Does Anthropic Owe Me $3000?

Welcome to the Tech Brew Ride Home for Monday, September 8, 2025. I am Brian McCullough. Anthropic agrees to pay one point five billion dollars to authors, do I get three thousand dollars for my book? Also, OpenAI is making a movie, Nasdaq wants tokenized stocks, three interesting raises, and at the end I will tell you about my weekend AI experiment. Here is what you missed today in the world of tech.

Anthropic will pay one point five billion dollars plus interest to settle a class action accusing it of training on millions of pirated books. The deal reportedly pays about three thousand dollars per book for roughly five hundred thousand books to start, requires data destruction, and heads to a hearing today. Anthropic says it faced inordinate pressure to avoid a business-ending trial, even as it boasts a massive valuation and remains unprofitable thanks to compute costs.

Should I apply for this? Three thousand dollars is not chump change, and I am pretty sure my internet history book is in there somewhere. Over the weekend, an AI project even cited research I did years ago, which I will get to at the end.

MG Siegler makes a spicy point: fresh off a huge fundraise, did Anthropic just pull up the drawbridge by setting a floor price for AI training infringement that only the biggest players can afford? Writers may be euphoric, but this could shrink the competitive field and set a new standard that tells rivals to open their wallets if they want to play. It feels a little Godfather, a little floor-setting maneuver that narrows the AI wars.

Meanwhile, full speed ahead on creation. OpenAI is backing Critters, a largely AI-made animated feature aiming for theaters in 2026, with a budget under thirty million dollars and a nine-month production timeline instead of three years. Chad Nelson is leading it with Vertigo Films and Native Foreign, using GPT-5 and image models, hiring human voice actors, and sharing profits among roughly thirty people. The bet is simple: prove AI can deliver big-screen quality and lower the cost of entry for creatives.

Tokenizing the stock market inches toward prime time. Nasdaq asked the SEC to allow trading of tokenized securities on its main market as early as the third quarter of 2026, with tokenized and traditional shares on the same order book if rights are identical. Regulators say tokenization will not sidestep securities laws, while banks and platforms explore the space and Europe’s gray zones show why a higher bar matters.

Interesting raise one: Mistral is finalizing about two billion euros at a twelve billion dollar valuation, pushing open source models, Le Chat, and a French data center, plus debt for Mistral Compute with Nvidia. ASML is reportedly committing one point three billion euros at a ten billion euro pre-money valuation, becoming a top shareholder with a board seat and tying two European champions together.

Interesting raise two: Databricks is set to close one billion dollars at a one hundred billion dollar valuation and says it is on track for four billion dollars in revenue in the fiscal year ending January 2026, up fifty percent year over year. Around six hundred fifty customers pay at least one million dollars annually, cash flow is positive, and the new capital fuels the AI talent wars.

Interesting raise three: ElevenLabs is running an employee tender of up to one hundred million dollars at a six point six billion dollar valuation. Their hyper-realistic voices power everything from multi-language audio to virtual agents and even podcasts.

Quick programming note: tomorrow is iPhone Day. The show will be a few hours late so I can watch and then break everything down for you.

About that weekend experiment. I read about AI-generated boring history videos flooding YouTube, so I tried it: I had ChatGPT write two ten thousand word histories, one on Google and one on Amazon, and I fed them into ElevenLabs with a soft British voice. I got two approximately ninety-minute, fall-asleep-to tech history narrations and posted them on YouTube. I checked every fact and source, and the scripts even cited interviews I did a decade ago.

So, can AI do what I do? Yes, but the curation mattered. I iterated prompts dozens of times, nudged for accuracy and detail, and verified everything. Is there a role for AI slop if it is brewed by someone you trust? I am cooking up another test on a topic I know less about to see how much work it takes to get it right. Links are in the show notes—sample a few minutes and tell me if this is good AI slop. Talk to you tomorrow for the iPhone Day.