706: No One Is Getting a Good Deal Tonight

The episode begins with a deep dive into Apple’s announced hardware updates, including new Mac minis and Mac Studios featuring M6 and M5 chips respectively. The hosts discuss the technical specifications of the M6, noting it as Apple's first 2nm chip with doubled neural engines and various core configurations, while the M5 Ultra and Max are examined for their chiplet-based design and performance scaling. Particular attention is paid to the pricing strategies, the configurations available, and the limitations in options, especially around RAM and storage. They also analyze the implications of these releases for power users and professional workflows, emphasizing that the ultra configurations are still just two max chips glued together, which influences their performance scaling and pricing. The discussion touches on the support for multiple displays, connectivity features, and how these updates fit into Apple's broader product and software ecosystem, including the timing of macOS 27's release.'

The podcast covers recent beta updates to macOS 27, highlighting subtle UI tweaks such as redesigned window control stoplights in Beta 6, with a humorous comparison to car headlights. They also discuss the new Preview icon and changes to Siri’s logo, as well as the broader design philosophy Apple seems to be adopting—modern, glassy aesthetics with refined liquid glass effects. The hosts express optimism about Apple's design direction, noting the improvements over previous liquid glass implementations. The discussion also includes AI-related insights, like the non-deterministic nature of language models at zero temperature and their hardware implications, referencing a detailed blog post on floating-point arithmetic. Additional updates include shared albums in Photos consuming iCloud storage and display support improvements, with commentary on the state and evolution of Mac UI components.

The core of the episode delves into the technical details of Apple’s new chips, especially the M6, M5 Max, and M5 Ultra. The hosts analyze core configurations, performance claims, and the new ARM scalable matrix extension (SME) that replaces the private AMX instructions. They discuss implications for AI workloads, highlighting how the new chips improve neural engine capabilities, GPU performance, and memory bandwidth. The conversation includes technical details like core sizes, performance scaling, and the challenges of modern chip design. They also speculate about Apple’s strategic shift away from ultra-defines as just glued-together max chips, and consider upcoming future chips, like the rumored M7, as optimal for their long-term AI computing goals.

Further discussion explores the high costs of Apple’s updated Mac mini and Mac Studio configurations, including detailed pricing comparisons and inflation adjustments. They touch on the limitations in options, with multithousand-dollar upgrades for larger RAM and SSD configurations, and the practical impact on consumers. The hosts analyze Apple’s display support enhancements, such as higher refresh rates and multiple external displays, and comment on the scarcity and pricing of high-end storage solutions. There’s also an examination of Apple’s pricing strategies, their focus on cloud and AI, and how these trends influence consumer options and perceptions of value.

A significant portion of the episode is dedicated to the hosts’ personal experiences with vibe coding, creating their own utilities and apps for self-use with minimal effort and advanced AI assistance. John describes building a custom window management app, a file decrypter for Audible, and automation tools, all through minimal prompts and AI guidance. They emphasize how AI enables even non-professional developers to create complex, useful software, lowering barriers for individual innovation. The discussion also covers their strategies for releasing private, non-sandboxed apps alongside App Store versions, and the ethical and practical considerations around using AI for development, including examples of progress, frustrations, and the potential for AI to transform personal productivity.