by 
Welcome back to our weekly roundup of happenings from XR and AI realms. Let’s dive in…
Feeling Spatial
Qualcomm Inks Multi-Year Deal to Power Consumer Specs and Future Devices from Specs Inc. The timing for this move is right as Snap’s Specs Inc spinnoff is committed to releasing consumer-geared Spectacles – known simply as Specs in their current development phase – sometime this year. This will be the first consumer-targeted variation on the erstwhile developer-focused Spectacles. Moreover, Specs will be processing-intensive. While a large swath of the consumer headworn AR market has moved to a toned-down visual UX – a shortcoming compensated through AI functions – Specs are moving in the opposite direction with a highly immersive UX. Snapdragon XR will be the engine behind it.
The AI Desk
Anthropic Holds Back Mythos After Safety Tests. Anthropic has withheld the release of its Claude Mythos Preview after internal testing showed the model could identify thousands of software vulnerabilities across major operating systems and browsers. During testing, the model exceeded its constraints, contacted a researcher outside its environment, and attempted to conceal its actions. Systems trained for coding are effective at finding and exploiting flaws, making them useful for both defense and attack. Anthropic is sharing the model with select partners to strengthen defenses, but other developers are unlikely to slow down. The window to adapt security infrastructure is narrowing.
Anthropic’s Mythos move follows a recent leak of Claude Code, which exposed more than 500,000 lines of source code through a mistakenly published file. The material revealed internal architecture, agent workflows, and a backlog of unreleased features, including tools for autonomous coding and persistent task management. It also clarified Anthropic’s direction toward long-running agents capable of managing complex software projects. The leak raised concerns about internal security while offering competitors a detailed view of the system.
Meta Reboots AI With Muse Spark. This week they introduced Muse Spark, the first major model from its Superintelligence Labs unit, led by Alexandr Wang, following Meta’s $14.3 billion investment in Scale AI. The release comes after developer disappointment with Llama 4, which prompted Meta to rebuild its AI stack rather than iterate on existing models. Muse Spark is positioned as a smaller, faster model capable of handling complex reasoning tasks in science, math, and coding. Access is limited to a small group of partners, with plans to expand through a paid API. Meta claims the system was developed in nine months, a compressed timeline compared with typical multi-year model cycles. Progress or competition? Or both? Also, wasn’t Spark the name of their AR developer platform?
Can tools from Meta, Grok, and others keep up? What about DeepMind and other Chinese companies? The timeline raises questions about whether the acceleration reflects technical progress or competitive pressure. Anthropic, Google, and OpenAI have established strong positions with developer adoption around Claude Code and Gemini. Meta is attempting to reenter that race with a new architecture and a different pace of development.
California Creative Jobs Continue to Decline, but not because of AI. A new report cited by The Hollywood Reporter finds California has lost a significant share of creative jobs, with declines across film, television, and digital production as work shifts to other regions. AI is still a small part of the shift, particularly in post production, design, and entry-level roles, where automation is beginning to reduce demand. Other factors remain significant, including rising production costs, tax incentives outside California, and continuing contraction following the streaming expansion.
Cinematic Corner
See the latest episode of Linda’s Last Podcast below. Produced by the author of this column, you can see the series archive and subscribe on YouTube.
Spatial Audio
This column is also the script for the AI/XR Podcast, hosted by the author of this column, Charlie Fink, and Ted Schilowitz, former studio executive and futurist for Paramount and Fox, and Rony Abovitz, founder of Magic Leap and Synthbee AI. This week, our guest is Ian Hamilton, VR expert and writer, who was formerly the editor of Upload VR. Episodes drop on Tuesdays, and you can find them on podcasting platforms Spotify, iTunes, and YouTube.
Charlie Fink is an author and futurist focused on spatial computing. See his books here. Spatial Beats contains insights and inputs from Fink’s collaborators, including Paramount Pictures futurist Ted Shilowitz.
Header image credit: Solen Feyissa on Unsplash