Models & Labs
Anthropic launches Claude Docs and Slides
The Verge AIhigh confidence
Why it matters
- →Anthropic directly competes with Google's integrated AI productivity tools by offering native document creation.
- →The consolidation of Cowork, Design, and new Docs/Slides into one interface simplifies the user workflow.
- →Real-time collaboration features allow multiple users and Claude to edit files simultaneously.
Anthropic has launched beta versions of Claude Docs and Slides, allowing users to create, edit, and collaborate on documents and presentations directly within the chat interface. The update consolidates previous separate tools like Cowork and Design into a unified 'one Claude' experience, available initially for Pro and Max subscribers on web, desktop, and mobile. These new features aim to compete with Google's Gemini-integrated productivity suite by enabling real-time collaboration and native file generation. Access for Team and Free users will follow in subsequent weeks.
Read originalMore from The Verge AI
© The Verge AIGeneral AIproductivity
Snap launches Specs Intelligence AI assistant
Snap is entering the crowded personal AI assistant market with Specs Intelligence, an anticipatory service designed to manage work and travel tasks by connecting to apps like Gmail and Slack. Unlike standard chatbots, it aims to surface relevant actions before you ask, leveraging a mix of open-source models and local LLMs hosted in the US. The tool launches on iOS today with a Mac preview coming later, positioning itself as a companion to Snap’s new AR glasses. While privacy assurances promise no data training or ad targeting, the real test is whether users will trust a social media company with deep access to their personal routines and priorities.
The Verge AI
© The Verge AIMarket & Regulationother
AI e-waste projected to fill 23M containers by 2050
A new Basel Action Network report exposes the hidden physical cost of AI, projecting that data center infrastructure will generate up to 617 million metric tons of e-waste by 2050. Unlike previous studies focusing only on GPUs, this analysis includes cooling, power, and networking gear, revealing that AI could account for 15-20% of all global electronic waste. The scale is staggering: enough trash to circle the world six times if packed into shipping containers. This shifts the narrative from AI's carbon footprint to its tangible toxic legacy, highlighting a massive gap in recycling infrastructure and policy.
The Verge AI
© The Verge AIMarket & Regulationbusiness
Apple Plans ARM-Based AI Servers with Nvidia Tech
Apple is reportedly planning a return to the server market by 2029, targeting the booming demand for efficient AI compute. The strategy hinges on pairing its high-performance M-series chips, likely the M8 Ultra, with Nvidia’s NVLink Fusion technology to enable multi-chip scaling. This move signals a shift from Apple's traditional consumer focus to enterprise infrastructure, leveraging the popularity of Mac Studios among developers who have faced hardware shortages. It represents a significant strategic pivot, potentially challenging established data center players by offering an alternative to x86 architectures.
The Verge AI
More in Models & Labs
Models & Labsmodels
llama.cpp adds support for DFM Mimir 1B HRM architecture
llama.cpp now runs the DFM Mimir 1B model, a rare example of a hierarchical recurrent transformer that alternates between two distinct processing cycles. This isn't just another standard decoder; it requires handling aliased cache slots and looped graph builds, pushing the runtime's architectural flexibility further than usual. While the decode speed is roughly four times slower than a dense model of similar width due to these 128 block passes, local inference for this specific architecture was previously impossible on consumer hardware. This release effectively opens the door for experimenting with non-standard transformer topologies in open-source environments.
llama.cpp Releases
Models & Labscoding
llama.cpp b11007 enables CUDA graphs for MTP
This release quietly patches a major performance bottleneck in speculative decoding. By enabling CUDA graphs for Multi-Token Prediction (MTP), llama.cpp reduces kernel launch overhead on NVIDIA GPUs, making draft-token generation significantly faster. It also updates CUDA builds to support versions 12.8 and 13.3, keeping pace with the latest driver stacks. While KleidiAI is temporarily disabled on Apple Silicon, the focus here is squarely on squeezing more throughput out of existing hardware for local inference.
llama.cpp Releases
© TechCrunch AIModels & Labsother
Meta develops camera-free AI glasses 'Luna'
Meta is quietly addressing the privacy backlash against its Ray-Ban smart glasses by developing a new model called Luna that strips away cameras entirely. This device relies on six microphones and a side button to interact with Meta's AI chatbot and Muse agent, positioning audio-first interaction as a viable alternative to visual surveillance. By decoupling AI assistance from constant recording, Meta is testing whether users will adopt wearable AI if the social stigma of being watched is removed. It signals a strategic pivot toward privacy-preserving hardware to broaden adoption beyond early adopters comfortable with always-on cameras.
TechCrunch AI