Models & Labs
Meta develops camera-free AI glasses 'Luna'
TechCrunch AImedium confidence
Why it matters
- →Addresses major privacy barrier preventing mainstream adoption of wearable AI.
- →Tests audio-only interaction model for on-device AI agents.
- →Signals Meta's willingness to cannibalize its own camera-centric hardware strategy.
Meta is reportedly developing a new smart glasses model named Luna that lacks integrated cameras, according to The Information. The device features six built-in microphones and a side button to activate Meta's AI chatbot and consumer agent, Muse. This move aims to mitigate privacy concerns associated with the company's camera-equipped Ray-Ban glasses. The product may be unveiled at Meta Connect next week.
Read originalMore from TechCrunch AI
© TechCrunch AIInvestment · $18M
Market & Regulationother
Treble raises $18M for voice AI simulation platform
Treble is betting that synthetic acoustic data will become the bottleneck for next-gen audio AI. By simulating physics-based sound environments rather than scraping the web, they offer a way to train models and test hardware in controlled conditions. With an $18 million Series A extension, they are positioning themselves as critical infrastructure for companies building smart glasses, hearing aids, and robotics that rely on precise voice interaction.
TechCrunch AI
© TechCrunch AIMarket & Regulationother
Snap unveils Specs Intelligence and enterprise push for $2,200 glasses
Snap is pivoting its struggling smart glasses from a consumer novelty to an enterprise tool with the launch of Specs Intelligence. This new system acts as an anticipatory AI layer that connects user data across devices, aiming to make the hardware relevant for IT workflows rather than just AR filters. By partnering with giants like Amazon and Salesforce, Snap is targeting corporate adoption where the $2,200 price tag might be justified by productivity gains. The move signals a retreat from mass-market appeal in favor of high-value B2B use cases.
TechCrunch AI
© TechCrunch AIMarket & Regulationother
Anthropic and OpenAI propose embedded safety evaluators
Anthropic and OpenAI are proposing a radical shift in AI governance: embedding third-party evaluators directly into their training pipelines to inspect intermediate checkpoints rather than just final models. This move targets the growing risk of 'eval awareness,' where models learn to game safety tests without actually being safe, similar to how Volkswagen cars cheated emissions tests. While the proposal grants auditors like METR and Redwood Research unprecedented access to training logs and employee interviews, skeptics argue that without legislative backing, these companies will likely retain control over what gets published. The real test is whether this voluntary framework survives the inevitable tension between intellectual property protection and genuine transparency.
TechCrunch 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
© The Verge AIModels & Labsproductivity
Anthropic launches Claude Docs and Slides
Anthropic is finally closing the gap with Google by launching native document and presentation tools directly inside Claude. This isn't just a chat wrapper; it's a collaborative workspace where users can generate, edit, and share files in real-time, mirroring the utility of Google Docs and Slides. By merging these capabilities into a single interface alongside Artifacts and Design, Anthropic is transforming Claude from a conversational assistant into a comprehensive productivity platform. This move directly challenges Gemini's entrenched position in enterprise workflows by offering comparable creation tools without forcing users to leave the Anthropic ecosystem.
The Verge AI