Skip to main content


Microsoft has eliminated the onboarding fees for developers submitting apps to its Windows Store. Nearly 200 countries' individual developers can now publish apps using just a personal Microsoft account without paying the previous one-time $19 fee. This move makes the Microsoft Store more accessible compared to Apple, which charges an annual $99 fee, and Google, which charges a one-time $25 registration fee. According to Chetna Das, senior product manager at Microsoft, this change removes significant barriers and encourages innovation by allowing developers to thrive in the Windows ecosystem. The Microsoft Store, which has over 250 million monthly active users, supports various app types including Win32, UWP, PWA, .NET, MAUI, and Electron apps. Developers can also use their own in-app commerce system and keep 100% of their revenues on non-gaming apps.


Despite Google's prior claims of a thriving open web and stable search traffic, a recent court filing reveals the company admits the open web, specifically open-web display advertising, is in rapid decline. This admission arises amidst a legal battle concerning Google's monopoly in advertising technology, with the US Department of Justice recommending a breakup of Google's ad business. Google argues that such divestiture would accelerate the decline, harming publishers dependent on open-web display ad revenue. The company highlights the rise of non-open web ad formats like Connected TV and retail media as reshaping the ad ecosystem. While Google maintains that search traffic and click volumes remain stable or have expanded to more sources, publishers have reported noticeable drops in traffic following changes in Google's search algorithms and the advent of AI chatbots. Google spokespersons emphasize that the decline pertains to specific advertising formats rather than the open web as a whole, reaffirming Google's stance that the web continues to thrive.


Google has introduced significant updates to its Gemini-powered products. The Gemini app now supports audio file uploads, fulfilling the top user request, with free users limited to 10 minutes of audio and five prompts daily, while AI Pro and Ultra users can upload up to three hours. Google Search AI Mode has expanded to include five new languages: Hindi, Indonesian, Japanese, Korean, and Brazilian Portuguese. Additionally, Gemini-powered NotebookLM now offers new report styles such as study guides, blog posts, flashcards, and quizzes in over 80 languages, customizable by users. These updates enhance user experience across Google’s AI tools, continuing a trend of rapid innovation in AI technology.


Showrunner, a startup aiming to revolutionize the entertainment industry with AI-generated videos, is working to restore lost footage from Orson Welles' 1942 film The Magnificent Ambersons. The original film was drastically cut by the studio without Welles' input, and the lost footage's negatives were destroyed. Showrunner plans to use a new generative AI model, FILM-1, to approximate Welles' lost scenes by generating keyframes and manipulating live actors' faces to resemble the original cast. They've enlisted AI VFX artist Tom Clive and filmmaker Brian Rose to assist with the project. Unlike previous controversial AI uses, Showrunner intends not to monetize this project, aiming instead to preserve and present the film as Welles originally intended, and is willing to share the results with the IP holders for academic or commercial use.


Anthropic has agreed to pay a minimum of $1.5 billion plus interest to authors in a landmark settlement over the use of copyrighted works to train its AI systems. The settlement, possibly the largest in US copyright litigation history, is expected to provide approximately $3,000 per book or work. It requires court approval with a hearing scheduled for September 8th. The final payout could increase depending on the number of claims submitted, as Anthropic will pay an additional $3,000 per work beyond 500,000 claims. Anthropic must also destroy the original files it downloaded and any copies as part of the settlement. This deal covers past claims only and does not permit future AI training use of these materials after August 25, 2025. The case follows allegations from authors who accused Anthropic of using pirated books for AI training, amid ongoing legal scrutiny of AI companies regarding copyright infringement. Anthropic maintains its commitment to safe AI development and compliance with copyright law.


Amid a summer of significant changes including a multibillion-dollar investment from Meta and leadership hires by Mark Zuckerberg, Scale AI is now embroiled in a legal battle. The AI data labeling firm has sued former head of engagement management Eugene Ling and his current employer, competitor Mercor, accusing Ling of stealing over 100 confidential files. This lawsuit highlights tensions within the competitive AI industry.


My first experience coding with ChatGPT in early 2023 reminded me of 'The Monkey's Paw,' a tale where granted wishes come with harsh consequences. ChatGPT executed my requested changes with brutal literalness, often disrupting unrelated code. While some parts were useful, the resulting code was over-engineered and cluttered with irrelevant fragments, making the process feel cumbersome. Despite these challenges, I continued exploring AI-assisted coding tools earlier this year...
This is The Stepback, a weekly newsletter that explores a significant story from the tech industry. For more AI insights, follow Hayden Field. The Stepback is delivered to subscribers at 8AM ET; you can opt in here. The idea of AI agents began, in part, with fictional representations like J.A.R.V.I.S. from the Marvel movies. Though not the origin, J.A.R.V.I.S. has influenced the vision of agentic AI — an intelligent assistant understanding user needs. However, despite this inspiration, such AI agents remain largely conceptual and are not yet ready for widespread, practical use. For the complete story, visit The Verge.
Anthropic has agreed to a proposed class action settlement with US authors accusing the AI company of copyright infringement. The lawsuit alleged Anthropic's AI models were trained on pirated books. The settlement allows Anthropic to avoid a trial that could have resulted in significant legal penalties. The case stemmed from claims that Anthropic used pirated materials in training its Claude AI models. Previous court rulings favored Anthropic on fair use of legally purchased books but the class action suit proceeded. The settlement is expected to be finalized by September 3rd, with details to be announced soon.
In July 2023, I decided to delete the Day One journaling app from all my devices, marking a significant change for me as a longtime diarist. This choice was influenced by Apple's announcement of its Journal app at WWDC 2023, which planned to use on-device machine learning to create prompts based on personal data like contacts, photos, music, workouts, podcasts, and location. The idea felt intrusive and unsettling, especially since Apple's Memories feature once resurfaced a deeply personal and painful photo. This experience highlighted how integrating AI into journaling can intrude on the personal and reflective nature that makes journaling meaningful.