radlad

Are we reading the same Wikipedia page?

> An operating system is difficult to define,[7] but has been called "the layer of software that manages a computer's resources for its users and their applications".[8] Operating systems include the software that is always running, called a kernel—but can include other software as well.[7][9] The two other types of programs that can run on a computer are system programs—which are associated with the operating system, but may not be part of the kernel—and applications—all other software.[9]

> You seem to misunderstand how VMs work. They are not OSes running on top of OSes, they are multiple OSes sharing the same hardware resources simultaneously side by side.

This is not strictly true. VirtualBox, VMware Workstation, QEMU run guests on the host. Even KVM operates in tandem with the host kernel. Maybe you're referring to Xen and ESXi VMs?

> Inferring that you think soon LLMs will either not need operating systems, or will become them.

You removed the key point of context:

> I almost do not need a traditional OS anymore. [...] it's the way I access my LLM. For now.

I believe we will move to something that resembles microkernels and/or thin clients. You'll still have something here, but it'll be about as impactful as BIOS is.

> This is the main reason you are getting backlash and being downvoted.

FWIW, I've received a single downvote.

"The definition"? Whose? There are many definitions of "operating system" and I think it may be you who has it wrong: https://en.wikipedia.org/wiki/Operating_system#Definition_an...

Having a kernel, drivers, or firmware does not disqualify an operating system focused on LLM-usage from being an "agent operating system."

Do you agree that Android is an OS, or would you say that the OS is Linux and Android is simply a distribution? Is the operating system "Windows 11" or is it Windows NT kernel? Is UEFI an OS? ChromeOS? Language isn't always black-and-white.

You seemed to miss the cloud computing point: are you running one OS, or two? Does one OS running on top of another OS with virtualized hardware count? If yes, then why do you think an agent OS is incompatible with a traditional OS?

> An operating system is not just something that opens apps, it is the thing that lets you type, use a mouse, and see things on a screen. LLMs cannot do any of that, they are an app which runs on an OS.

An OS is all of these things. I never once claimed that an agent operating system would have a kernel which "is" an LLM. This is an absurd interpretation.

They load only when operating on files in that directory. i.e. Lazy loading?

For example, I have one in my schema migrations directory that explains how to write migrations with appropriate concern for forwards- and backwards-compatibility in our deployment model.

I'm obviously not saying that LLMs replace the glue between hardware and software. That would be absurd.

An operating system is just the interface a user has to the programs they run on their computer. Surely you can imagine a world where you boot into a ChatGPT-like interface instead of the desktop OS you are familiar with?

Different computing paradigms already exist: mobile OS sure, but also ChromeOS... heck, install Arch Linux and you're dropped into a shell by default - not a GUI. How is this so different?

And it doesn't even have to replace your existing OS to be your operating system: If you purchase a cloud VM (say from AWS or DigitalOcean), you're likely running an operating system on top of another operating system already.

I'm not advocating for this model. But I'm also not sure it has any more lock-in than choosing an OS today: Windows, macOS, Linux, ChromeOS, Android, iOS.

I already mentioned this:

> This obviously isn't entirely true: [...] it's the way I access my LLM.

It doesn't change my opinion that there's a very real possibility that we move to thin clients or microkernels intended primarily to run chat interfaces (or some other UI that organizes LLM sessions and their outputs) either locally on the hardware, or against cloud computing environments. An operating system is just the interface a user has with their computer.

More to the point - it is possible to run an operating system on top of another operating system: this is what VMs are. I'm obviously not claiming that LLMs replace the glue between hardware and software(?!)

> I don't get asked about access for Chrome at every update.

Is it possible that you clicked "Approve" while the parent clicked "Not now"?

One of my personal issues is that I was using software called "Secretive" which installed a Secretive Agent system service. I later patched and forked it, but reused the name. Despite the app bundle ID being different, and having a different signer, they show up collapsed in the "Login items" panel, with no way to enable one and disable the other: https://i.servepic.com/t5DvVjB0.png

There are also numerous entries for software that has been completely removed from my system for years. I wholeheartedly believe OP's experience.

I think because the idea of an "agent operating system" is starting to feel more and more inevitable. Since I can use AI to:

  - Pull information from external sources
  - Transform and/or augment that information
  - Publish information

I almost do not need a traditional OS anymore. This obviously isn't entirely true: I still use it to interact directly with humans, and it's the way I access my LLM. For now.

That's what MEMORY.md and subdirectory-specific CLAUDE.md does.

Yes, my question is directed around how you are determining "task result." Is this measuring whether the code works, or whether it is maintainable?

As with human-powered coding, we read code far more frequently than we write it. It's worth spending a little extra (time|tokens) during authorship to make future maintenance feasible.

My CLAUDE.md, memories, and skills are all about either (a) adherence to project standards and guidelines, (b) product decisions which impact future code, and (c) instructions on how to prototype and work in my environment.

Removing these instructions would mean more turns with the AI to get the desired result.

'better results' in terms of what though? A benchmark, or code that I would actually click "approve" on in a pull request scenario?

I find it impossible to daydream while focused on traffic.

This is blatantly false based on my usage of Fable and Opus. Though they are much better than they used to be.

As a native English speaker, I worry about picking up LLM-speak simply due to interacting with Claude Code all day.

I'm sorry to hear that's how it's going at your work.

The best engineers hand-edit .git

Edit: But actually, one of my favorite Git explainers is https://wildlyinaccurate.com/a-hackers-guide-to-git/

Considering most teams have only switched to heavy AI use in the past few months, the verdict is still out on this.

That said, I've had lots of success using AI to learn, refactor and clean up codebases.

I notice another trend were a lot of AI naysayers haven't really spent a ton of time getting intimately familiar with AI.

Definitely not - the biggest risk with this increased speed is going full-bore in the wrong direction. A product mindset (and a critical eye to architecture) matters more now than ever.

> I use LLMs for exploration and for review, but I write the code myself. I find it hard to believe why so many engineers try to avoid it. It’s not consuming much of my time. And it’s actually the most enjoyable part.

At my workplace, there is more work to be done than there is engineers, and approximately 2 engineers per service. I can spin off multiple Claude Code instances on unrelated work, steering them occasionally, and then finally reviewing the output. After I have reviewed it, I post it for team review.

You're absolutely right that my depth of familiarity is lesser with this code, but we are absolutely shipping more as a result of increased parallelization.

The bottleneck now is typically reviews - both pre-push and team reviews.

I find project-wide or subpackage-wide (for sufficiently large projects) CLAUDE.md's that document patterns (including for tests) solves this.

/init will make a project-wide one, or you can instruct it to "Create CLAUDE.md in any sub-directory that is sufficiently complex" then modify from there.

I don't expect that would be the case. This is what's called BLUF or Bottom Line Up Front: https://en.wikipedia.org/wiki/BLUF_(communication)

The model will still have read the entirety of the document before composing its response. And I believe that even in auto mode, there are thinking tokens behind the scenes.

I could be crazy, but don't all the external antennas allow for more flexibility with regard to directionality of signal?

I have a U7 Lite and it is very directional compared to other routers I have used (spider style, trash can style, etc.)

Seems kinda weird - it's cognitive load I'd love to avoid. If I'm going to take it on, I might as well try other providers.

I can empathize with this view but I've also seen one line clear bug fixes rejected because CC signed off on it.

There's no other correct fix - why do you care which pen I used to write it?

What fundamentals are those?

This sounds like what regular commit messages do. How are conventional commits specifically helpful?

As I recall, Amazon also famously didn't turn a profit for ages - but they were also capable of turning one much earlier than they did.

Are AI companies capable of turning a profit today if they turn some knobs?

Regarding seed ratio, generally by perma seeding. Many private sites either use seed time requirements instead of ratio or offer bonus points for seed time which can be exchanged for ratio. But also as new editions and formats are released, the library has a bit more turnaround than your music sites of yesteryear.

Out of curiosity, what movie?

> hey there's a project idea: a "todo list" for rippers that scrapes imdb and checks what's not in pirate ba

Private sites do things like this, archival efforts and have request systems.

Don't want to name drop, but this is unfortunately a real phenomenon.