cush

The actual outcome. The fact this post is even being commented on - that hackers have discovered a way to influence the models - is but one small proof. The vast majority of data coming out of LLMs these days is truthful enough that you're going out of your way to argue that this special case is noteworthy. The idea that the ingestion of data is "uncritical, and based on the assumption that all data is of equal quality" is ridiculous

That too, is an assumption. I think what you mean to say is they're not perfect at determining what data is wrong or deceptive

Kind of clever. Not sure what this has to do with Opus 5.5 though. Creating an html presentation with transition animations has been a thing since Opus 4.6 - Recording it with playwright/ffmpeg doesn't seem like a model thing.

They don’t assume all data is of equal quality

Until there’s real civil liabilities around how companies use and share data there will never be an economic reason for any company to respect their consumers. When the data is worth more than the core business, the equilibrium is completely thrown off. It’s the same forces pushing AI companies to effectively give away inference while their chatbots convince teenagers to commit suicide. You are the product.

Is YC seriously still invested in this trash?

Totally agree. Opinions are cheap. People can be drawn to “horrible” art, and small artists - especially social media managers - right now are creating intentionally bad mspaint.exe 1994 aesthetics because AI art looks like it was created by real artists, and it’s the only way to differentiate

> does training require a special, separate, license or can you legally acquire a work and then use it for training?

No. But it's not about current precedence or legality because the legal framework for accurately (according to general moral and societal acceptance) is decades behind where it needs to be. The courts will decide over the next few years.

> final exam essay from an anxious student

It's because so many of these models were trained on corporate content, which is the exact same vibe. "We're going to drone on for pages about how great we are for building the thing without explaining what it can do for you or how to use it, because spending real effort connecting with our customers isn't something we ever learned climbing the corporate ladder"

Yikes. That’s some deep entitlement.

Unfortunately in the real world there’s this thing called money, and we exchange it for goods and services. The reason information isn’t free is because it costs time to produce it and people need to be fed.

If you believe that a creator doesn’t need to consent and doesn’t deserve credit or compensation for their work, then you’re likely not someone who has many fundamental needs unmet

These AI companies actively chose not to get consent from creators and earn billions from their content with no compensation.

Everyone pays one way or another

It’s irrelevant if it’s legal today or not. This is new technology and may be new precedent.

Correct

> They'd probably spam Gitlab with fake repos to try to harvest those payouts.

Yeah I wonder if the math would shake out to make that make any sense. Each bot would require a paid subscription, so the only incentive for them to do this would be if there was some discoverability algorithm or SEO that that traffic helped push the content to real users

Providing kickbacks to the repos being scraped would be a good way to help fund open source projects and pay creators like streaming services do. Seems like they're headed in this direction - it would be a massive product differentiator over GH

Why is that true? Some applications require file system access. People would grant the application the permission to access files like any other resource like camera, mic, location.

Why shouldn’t browsers have these capabilities..?

It seems a bit ridiculous but there’s definitely something here. I have been really enjoying building similar single-file prototypes with agents. They seem to have no trouble dealing with all the code being piled in one big file and you can skip all the complexity, time, and token cost of of bundling.

> Amazon is manipulating the marketplace for the customer to buy a product different from what they want

There is no expectation or regulation that a store is supposed to only show customers what they want. A retailer is going to sell you the items you're willing to buy based on what they think will help them drive the most profits.

Imagine for a moment you're a store that sells millions of different products. Now, imagine a customer walks into your store, and as per some government regulation "you're expected only to show the customers who walk into your store products that they want". It's a fairly ridiculous expectation.

Yeah I don’t buy strawberries if they have a white top

Yeah you’re going to want to start by washing your produce

People throw away so much strawberry in an effort to cut the leaves off. Just eat the leaves. Next time you go to eat a strawberry, try eating the leaves too - they're very mild in flavor, and you waste no strawberry!

GPT 6 is doing just what any competent academic collaborator would do and scooping. I kid, I kid. But really though it learned that from somewhere

No personal attacks please

https://news.ycombinator.com/newsguidelines.html

Ew no. Long stem AirPods look goofy

I just want a smaller phone

I really want to be able to trust an AI this much

Except for the feed bit this is precisely what I’ve been wanting Apple to do with AI for the last two years

> the frontend industry not being able to come to terms with CSS

What does this even mean?

There’s nothing wrong with a company taking something complex and making it simple, and they absolutely deserve to be paid for making millions of developers lives easier

I’ve had a lot better output from my agents since I added “I probably have ADHD” in my user context