy, or arguing
For software that can kill people or substantively affect someone’s life in a negative way, the bureaucracy is there for good reason. I don't think anyone should want someone at Phillips to vibe code the control software for an X-Ray machine or an employee at CrowdStrike vibe coding the next update before pushing it out to millions of machines.
We are forced to endure low-quality software because there is little or no accountability. I can only imagine what you propose would make an already poor situation worse.
> maintain and keep your fork of software with the changes you want or need
Then you'll have the same problem everyone who forks a piece of software ends up having, sooner or later: as the original evolves, keeping your fork up to date with the upstream changes becomes harder and harder. The bigger and more invasive the changes are, the harder synchronizing with newer releases become.
This just reflects how absolutely clueless you are about the amount of attention to detail that the OpenJDK folks put into developing the language and ensuring that it works for all its users (which are serious users delivering actual value.) And I say this not even being a Java programmer myself, just an envious C++ dude watching from the sides.
Someone brings this up on just about every AI-related thread. I think it's nonsense. Nobody wants to maintain a fork of any remotely complex software, not even with AI. And in a corporate setting, nobody wants to use your custom fork; they just want to use the standard software they already know with the quirks they've already learned.
This is far more likely to happen in a corporate setting with competent engineers who want to build/solve now versus relying on others. I've seen it 100% of the time with varying results.
Except. . . (accept)... SaaS solves the problem of needing one piece of software to communicate between multiple (many) users and/or other pieces of software.
Can't picture a functioning world where every piece of software is custom and requires factorial amount of AI comparisons and reviews to patch the API to communicate. In fact, it's impossible! There's not enough compute to handle a factorial explosion.
Given how much of the industry is SaaS that would be such a self own! Wow, fantastic, you can build all your software in house, zero dependencies. Wait, your users can do the same? And they don’t need to pay you anymore for any of your cool services because they just asked their agents to recreate your infra from scratch? Interesting, truly the future of humanity
> Wow, fantastic, you can build all your software in house, zero dependencies. Wait, your users can do the same? And they don’t need to pay you anymore for any of your cool services because they just asked their agents to recreate your infra from scratch?
The cherry on top is that OpenAI and Anthropic brainwashed your coworkers and your company's C-suite into uploading the entirety of the "proprietary" codebase onto their servers thousands of times per day over the last three years.
I think it's even more pervasive than that. Why bother forking software at all? At the point in which code generation is meaningfully trivialized, software becomes entirely disposable. Anything you want, have a model spin it up. You don't even need libraries, the model can just make everything in-situ, who cares? Why on earth would I ever want to use SQLite if I have access to a sufficiently advanced code generator which can generate me a similarly high quality database system, with the added benefit of conforming to whatever my problem domain is, conforming to whatever branch of database theory I want?
Even SaaS isn't safe. I don't even have to describe your product to my system, I just have to give it a harness with access to the interface and have it replicate it locally. Frankly you can probably already prompt for that.
The only thing holding this future back right now are pricing problems and code generation quality. Both of those barriers are constantly being knocked down. We might never arrive at that future, but it's definitely a higher probability than solving AGI's scaling issues, and would arrive much sooner for technical users.
> Why on earth would I ever want to use SQLite if I have access to a sufficiently advanced code generator which can generate me a similarly high quality database system
Because SQLite has 10k requirements that wouldn't even cross your mind to write down, but 80% of which are useful to you.
>Why on earth would I ever want to use SQLite if I have access to a sufficiently advanced code generator which can generate me a similarly high quality database system
https://www.youtube.com/watch?v=V_qzqY1bb7I your sufficiently advanced code generator may generate you a high quality database system for some measure of quality, but it will not have SQLite's reliability over the extremely long tail of edge cases proven through its testing and use in real life
Then you've failed the criteria of sufficiently advanced. It's perfectly fine to cast doubt we'll see scaling to this generalization, but you're not casting doubt you're outright rejecting the premise in-confidence. It betrays that you have no idea what you're talking about. May I see your quantification of this long tail? Do you even know how to formalize the mapping from n-bit precision of weights and/or activations to the standard deviation of a transformer's output distribution, such that we could decide whether the long tail of a given behavior is unreachable? Something tells me that no, you don't know how to do that in the slightest. So what drives you to speak with such confidence?
That's before we get into the entire non-linearity of agentic systems introducing massive decidability problems on this in the first place. A little bit of epistemic humility please.
Oracle is just selling AI shovels. Of course it would have a hard time defending IP issues in slop code when it had famously sued Google for Java infringement.
Also, Oracle knows that it will be stronger with strong IP laws. Other companies will realize that fact soon.
So Java, which powers a sizeable chunk of the real-world since two decades now, without needing the help of LLMish sloppy-pasta, is going to continue to power the Real World [TM] with its language and JVM that have been created with tens of millions of human man-hours without LLMish sloppy-pasta?
I'm shocked.
I'm not only shocked: I also see there a delicious irony in the countless of LLMish sloppy-pasta code that's now running and going to be run on man-made JVMs.
> countless of LLMish sloppy-pasta code that's now running and going to be run on man-made JVMs.
Don’t worry, I think most LLMs, when starting a greenfield codebase, don’t reach for Java, so it may be a smaller amount of code affected than you feared.
I would laugh for a week if the courts eventually found issues with using LLM content and forced companies to either publish their LLM generated code as public domain or they were forced to roll back their codebases to a pre-LLM state.
Probably won’t happen but clearly Oracle sees a potential for legal issues with LLM output.
Thing is public domain can be used any way you want. The Libre style licenses are specifically engineered to make the source available to the users of the source, no such clause exists in using public domain.
Oracle is not a software company, but a litigation company. It will be hard to sue people because LLMs generated code similar to some code copyrighted by Oracle if they themselves use LLMs internally to generate code.
Probably it would get picked up by someone else (e.g. the Linux Foundation)? It's GPLv2-licensed, so it's not like the code is going to be risky for someone to fork
I don't think Oracle will die, but even entertaining the hypothetical I suspect that some other organization (e.g. Apache) would pick up and maintain a fork.
IBM, Red Hat, Microsoft, SAP, Amazon and others all work on the JDK. They would have to step it up to cover the loss of Oracle's contributions, but I am sure they could manage.
Regardless, your timeframe on Oracle's death seems way too soon.
Hilarious thing is, if I want a JDK, I'll ask Fable to write one, get Sol to review it and write tests for it, and it will be working by Monday morning. No need to get permission or anything else from Oracle. Building a complex product from many smaller, individually-testable components is exactly what LLMs are good at, and a Java SDK certainly qualifies.
And if I run out of tokens, well, that's OK too. K3 running on my own box will finish the job. I'll just have to wait another week, that's all.
If I want a JDK, I'll ask my package manager to download and install one, and it will be working a minute later. Why would you want to wait another week for something that already exists?
Please do this. I would love to see this timeline actually hold true. Production quality is the bar though. Not something that only works in a very narrow set of circumstances with a ton of caveats.
Maybe so, but perhaps the most successful part of the AI marketing pitch has been exploiting management's hatred of labor (and vice versa).
A significant motivator for a lot of the nonsensical AI layoffs and initiatives over the last three years has been that it's a great opportunity for management to get ill on their slaves.
Tech labor got a little too big for their britches during the hiring spree of 2021, and AI was a great opportunity to take them down a peg, even if it offered little in the way of ROI.
If that's a reference to Thaler v. Perlmutter, the only thing that's been established is that an LLM can't be considered an author under the Copyright Act, only a human being can. It says nothing about the consequences of a human claiming authorship of LLM-generated code, which would be relevant here.
Thaler v. Perlmutter stands for a much narrower proposition and at any rate is not binding nationally, SCOTUS having denied certiorari.
https://www.theregister.com/ai-and-ml/2026/08/03/as-larry-el...
The register article is about this post:
https://openjdk.org/legal/ai
It is 'OpenJDK Interim Policy on Generative AI' and their lawyers are writing the final version, according to this page.
It sounds like a sensible action, given past scars around Java and copyright, plus this is from a big old corp.
That said I personally don't expect that final proposal will end up any better.
How many of these were self-inflicted?
Especially for a project that runs to many major businesses, this could pose a massive risk.
Seems like a better way to do this than ban all AI generated code, but it's probably a matter where they're far more concerned about their own inconvenience than someone else's.
I’m sure someone could still just prompt their way through but it’s a decent soft gate IMO.
Humans still need to review code for taste, making sure AI is writing sensible, well organized output. That's a core "AI skill" for now, at least until it gets better...
in any of those cases there's no reason to get swamped by cheap PRs for now
I think for Java specifically, it's also very much: which prompts make sense to write. Java has had a storied history of features that made future development hard. That's resulted in a culture where they like to think quite deeply about how to add a feature BEFORE they sanction any sort of programming.
The same problems that bedevil AI-written PRs here bedevil AI-written PRs at my place of work.
I'm sure it'll get better eventually, maybe, but something something Pareto principle.
>I'm sure it'll get better eventually, maybe
Why would it get better?
Selling it is one thing. Having to use it one's self is quite another.
Do as Larry does --- not as Larry claims to do.
I would short his stock but it has already lost half it's value over the past year. And his credit rating is one notch above junk.
AI is going to make fools out of a lot of billionaires like Larry.
Even the biggest institutional AI investors have to ban their contributors from spamming them with AI generated slop.
I accept @jerf's explanation is the right one but this is an absolutely almighty signal that AI-first developers should heed. And someone should ask Altman about it on the record.
The "limited ROI on AI investment" articles will continue to percolate slowly into the brains of the LinkedIn hive-mind until we hit a tipping point, and then we'll finally shut up about how a handy dev tool with some decent use-cases is the dawning of the singularity that will replace all white collar labor and get back to actually building business value.
Linux's LLM policy allows AI-generated code because the project was never vulnerable to this DoS attack in the first place. They kept the social defense they've had for decades.
The killzone here is Github culture, where people decided it was normal to accept code from anonymous randoms with anime avatars. They're doomed.
It would be interesting to see how this effects the copyleft Licences with contributors are using AI for the PRs
Which recent verdict?
> It would be interesting to see how this effects the copyleft Licences with contributors are using AI for the PRs
I know Linux and GCC have been diligent about tagging and tracking LLM based contributions. In the worst case they can chuck it all out and handwrite it back.
There is this old one around contents like images and all, the ripple affect is pretty much everywhere.
Its sort of coupled with the recent penalty of $1.5B on Anthropic. The catch 22 is that some portion of the LLM training data can be classified as IP theft. Even though Anthropic has been fined the data still exists and can be used by LLM to generate code for you. So if you are claiming something as an IP and it has stolen part in it then it leaves you in a hard place.
Language translations may save you in some cases, though the whole definition of cleanroom has been in debate recently too where people are trying to rewrite opensource/famous libraries in different language and claiming IP rights over them.
Settlement not penalty.
Anthropic will have judged the benefits of the settlement, not just the headline cost. It could have been a strategic move by Anthropic: we can't know without information we don't have. https://news.ycombinator.com/item?id=49014389 1.5B looks like ~2% of funding/income.
Refusing to use AI honestly seems more like vanity to me. Like "no machine could ever do what I do".
That said I would not necessarily accuse people who don't want to use AI of vanity - there are good reasons not to want to use it - but if someone else did it I'd find it easier to understand.
code is a liability, and they likely have more to lose than gain by allowing AI contribution.
It's shocking given Oracle's AI spending spree over the last couple of years.
If you know how to work with it it's great, but if you use it indiscriminately it does more harm than good.
The biggest use of Java currently is Android, and Google is on a path to migrate to Fuchsia eventually.
The second biggest use is Apache Spark/Kafka, but with LLMs now, you can pretty much re implement all of that functionality in pure C.
They just want a legal angle, since this is how they make most of their money.
There is also the approach of Microsoft putting AI all over the place on .NET, and CoPilot driven development all over the place.
> If I use a generative AI tool to create 100 lines of code, and then edit ten of those lines myself, may I contribute the result?
> No. Your contribution would still include, in part, AI-generated code.
I totally agree with you on this! The engineer is still actively engaging while getting the performance gains of not having to type out functions. The Engineer is still in charge vs agentic "engineering" a model + harness can spit out whatever and the engineer is left to review tons and tons of code
What if someone generates code with AI and then goes into the IDE and then bathes it, dresses it (including adding comments) etc, in a way a human would? What then? No, here I am not exploring a way to fool the code DNA checking, but rather trying to find out what the real problem is with the AI generated code? (Other than license issues, too many PRs etc)
Too many PRs wouldn't be a problem if they were easily digestible: But they often include these giant refactors with confusing changes that aren't elaborated.
The real underlying problem is with authors who do not understand the code they submit. If you cannot defend the PR, you shouldn't be submitting it.
Will you submit an AI-generated pull request to OpenJDK? No? Voilà, you have successfully self-enforced the OpenJDK policy.
Or do you go to the trouble of finding an interesting issue to work on, get your agent to code it up, manually polish it to make it less AI-looking in case there are doubts, and then submit it? Why would you? To prove some kind of point, to yourself, that you can never disclose publicly? Most people have better things to do. Voilà, the policy is, again, self-enforced. Enjoy your day on the beach instead of trying to trick a project that is politely asking you not to trick it!
sips tea
As we move forward it will be easier than ever to just maintain and keep your fork of software with the changes you want or need. No more approval, bureaucracy, or arguing. Just tell the AI agent want you want changed and you have it.
This will be used for huge things too. Like maybe you want a specific fork of Java that only supports for each iterators, goodby linters, hello compile time error.
Software as a list of requirements and that's it. The local LLM appliance everybody has taking in a document specifying hardware, interfaces, and requirements and spitting out software changeable locally via conversation with its users.
In the same way you have a cookbook with recipes to make dinner instead of ordering out.
For that, you'd want the list of requirements and hardware documentation to be written in a precise, formal language. That's no different than writing them in a programming language (though a declarative one, instead of the more common imperative ones).
I've in the past (way before LLMs existed) thought about automatically generating device drivers from hardware documentation. But besides the need for very precise documentation, hardware never works exactly as documented; a human-written device driver can avoid problematic areas (perhaps even by accident), while a computer-written device driver would end up exploiting every corner case of the documentation.
I don't imagine this is practical or desirable for all situations. Good software is built from being battle tested by many users in many environments. Even with the advancements in AI tools, I don't imagine they'll become omnipotent anytime soon.
> No more approval, bureaucracy, or arguing
For software that can kill people or substantively affect someone’s life in a negative way, the bureaucracy is there for good reason. I don't think anyone should want someone at Phillips to vibe code the control software for an X-Ray machine or an employee at CrowdStrike vibe coding the next update before pushing it out to millions of machines.
We are forced to endure low-quality software because there is little or no accountability. I can only imagine what you propose would make an already poor situation worse.
Then you'll have the same problem everyone who forks a piece of software ends up having, sooner or later: as the original evolves, keeping your fork up to date with the upstream changes becomes harder and harder. The bigger and more invasive the changes are, the harder synchronizing with newer releases become.
Can't picture a functioning world where every piece of software is custom and requires factorial amount of AI comparisons and reviews to patch the API to communicate. In fact, it's impossible! There's not enough compute to handle a factorial explosion.
I really doubt SaaS going anywhere.
The cherry on top is that OpenAI and Anthropic brainwashed your coworkers and your company's C-suite into uploading the entirety of the "proprietary" codebase onto their servers thousands of times per day over the last three years.
Even SaaS isn't safe. I don't even have to describe your product to my system, I just have to give it a harness with access to the interface and have it replicate it locally. Frankly you can probably already prompt for that.
The only thing holding this future back right now are pricing problems and code generation quality. Both of those barriers are constantly being knocked down. We might never arrive at that future, but it's definitely a higher probability than solving AGI's scaling issues, and would arrive much sooner for technical users.
Because SQLite has 10k requirements that wouldn't even cross your mind to write down, but 80% of which are useful to you.
https://www.youtube.com/watch?v=V_qzqY1bb7I your sufficiently advanced code generator may generate you a high quality database system for some measure of quality, but it will not have SQLite's reliability over the extremely long tail of edge cases proven through its testing and use in real life
That's before we get into the entire non-linearity of agentic systems introducing massive decidability problems on this in the first place. A little bit of epistemic humility please.
Also, Oracle knows that it will be stronger with strong IP laws. Other companies will realize that fact soon.
I'm shocked.
I'm not only shocked: I also see there a delicious irony in the countless of LLMish sloppy-pasta code that's now running and going to be run on man-made JVMs.
Don’t worry, I think most LLMs, when starting a greenfield codebase, don’t reach for Java, so it may be a smaller amount of code affected than you feared.
Prompt: “Top language pick for an Android app? Respond with only the language.”
Responses (admittedly via the aichat CLI, not an agent harness; and temp=0):
Kimi K3: Kotlin
GPT 5.6 Sol: Kotlin
Claude Fable 5: Kotlin
DeepSeek v4 Flash 0731: Kotlin
Are there any other “inherently” heavily JVM-biased domains?
I think even in Minecraft game mods Kotlin is possible nowadays…
Probably won’t happen but clearly Oracle sees a potential for legal issues with LLM output.
Regardless, your timeframe on Oracle's death seems way too soon.
And if I run out of tokens, well, that's OK too. K3 running on my own box will finish the job. I'll just have to wait another week, that's all.