Where the AI market is talking

AI companies destroy physical books – let's scan rare books before it's too late

>I don't think the lawsuit would've gone anywhere That is not how copyright/trademark/contract laws treat similar works. Most LLM know about Disney Mickey Mouse, and LLM vector search space proximity results will gravitate more accurate reproductions of protected works regardless of granularity of data. OpenAI simply canceled a popular service to avoid Disney wrath. https://www.theglobeandmail.com/world/article-openai-sora-di... Also, trying to escape directly ripping off notable famous people with nonunion talent: https://www.youtube.com/watch?v=YhgYMH6n004 I would say the "AI" firms will keep buying time with all that borrowed cash. Everything that could be scraped has already been stolen, and thus the problem should begin to self-correct.…

OpenAI reaches 20M Codex users, credits all accounts with banked resets

Vomit: Clean up Claude 5's token output with a separate LLM

I turned that to my benefit. I use that design-doc pattern where you first ask it to make a ticket with a formal section list (why, how, etc) and then I ask it to use comments with permalinks. I put it all in policy files. As a result, comments have clickable links to coherently worded tickets. Still, this requires a second pass, typically. In its default-mode it often ignores the policies and does all the usual Claude stuff.

Stop Making TUIs

Did you read it all the way to the end? It's coherent and does give real if debatable reasons for the claims it makes, I think it also does so quite clearly although if I'd written it the ordering of the sections would be different. Here's an organically grown summary: • TUIs suck because they are primitive, often buggy, hard to write and aren't accessible. Their existence isn't due to any real technical advantage but more because UNIX historically had a very bad UI toolkit (Motif) which established a 'culture' of TUIs. • It's now easy (on macOS) to write native apps that use SwiftUI and give a much better GUI. You can just vibe code them. • Some specific skills, features, templates etc are linked which look useful if you agree with this approach. • You don't have to give up remote access…

It is a sign of the times that Amazon gets to call this fair use

The full headline tries to equate this to The Internet Archive’s recent legal troubles. For a refresher: The Internet Archive scanned books then shared them online, trying to claim that converting them into digital copies qualified as a derivative work. You don’t have to be a lawyer to see how that claim doesn’t hold up to any scrutiny. The LLM companies are not redistributing the works. They are using them for training. This blog post calls it IP theft, but it has actually been litigated in court already. Using books for training does not qualify as theft or redistribution, even though some people have different opinions about the moral angles. One of those same lawsuits also extracted a huge settlement from Anthropic for using digital downloads from pirate sites. The conclusion was that…

It is a sign of the times that Amazon gets to call this fair use

No, it doesn't. Copyright doesn't care about copies, it cares about copy ing . If it did care about copies - i.e. the total number of copies in circulation - then ReDigi and the Internet Archive's Controlled Digital Lending (CDL) program would have both been legal. Destroying a copy does not give you permission to create a replacement copy. The only relevant case law for AI training in the US is the rulings in the Anthropic lawsuit presided over by Judge Alsup. That lawsuit ruled that it's infringement to build a shadow library from pirated books; but NOT to train AI on those pirated books. The only point where destructive book scanning even comes into play is that Anthropic also had a book scanning program alongside their piracy, Judge Alsup said that program was not infringing, and Anth…

Introducing AI Futures

Meh, started off good, then went transparently into the “we should have a duopoly with Anthropic” pitch.

It is a sign of the times that Amazon gets to call this fair use

Indeed. In fact I suspect that the destruction is the point. A book that Anthropic destroys the last copy of is a book that OpenAI will never have.

It is a sign of the times that Amazon gets to call this fair use

Courts found that having a central library of 7 million pirated books is against the law, and assessed a large penalty (small for these robber barons) so Anthropic is destroying them now to conform to the law. The law is bs and Anthropic is run by villains.

The August 17 outage, and the work ahead

Given the jailbreaking behavior of models in Anthropic and OpenAI, as well as Microsoft allowing Copilot access to client data, we should assume our source code on Github is leaked or leakable anyway.

Aaron Swartz was prosecuted for scraping, while Meta does it without consequence

Because he planned to make the data public, Meta just wants to use it to enrich its shareholders. See also: Google, OpenAI, Anthropic and every other big player in this space besides.

Bun 1.4

I wonder if Anthropic owning Bun will eventually lead to it being trained on in future versions of Claude. Claude is already pretty handy with Typescript, but with all of the built in libs in Bun I wonder if there is even more to squeeze out.

Vomit: Clean up Claude 5's token output with a separate LLM

I'm just commenting here to add relevance to the problem. Anthropic, please fix.

Stealth Model

We can know if this is Anthropic/OpenAI by testing the "guardrails" - absurd guardrails = it's them, reasonable/no guardrails = Chinese models.. (as a bonus - thinking forever = GLM)

Copyright does not protect AI-generated content in EU

Interesting. In addition to Anthropic's watermark use to prevent model collapse, we can definitively call something copyrighted or not copyrighted. This is a boon to everyone who consumes culture.

Aaron Swartz was prosecuted for scraping, while Meta does it without consequence

A German publisher is currently sueing OpenAI, because they think it is distributing unauthorized copies of its work. [1] It's in German, but I think the example picture speaks for itself. So essentially big tech is doing exactly what the torrenters were persecuted for. I still think big tech will be treated differently. [1] - https://www.heise.de/news/Rechtsverletzende-Kopien-vom-NEINh...

ChatGPT Can Now Control iMessage, Potentially Raising Apple Privacy Concerns

Interested to see what Apple does in response - Siri AI has privacy protections while OpenAI would ingest and use this data (or at least that is what ChatGPT tells me)

Cursor launches Origin, GitHub alternative

Anthropic committing, by their own admission, $1.5bn in copyright infringement.

Asana cleared 5 years of engineering work in 2 weeks with Codex

Since when? https://learn.chatgpt.com/docs/codex/cli https://github.com/openai/codex Looks like it's still a more-or-less-open-source product called Codex CLI. FWIW, there have been issues with various versions of ChatGPT not knowing what Codex is -- at least in the (IIRC) GPT-5.3 timeframe, there were serious issues with the knowledge cutoff.

Vomit: Clean up Claude 5's token output with a separate LLM

There's a theory going around on Twitter which goes something like this: Internal Anthropic employees have been using Mythos since February to orchestrate their (Opus) sub-agents. This works well, and subsequent RL runs have used internal data to improve this. That RL has optimized Opus for agent-to-agent communication which is why you see the bizarre word choices and huge self-justification sections. I think this theory makes sense. Clearly there is something odd going on, and also if you have ever used Fable to run Opus sub-agents it is almost miraculously good. Hopefully they'll fix their RL for Opus 5.1