Mark Zuckerberg just revealed plans to bring a subscription tier and product ads to the standalone Meta AI app, targeting power users who want more compute. Meta also raised its AI spending outlook to as much as $72 billion this year. Why it matters: Meta wants a direct rival to ChatGPT and Copilot, but monetization will take a back seat to user growth for at least another year.
Meta just updated its privacy policy for Ray-Ban smart glasses, quietly removing the toggle that allowed users to stop voice recordings from being stored in the cloud. Audio logs and transcripts will now be kept for up to a year to train Meta AI models. While you can still delete recordings manually in settings, you can no longer opt out entirely.
Meta just launched a standalone Meta AI app powered by Llama 4, putting hands-free voice conversations front and center. The app absorbs the Ray-Ban Meta companion controls, includes a native full-duplex voice demo, and taps into your Facebook and Instagram activity for personalized answers. It basically elevates Meta AI from an in-chat bot into a full-blown daily companion.
Google just made Meta’s Llama 4 models generally available as a fully managed API on Vertex AI. Developers can now run Llama 4 Scout and Maverick via standard OpenAI-compatible endpoints without provisioning GPUs or managing infrastructure. Why it matters: it brings Meta's flagship open models straight to enterprise cloud pipelines with zero operational overhead.
Meta and Booz Allen just launched a fine-tuned version of Llama 3.2 aboard the International Space Station. Dubbed "Space Llama," the multimodal system runs entirely offline on HPE space computers with Nvidia acceleration, slashing processing time from minutes to just over a second. Why it matters: It proves open-source edge AI can keep critical missions running in places where the cloud simply can't reach.
OpenAI is reportedly planning a "handoff" feature for its upcoming open reasoning model, letting it ping larger cloud models when local queries get too tricky. The model is being trained from scratch for an early summer launch to challenge DeepSeek R1. It is a clever move that bridges open weights with OpenAI's paid API ecosystem.
Meta just rolled out live translation to all Ray-Ban smart glasses users across supported markets. The feature enables real-time two-way translation between English, French, Italian, and Spanish—and by downloading language packs in advance, it works completely offline without Wi-Fi or cellular data. It turns the smart frames into a genuinely practical travel translator.
Apple CEO Tim Cook is reportedly obsessed with shipping true augmented reality glasses before Meta can pull it off. According to Bloomberg's Mark Gurman, Cook views the Vision Pro as merely a stepping stone toward lightweight specs, making the project his personal top priority. It's still years away given battery and display hurdles, so Apple may drop basic AI glasses first to stay competitive.
James Cameron just shared his vision for generative AI in Hollywood, arguing that VFX-heavy blockbusters must use AI to cut budgets in half by doubling production speed rather than firing crew. He also pushed back on training data debates, insisting regulators should judge whether final outputs plagiarize instead of policing what models learn from.
San Francisco AI startup Deep Cogito just emerged from stealth and open-sourced Cogito v1, a lineup of hybrid reasoning models ranging from 3B to 70B parameters. Trained on top of Llama with an AlphaGo-style self-play loop, the models toggle between instant answers and deep reflection while adding native tool calling.
Nvidia just released Llama-3.1-Nemotron-Ultra-253B-v1, a fully open-source dense model built on Meta's 405B base. By overhauling the architecture to fit on a single 8x H100 node, it edges out the massive 671B DeepSeek R1 on benchmarks like GPQA and LiveCodeBench, though DeepSeek still leads in heavy math.
LMSYS just dropped over 2,000 battle logs for Llama-4 on Arena and took aim at Meta for transparency issues. The evaluation platform revealed that Meta failed to disclose its experimental Maverick model was heavily tuned to game human preference prompts, prompting an immediate overhaul of leaderboard policies.
Meta’s head of GenAI just denied rumors that Llama 4 Maverick and Scout were trained on test sets to game benchmark scores. The speculation flared over the weekend after users noticed performance gaps across platforms. Al-Dahle blamed the mixed quality on early rollout issues across cloud providers, promising bug fixes over the coming days.
A Meta AI insider just blew the whistle on Llama 4, claiming the model lagged far behind rival open-source models and leadership pushed to blend benchmarks into post-training to fake results. The researcher submitted their resignation in protest and claimed Meta's VP of AI also stepped down for the same reason.
Meta just unveiled Llama 4 Scout and Maverick, marking its official pivot to open-weight, natively multimodal mixture-of-experts architectures. Both models operate on 17B active parameters, with Scout pushing context windows to a massive 10 million tokens, while an upcoming 2T-parameter teacher model, Behemoth, trains in the background.
OpenAI's GPT-4.5 was judged human 73% of the time in a new preprint Turing test study from UC San Diego researchers, easily beating actual humans when given a persona prompt. Without a persona, its success dropped to 36%, showing LLMs excel at convincing mimicry in short interactions.
Wikimedia Foundation just revealed that multimedia download bandwidth on Wikimedia Commons has surged 50% since January 2024, driven entirely by AI scrapers harvesting training data. While bots make up only 35% of pageviews, they account for 65% of the most resource-heavy traffic by digging up uncached, obscure files.
Meta just signed a multimillion-dollar, multi-year deal with UFC to become its official fan technology and wearable partner. The integration spans Meta AI, smart glasses, Quest headsets, and exclusive Threads content, with plans to build an AI-powered fighter ranking system. Why it matters: Meta gets massive mainstream sports visibility for its next-gen hardware and AI directly inside the Octagon.
Meta is reportedly releasing its first own-brand smart glasses, codenamed Hypernova, in late 2025 with a price tag between $1,000 and $1,400. The device adds a monocular heads-up display to the right lens and bundles an sEMG neural wristband for pinch-to-select gesture navigation. Why it matters: It bridges the gap between audio frames and full AR, testing whether consumers will pay top dollar for a wearable screen.
Joëlle Pineau, a central figure in Meta's AI research and longtime leader of FAIR, just announced she will depart the company on May 30 after nearly eight years. During her tenure, she helped steer milestone open-source projects including PyTorch, Llama, and SAM. Her exit comes amid rising geopolitical tensions and shifting internal priorities, marking a major transition for Meta's fundamental research team.
