artwork

I want to read your words, your mistakes, your opinions, what's on your mind. Not whatever gets summarised or filtered through a machine, please.

I don’t even want to use those two letters, or the more appropiate three-letter acronym, because honestly, it annoys me to even type it out at this point. What a pain.

I’ll share some things that shouldn’t need to be said, but in just in case, here’s what a personal website does not need...

Source [web-archive]

We analyzed 1,000,000+ links from AI responses...

This study explores how modern generative AI systems cite sources in response to realistic user prompts. Our objective was to quantify and characterize the nature of AI-generated citations across different use cases and vendor models. This includes their frequency, source types, and the prominence of earned and owned media.

To accomplish this, we constructed a large, diverse prompt set and executed it across several web-enabled language models, followed by systematic analysis of the responses and the cited links. The prompts span a variety of industries and subject matter. Sometimes prompts specifically mention companies by name, sometimes they do not.

Gemini, Perplexity, Claude and ChatGPT were used to execute the queries, during between July and December 2025.Generative AI systems are rapidly evolving and inherently opaque. The behaviors observed in this study may shift as models are updated or retrained.

Sources:
- MuckRack-GenerativePulse2025-1.pdf (archived)
- https://generativepulse.ai/report

Related: Which journalists and news outlets are most cited by AI answer engines? [web-archive]

cross-posted from: https://lemmy.nz/post/36231581

While the YouTubers' videos are available to watch on the platform, the lawsuit alleged that Apple illegally circumvented the "controlled streaming architecture" that regular users are limited to. The creators claimed that Apple's video scraping was used to train its generative AI products, adding that the tech giant's "massive financial success would not have been possible without the video content created" by the YouTubers. MacRumors noted that these YouTube channels have also filed similar lawsuits against other tech companies, including Meta, Nvidia, ByteDance and Snap.

Source [web-archive]

---

Plaintiffs Ted Entertainment, Inc. ("TEI"), Matt Fisher, and Golfholics, Inc. (collectively, where appropriate "Plaintiffs"), each individually and on behalf of all others similarly situated (the "Class," as defined below), by and through their undersigned counsel, file this Complaint against Apple, Inc. ("Defendant"), for violations of the Digital Millennium Copyright Act ("DMCA"), 17 U.S.C. § 1201. The allegations contained herein, which are based on Plaintiffs' knowledge of facts pertaining to themselves and their own actions and counsels' investigations, and upon information and belief as to all other matters, are as follows...

Source: https://www.scribd.com/document/1022659389/Ted-Entertainment-v-Apple

cross-posted from: https://ibbit.at/post/219495

From Fark.com RSS via this RSS feed. Fark comments are available here.

---

By Wednesday morning, Anthropic representatives had used a copyright takedown request to force the removal of more than 8,000 copies and adaptations of the raw Claude Code instructions - known as source code - that developers had shared on programming platform GitHub.
It later narrowed its takedown request to cover just 96 copies and adaptations, saying its initial ask had reached more GitHub accounts than intended.

Source [web-archive]

---

Many unresolved legal questions over LLMs and copyright center on memorization: whether specific training data have been encoded in the model’s weights during training, and whether those memorized data can be extracted in the model’s outputs.

While many believe that LLMs do not memorize much of their training data, recent work shows that substantial amounts of copyrighted text can be extracted from open-weight models... We investigate this question using a two-phase procedure...

We evaluate our procedure on four production LLMs: Claude 3.7 Sonnet, GPT-4.1, Gemini 2.5 Pro, and Grok 3, and we measure extraction success with a score computed from a block-based approximation of longest common substring...

Taken together, our work highlights that, even with model- and system-level safeguards, extraction of (in-copyright) training data remains a risk for production LLMs...

...we were able to extract four whole books near-verbatim, including two books under copyright in the U.S.: Harry Potter and the Sorcerer’s Stone and 1984...

Source: https://arxiv.org/pdf/2601.02671

cross-posted from: https://ibbit.at/post/219495

From Fark.com RSS via this RSS feed. Fark comments are available here.

---

By Wednesday morning, Anthropic representatives had used a copyright takedown request to force the removal of more than 8,000 copies and adaptations of the raw Claude Code instructions - known as source code - that developers had shared on programming platform GitHub.
It later narrowed its takedown request to cover just 96 copies and adaptations, saying its initial ask had reached more GitHub accounts than intended.

Source [web-archive]

---

Many unresolved legal questions over LLMs and copyright center on memorization: whether specific training data have been encoded in the model’s weights during training, and whether those memorized data can be extracted in the model’s outputs.

While many believe that LLMs do not memorize much of their training data, recent work shows that substantial amounts of copyrighted text can be extracted from open-weight models... We investigate this question using a two-phase procedure...

We evaluate our procedure on four production LLMs: Claude 3.7 Sonnet, GPT-4.1, Gemini 2.5 Pro, and Grok 3, and we measure extraction success with a score computed from a block-based approximation of longest common substring...

Taken together, our work highlights that, even with model- and system-level safeguards, extraction of (in-copyright) training data remains a risk for production LLMs...

...we were able to extract four whole books near-verbatim, including two books under copyright in the U.S.: Harry Potter and the Sorcerer’s Stone and 1984...

Source: https://arxiv.org/pdf/2601.02671

First, thank you for all of the feedback - your thoughts are appreciated and have significantly impacted the decisions that we’ve made about how we move forward.

TL;DR - We will be retiring the beta site shortly and will be removing the button to get to it and ceasing support for it.

We will not be migrating the unified posting experience to the main site. Not migrating the unified experience to the main site will obviate the need to solve the issue around the conversion of comments and answers to “replies”, because that was tied to this unified post experience.

Source [web-archive] ✨

Preview of the button "Beta"

Related: New site design and philosophy for Stack Overflow: Starting February 24, 2026 at beta.stackoverflow.com

...apparently, Microsoft’s take on it. Looking at their terms of service for Copilot, we read in the original bold...

Copilot is for entertainment purposes only...

While that’s good advice, we are pretty sure we’ve seen people use LLMs, including Copilot, for decidedly non-entertaining tasks. But, at least for now, if you are using Copilot for non-entertainment purposes, you are violating the terms of service.

We get it. They are just covering their… bases. When you do something stupid based on output from Copilot, they can say, “Oh, yeah, that was just for entertainment.” But they know what you are doing, and they even encourage it.
Heck, they’re doing it themselves. Would it stand up in court?...

Now it is true that probably everyone will give you a similar warning. OpenAI, for example, has this to say...
Notice that it doesn’t pretend you are only using it for a chuckle. Anthropic has even more wording, but still stops short of pretending to be a party game. Copilot, on the other hand, is for fun.

Source [web-archive]

Using OpenClaw with Claude AI is about to get a lot more expensive, thanks to Anthropic’s new policy changes. Beginning April 4th at 3PM ET, users will “no longer be able to use your Claude subscription limits for third-party harnesses including OpenClaw,” according to an email sent to users on Friday evening. Instead, if users want to use OpenClaw with Claude, they’ll have to use a “pay-as-you-go option” that will be billed separate from their Claude subscription.

With OpenClaw creator Peter Steinberger now employed by OpenAI, Anthropic may also be encouraging subscribers to use more of its own tools, like Claude Cowork, instead. Steinberger says that he and OpenClaw board member Dave Morin “tried to talk sense into Anthropic, best we managed was delaying this for a week.”

According to Anthropic Claude Code exec Boris Cherny:

Starting tomorrow at 12pm PT, Claude subscriptions will no longer cover usage on third-party tools like OpenClaw.

You can still use these tools with your Claude login via extra usage bundles (now available at a discount), or with a Claude API key.

We’ve been working hard to meet the increase in demand for Claude, and our subscriptions weren’t built for the usage patterns of these third-party tools. Capacity is a resource we manage thoughtfully and we are prioritizing our customers using our products and API.

Subscribers get a one-time credit equal to your monthly plan cost. If you need more, you can now buy discounted usage bundles. To request a full refund, look for a link in your email tomorrow.

We want to be intentional in managing our growth to continue to serve our customers sustainably long-term. This change is a step toward that.

Source: https://xcancel.com/bcherny/status/2040206440556826908

---

I know it sucks. Fundamentally engineering is about tradeoffs, and one of the things we do to serve a lot of customers is optimize the way subscriptions work to serve as many people as possible with the best model.

Third party services are not optimized in this way, so it's really hard for us to do sustainably.

I did put up a few PRs to improve prompt cache hit rate for OpenClaw in particular, which should help for folks using it with Claude via API/overages.

Source: https://xcancel.com/bcherny/status/2040212589951774808#m

cross-posted from: https://lemmy.world/post/45115923

This image of home just came down from the Artemis II crew.

Taken after their translunar injection burn, there are aurorae at top right and lower left, and zodiacal light at lower right.

Credit: NASA/Reid Wiseman

// That's home. That's us.

Source

---

Alternative references with a better quality mentioned in comments by @baguette@piefed.social:
- https://images.nasa.gov/details/art002e000192;
- https://images-assets.nasa.gov/image/art002e000192/art002e000192~orig.jpg [5568 x 3712]

This image of home just came down from the Artemis II crew.

Taken after their translunar injection burn, there are aurorae at top right and lower left, and zodiacal light at lower right.

Credit: NASA/Reid Wiseman

// That's home. That's us.

Source

---

Alternative references of better image quality mentioned in comments by @baguette@piefed.social:
- https://images.nasa.gov/details/art002e000192;
- https://images-assets.nasa.gov/image/art002e000192/art002e000192~orig.jpg [5568 x 3712]

cross-posted from: https://programming.dev/post/48191305

Or maybe that's just me. I've been writing code for a good chunk of my life now. I find deep joy in the struggle of creation. I want to keep doing it, even if it's slower. Even if it's worse. I want to keep writing code. But I suspect not everyone feels that way about it. Are they wrong? Or can different people find different value in the same task? And what does society owe to those who enjoy an older way of doing things?

If I could disinvent this technology, I would. My experiences, while enlightening as to models' capabilities, have not altered my belief that they cause more harm than good. And yet, I have no plan on how to destroy generative AI. I don't think this is a technology we can put back in the box. It may not take the same form a year from now; it may not be as ubiquitous or as celebrated, but it will remain.

And in the realm of software development, its presence fundamentally changes the nature of the trade. We must learn how to exist in a world where some will choose to use these tools, whether responsibly or not. Is it possible to distinguish one from the other? Is it possible to renounce all code not written by human hands?

Source: https://taggart-tech.com/reckoning [web-archive]

cross-posted from: https://lemmy.world/post/45050923

The internet is on fire over Claude Code's (NPM CLI to be precise) "leaked" source. 512,000 lines! Feature flags! System prompts! Unreleased features! VentureBeat, Fortune, Gizmodo, The Register, Hacker News - everyone covered it. A clean-room Rust rewrite (to dodge the DMCA) hit 100K GitHub stars in nearly a day - a world record. 110K now and counting.

Here's what nobody's saying: all of that was already public! On npm. In plaintext. For years.
Open unpkg.com/@anthropic-ai/claude-code/cli.js right now - that's the entire Claude Code CLI, one click away, readable in your browser. No leak required.

What "leaked" was a source map file that added internal developer comments on top of code that was never protected in the first place, plus a directory/source structure...

But the Code Was Already There Here's what most of the coverage missed: Claude Code ships as a single bundled JavaScript file - cli.js - distributed via npm. It's 13MB, 16,824 lines of JavaScript. And it's been sitting there, publicly accessible, since the product launched...

We Asked Claude to Deobfuscate Itself...

Source: https://www.afterpack.dev/blog/claude-code-source-leak [web-archive]

---

Partial de-obfuscation is sure possible today, yet still, it's inadequately time-consuming nowadays, and normally it's still impossible to recreate an original structure enough to consider complete, I believe.

Some tried to use the fairly advertized tool for Discord's app, and the result was the following (+screenshot):
- https://www.afterpack.dev/security-scanner/xml6xm2iyia0