Today's Headlines

  • SpaceX completes its acquisition of Anysphere, the maker of Cursor, in an all-stock deal at an implied equity value of $60.0 billion
  • Textbook authors file a class action against nine OpenAI entities and Microsoft in the Southern District of New York
  • OpenAI's CFO tells investors that enterprise revenue has passed the consumer business, with the run rate at $40 billion
  • Anthropic explains how Claude's text watermark survives editing and how it behaves inside code
  • A developer reports a 232x faster GPU kernel produced through repeated passes with OpenAI's Codex

Today's three stories moved in the structural parts of the industry. Ownership, the use of copyrighted work, and the centre of gravity in revenue each shifted on the same day, on three different faces.

What SpaceX moved is ownership. The leading AI coding environment became a wholly owned subsidiary, paid for in stock.

What the textbook authors moved is procedure around the use of copyrighted work. They carried the question of training data into federal court, framed by the Copyright Act and the Digital Millennium Copyright Act.

What OpenAI described is the centre of gravity. The majority of revenue now comes from enterprise, its finance chief told investors.

Today's Top Three

SpaceX completes its acquisition of Cursor, paying in Class A stock

SpaceX completed its acquisition of Anysphere, the company behind the Cursor coding environment.

The structure is a reverse triangular merger. According to the Form 8-K filed with the Securities and Exchange Commission, a June 16 merger agreement provided for X67 Inc., a wholly owned SpaceX subsidiary, to merge into Anysphere, and the merger became effective on August 14. Anysphere, Inc. survives the merger as a wholly owned subsidiary of SpaceX.

The consideration is stock in place of cash. Cursor's outstanding common and preferred shares converted into the right to receive an aggregate of 389,289,254 shares of SpaceX Class A common stock. The share count rests on an implied equity value for Cursor of $60.0 billion and a price per SpaceX share equal to the volume-weighted average closing price over the seven consecutive trading days immediately preceding the closing.

Vested equity settled on the same day. The 8-K puts the vested Cursor restricted stock units at an aggregate of 1,752,426 SpaceX shares before any withholding for applicable taxes, with cash paid in lieu of fractional shares.

Unvested equity carried over. Unvested restricted stock units and Cursor stock options were assumed and converted into approximately 29,128,326 SpaceX restricted stock units and approximately 44,365,047 options over SpaceX Class A common stock.

The legal basis for the issuance appears in the same filing. SpaceX issued the merger consideration in reliance on the registration exemption in Section 4(a)(2) of the Securities Act of 1933, as a transaction by an issuer involving a private placement. Chief Financial Officer Bret Johnsen signed the report.

Cursor described the same event on its own blog, saying the process that began with the SpaceXAI partnership announced in April is now complete. The company points to access to one of the world's largest GPU fleets as the aim, and cites Grok 4.6, released Wednesday, as an early result of the collaboration.

Textbook authors file a class action against OpenAI and Microsoft in the Southern District of New York

Seven textbook authors filed a class action complaint against nine OpenAI entities and Microsoft.

The case stands at the filing stage. The complaint reached the United States District Court for the Southern District of New York on August 14, 2026, as case number 1:26-cv-06966, running 26 pages and carrying a demand for a jury trial.

The plaintiffs are authors of college textbooks. They are Michael Sullivan, professor emeritus of mathematics at Chicago State University; Kenneth S. Saladin, in anatomy and physiology; Zvi Bodie and Alan J. Marcus, in finance; Kevin T. Patton, also in anatomy and physiology, together with his corporation Lion Den Inc.; and Dana Loewy, in business communication.

The defendants include the post-restructuring entities by name. OpenAI Foundation, formerly OpenAI Inc., and OpenAI Group PBC head a list of nine OpenAI entities, with Microsoft Corporation as the tenth defendant.

The complaint pleads four counts. They are direct copyright infringement against OpenAI under 17 U.S.C. §§ 106 and 501, vicarious infringement against Microsoft, contributory infringement against all defendants, and distribution of works with copyright-management information removed under 17 U.S.C. § 1202(b), also against all defendants.

The factual allegations concentrate on sourcing and processing. The complaint traces the Common Crawl, Books1 and Books2 training datasets to shadow libraries such as LibGen, reads the 55 billion tokens in Books2 as suggesting more than 100,000 books, and alleges that OpenAI built and ran a script that stripped copyright notices, titles, ISBNs, the © symbol and DOIs from the front and back matter of those books.

Standing is pleaded in a way particular to textbooks. Each plaintiff assigned exclusive publication and distribution rights to a publisher under a publishing contract, so the complaint proceeds under 17 U.S.C. § 501(b), which allows a beneficial owner of an exclusive right to sue.

The market theory separates textbooks from trade books. The complaint argues that textbook sales run through adoption, where an instructor or a school selects the book and a student or a school system pays for it, which makes the purchasing decision structurally different from a reader choosing a novel. The proposed class reaches textbook authors holding legal or beneficial ownership of a United States copyright registered with the Copyright Office.

The relief sought is written by reference to the statutes. The complaint asks for class certification, a judgment of infringement and of a § 1202(b) violation, statutory damages or at the plaintiffs' election actual damages and the defendants' profits under §§ 504 and 1203, declaratory and permanent injunctive relief including restoration of copyright-management information, and interest, costs and attorneys' fees, leaving the amount to the statutory framework.

The complaint also addresses timing in advance. On the three-year period in 17 U.S.C. § 507(b), it argues that the contents of the training datasets stayed outside public disclosure until Alex Reisner's articles in The Atlantic on August 19 and September 25, 2023 made the copying of textbooks identifiable.

The same group of plaintiffs sued a different defendant in July. According to the Text and Academic Authors Association, they filed a comparable action against Meta and Mark Zuckerberg on July 2, 2026.

OpenAI's CFO tells investors that enterprise revenue has passed consumer

OpenAI Chief Financial Officer Sarah Friar told an investor meeting on August 14 that enterprise revenue has passed the ChatGPT-led consumer business.

The attribution runs through two steps. CNBC quotes Friar as saying, "We entered the year at 60-40, but enterprise has accelerated much faster than expected and those lines have now crossed," and "The majority of our revenue is now enterprise," citing a person who attended the meeting and asked to stay anonymous.

The figures arrived through separate channels. CNBC confirmed an annualized revenue run rate of $40 billion, a number Bloomberg was first to report. Slides viewed by CNBC show the July run rate up 20% month over month, with business customers up 32%.

The crossing runs ahead of earlier guidance. Friar had told CNBC that the two businesses would reach parity by the end of 2026, which places Friday's account ahead of that schedule.

The meeting landed in a week of executive changes. Revenue chief Denise Dresser stepped down a day earlier, eight months into the job, and Brad Lightcap said two days before that he was ending an eight-year run at the company. CNBC reports that the meeting itself had been scheduled before those departures.

Friar also described a shift in customer behaviour. She said the era of tokenmaxxing has passed and that enterprise customers now look at cost per unit of intelligence. Advertising is approaching a $1 billion run rate, with testing inside ChatGPT under way since February, and she declined to discuss the timing of an initial public offering, citing the confidential SEC filing.

Read the three together and each one lands on a different face. Who owns what was settled in an 8-K, what may be used was carried into a courthouse as a complaint, and where the money comes from surfaced in a briefing to investors, all within the same day.

Other Developments

Products

Developer Stephen Cresswell published version 3.0.0 of Yadda, a behaviour-driven development library for JavaScript, to npm. Writing on his own blog, he says Claude Code built most of this release, and frames it as a case for executable specifications gaining value in agentic development.

Research

Independent developer sankalp reported using OpenAI's Codex to iteratively optimize a GPU kernel for the qr_v2 problem in a GPU Mode contest, reaching a 232x speedup over the baseline. The figure rests on the developer's own account, with the blog post standing as its sole source.

Policy

TechCrunch wrote up Anthropic's August 14 explainer on how Claude's text watermark behaves in practice. The mechanism builds a statistical pattern into low-stakes word choices — whether the weather reads as "gloomy" or "grey" — on top of the SynthID-Text approach Google DeepMind published in 2024, and detection stays with the holder of the key. Light editing leaves the mark intact, while a full rewrite that replaces every word removes it. In code, the watermark rides only in places that keep some freedom of wording, such as comments. Anthropic plans a detection API, and the work sits alongside the transparency Code of Practice under the EU AI Act, so other major developers who signed it are expected to ship watermarks of their own. TechCrunch cites Business Insider on dozens of users announcing cancellations on X, while Anthropic told the same outlet that cancellations remain at their usual level.

Source: Selected by the editors from the AI news inbox (7 items collected on August 16, 2026 — 3 primary, 4 secondary).