ethbr1

The difference is in enforceability.

Open weight models are less like Napster and more like DeCSS -- once you have the digital artifact, there's little external evidence you're using them.

Napster was easy to target because it was an open P2P network and specific key US individuals.

If the US government banned open weight models tomorrow (national security grounds), they'd already get a lot of pushback, only increasing day by day as more 'less than SOTA' solutions using them are deployed.

The US government could likely enforce this on its own supply chain (military and federal contracts, maybe some state funding) easily enough.

Enforcing it on private companies would be more difficult... maybe they could push that through, but it would likely take Congress to pass a law. And Congress is substantially less enamored with supporting OpenAI / Anthropic / Google / Meta.

And even if that gets pushed through, enforcement is going to be a bitch on smaller companies using non-US clouds.

Given this is HuggingFace, I'd expect that's less about thought leadership and more using what they know well, in a critical situation.

Correct. I just wanted to clarify the statement that parent made, as it seems like lots of people have a misassumption about the copyrightability of autonomous in the United States.

Expect it will be clarified and/or changed by law given how much money is at stake, but the current state is what the current state is.

If I were developing key IP with agents, I'd be very careful to document my human contribution.

There are lots of famous artists that are older than copyright.

There are lots of famous artists that made art before color image capture and reproduction (1930s-60s) and digital image capture and reproduction (90s-00s).

Copyrightless artist fame and economic viability is enabled by a lack of widely-accessible, cheap reproductive methods.

You mean like right now? Here is A24 claiming copyright for Backrooms related media...

Since resolved: https://kotaku.com/backrooms-director-kane-parsons-a24-copyr...

Turns out when you outsource copyright policing to minimum wage folks, ambiguity goes out the window.

Society actively misses out a lot. We could have had tons of derivative art that has been buried for the sake of propping up companies. We could have had Aaron Swartz. Abolish copyright.

The problem with absolutist arguments is that they ignore inconvenient facts.

At a time when creative and art economics is under siege, how would copyrightless art make enough money for the creators?

The fact of the acquisition of large swaths of copyright rights by large corporations does not negate the fact that artists need food (and ideally, a place to live and a way to provide for their family).

There are positions that might enable that (Hey, what if we banned the assignment of copyright to corporations? Human only? Original creator only?), but none of them are stripping all rights from intellectual property.

I'll link to a previous comment of mine: https://news.ycombinator.com/item?id=48968156

You need to have a very specific and 'creative' / 'substantial' expression of an idea for copyright to apply. The output of an LLM can be easily be such, but usually not.

This is incomplete with current US law. You need the above (the typical copyright qualifiers) AND evidence of substantial human involvement in the creation.

Minimally directing an autonomous agent does not qualify.

Because guaranteed to never fail requires some UX inconveniences, and the industry (and the market) doesn't seem in a mood to tolerate inconvenience for the sake of security.

Mac Minis are out of stock because they're the cheapest entry point to reasonably capable local agents (read: 64+GB of GPU accessible RAM), which TikTok is all about.

They're being bought as a hardware solution, not a sandbox solution.

As proof, look at all the people wiring up their personal emails, etc. as soon as they turn them on.

They are, unless governments actually crack down.

A lack of enforcement actions has caused situations like this.

Why would LG do better when it's cheaper to not?

"Our ranking on some metric wasn't as high as we would have liked" seems like a bad reason to self-exclude.

I'd offer that the AI harness ecosystem is discovering the same thing any rapidly-growing software space did before it -- lines of code are a maintenance burden.

That's not something a team sees when they're doing greenfield, forward development. (Most projects, now)

It only crops up when they have enough on-the-books code that features / bugfixes start requiring introspection of existing code. (Now to ~1 year from now)

And a lot of these projects don't seem to be approaching codebase size growth from a future code debt perspective...

Maybe the important take-away would be to recognize that this was a rant.

Ownership and honesty about things is healthy, as is letting out strong opinions sometimes.

Nothing wrong with saying "I wanted to write a (technically-informed) rant about something that's really bothering me, and I did."

F.ex. "clown-car turboslop" would be fairly polite by the standards of younger Torvalds (though he admittedly recognized later in life that wasn't a great communication style)

Poorly socialized people tend to have poorly developed social mores.

And a lot of developers, especially on the internet, tend to be poorly socialized.

Salt that with the fact that they often are more fact-intelligent than the people around them in real life, and you get a propensity to shitpost whenever thry feel it's warranted, especially about technical things.

I think you're vastly overestimating the technical competence and background of most people using agentic coding harnesses today.

The downside to democratizing code creation is... it puts it in the hands of non-developers.

F.ex. a team I work with asking OpenClaw "What is SRE?"

Going from "build and running executables" to that needing to be done on the same system and under control of the LLM is a large leap.

Separation of duties and ephemerality are well-known security mechanisms that many harnesses jump right over in the pursuit of easier UX.

Mingling permissions + trusting vibe-coded security boundaries in the harness itself (developed by folks whose appetite for risk would make even an 00s front-end developer blush) is rolling a handful of dice all at once.

But fundamentally, this is a disagreement between two risk appetites who will never agree: {it works most of the time} vs {it's guaranteed to never fail}

Skepticism is the flipside of knowledge.

If an LLM has knowledge encoded inside it (and it's hard to argue it doesn't), then cognitive dissonance can be experienced. And once experienced, must be dealt with, especially in longer-running agentic loops.

A friend was joking the other day about sending some messages under a previously-used Slack identity for an agent (since turned off), then asking the agent about the messages.

The agent maintained it hadn't sent those messages (no memory) and then was forced to reconcile the idea that the messages indeed appeared to come from it.

Its extremely-agitated conclusion was that there had been a security breach and the entire network should be locked down.

I feel like that's a lot of LLM use in a nutshell: sure, there's a more surgical scalpel available, but isn't it so much easier just to fling prompts at a magic box?

And as Hillary showed with her email server, digital records are more ephemeral these days, so there will be a lot of things we simply never know.

Ironically, part of the story behind the email server was that State Dept IT was so bad that provisioning a new email server was a months-long request.

So naturally, shadow IT became prevalent.

More to the point, because you've been chasing them around goalposts, you've both accepted that the question at hand is unknowable precisely because the source isn't available.

Do you believe in "guilty until proven innocent"? Do you believe that they're doing bad things with the bun code, and you can't be convinced otherwise in your judgment of them until they release the code?

Hell yes, and anyone who trusts one of the AI labs for anything important is insane.

Whether it's their initial mass piracy and then lying about it, dealmaking with the US government, or continually shifting opt-out lines on what they are and aren't training on, they haven't earned the trust to skip the "but verify" step.

Their economic incentives are too transparently aligned with ends at any costs.

And as the TCO enterprise adage goes: the cost of continued support.

It's one thing to value ones own time spent supporting something at $0, but most people probably shouldn't.

And it's incredibly inefficient to debug and maintain your own custom stack, when Valve can assign a team to do the same thing at orders of magnitude lower unit cost for standardized hardware/software configurations.

This. The biggest value proposition of a console that plays Steam games is the ability to acquire Steam games during Steam sales.

The closed console ecosystems seemingly have far less incentive to ever discount games, even old ones.

Radiolab did a great episode that covered the history of Chinese character computing: https://radiolab.org/podcast/wubi-effect

To summarize, in the 70s and 80s, China was facing an existential threat with their inability to access an economic accelerator (widespread computing) in their native language.

To the extent that there was serious consideration at the highest levels of converting the entire country to an alphabet-based writing system.

I'd expect they're looking at AI the same way:

We have to have access to this. Most of the frontier labs are American (or European). Therefore we need a solution we have continued access to.

Open weights feels simultaneously Chinese in nature (progress through making a design copyable and improvable by a large number of people) and economic (providing an incentive for the world to use Chinese models over other frontier).

It seems like there are four things being conflated (for simplicity's sake, no malice) here:

- Actual day-to-day acquaintances' responses to someone's perceived ideological heresies (read: your friends having a problem with your views)

- Individual and bandwagon responses to someone's perceived ideological heresies (read: "the internet" having a problem with someone's opinions)

- Corporate responses to someone's perceived ideological heresies (read: Google banning accounts or firing an employee)

- Government responses to someone's perceived ideological heresies (read: the executive branch punitively cutting unrelated funding or the DOJ starting frivolous lawsuits)

Ordered thusly because I think people (regardless of their orientations) have increasing issue with those things the farther down the list you go.

And especially #3 (corporate) and #4 (government) seem to have gotten out of control after 2016, somewhat as a response to worsening disingenuous behavior by #2 (the internet).

It's one thing if friends decide to not be friends because of expressed views. It's worse getting mobbed by 100+ strangers from the internet. And it should probably be unacceptable that there are corporate or government consequences for some of the low-grade "Well that's just your like your opinion, man" stuff in the last decade.

How do you know it's vibe coded slop? Have you seen the source code?

I believe you've just made your opponent's point for them.

Afaik, under current US law (Thaler v. Perlmutter 2025^), "the Copyright Act of 1976 requires all eligible work to be authored in the first instance by a human being".

With regards to AI-generated code, this means the copyrightability (at all, not with regard to who owns it) turns on whether or not a human was substantially involved in its creation.

Existing decisions require evidence of pretty heavy and continued human guidance to qualify.+

To wit, autonomous agent created code (prompt -> machine churns for hours -> output) is explicitly not eligible for copyright.

Functionally, this is a double-edged sword.

On the one hand, it means anything coded with autonomous agents by Meta, Google, et al. can be legally reused if it leaks (because no one could hold copyright on it).

On the other hand, it leaves copy-left open source licenses in a weird place. If you convert open source (even MIT/BSD-style) code into something else with an autonomous coding agent... the result has no copyright (nor can ever in the US).

In this instance, Bun was MIT-licensed, no? Then it was shoved mostly-autonomously through an LLM for the port.

Now Bun-Rust is technically still MIT licensed, but if push came to shove it seems like US law's current position is that Bun-Rust would now have no copyright license (because the manner in which it was developed renders it ineligible for copyright).

That's on the copyright side.

On the infringing usage side (i.e. whether you were entitled to shove a copyrighted work into a coding agent to produce something)... that's still TBD.

^ https://media.cadc.uscourts.gov/opinions/docs/2025/03/23-523...

+ https://garrettham.com/ai-generated-works-copyright/

To the end that they exist at all.

One of the biggest organizational anti-patterns of the modern SaaS world is the move away from release notes.

It's a question of whether a company has a competent enough change control process to be able to generate a list of changes for a given release.

If a tech product company can't meet the low bar of documenting its changes, then it probably shouldn't be trusted?

Release notes aren't for end users in the same way that the wiring diagram taped to the back of your appliances isn't for end users.

They're for experts to either directly use... or to communicate in simpler language to end users.

I'd vote fire, but people say Grug brain ethr1 know too much about old technology.

It's definitely a factor.

In the same way that the US going to war with an important Chinese oil supplier: 85% of Iranian oil going to China and comprising ~14% of China's oil imports.

There's always geopolitical reason behind the reason.

I think the market for Oracle Cloud is the same for early GCP: companies with large enough needs and strong enough engineering teams that they can leverage "X runs on Oracle Cloud" into deep discounts. And then cover the gaps with engineering.

I halfheartedly honestly concur

What will trigger the downturn will be folks who've overextended themselves on margin forced to sell in order to liquidate assets

Curious question... do the ultra wealthy ever get margin called on personal loans secured by unsold equity?

I'd assume the folks we're talking about doing that have sufficiently small loan to value (a year's expenses vs all their stock) to avoid, even in a downturn.

Maybe at the more upper income segment where LTVs on smaller equity totals are higher?