Meta

Meta rolls out consumer subscription plans for Instagram, Facebook, and WhatsApp globally, with prices from $2.99 to $3.99 per month, offering extra features. The company also begins testing new subscriptions for businesses, creators, and Meta AI users.

A new analysis shows that top AI forecasters adjust their AGI timelines based on which lab is currently leading the field, with predictions swinging from earlier to later and back again as the dominant lab changes from ChatGPT to xAI/Meta/Gemini to Anthropic.

This article dives deep into Ollama's configuration engine, covering how to fine-tune local language model parameters using the Modelfile, optimize hardware performance with server environment variables, and format prompt flows with Go template syntax.

Meta is rolling out paid add-ons for Instagram, Facebook, and WhatsApp worldwide while building a separate paid AI offering. This marks the first time Meta has clearly monetized its AI investments.

ElevenLabs has released Music v2, an upgraded AI music generation model that can shift between genres like opera, heavy metal, and rap within a single song. A new inpainting feature allows users to regenerate specific sections without affecting the rest.

A group of former researchers from Google DeepMind, Apple, OpenAI, and Meta have launched a startup called Trajectory, aiming to help companies continuously improve their AI products by training on real-world user interactions. The company has raised a $15 million seed round at a $115 million valuation, led by Conviction. Trajectory's platform enables continuous learning for AI models, updating them based on real-world failures. It currently works with AI-native companies like Clay and Harvey, and plans to expand to Fortune 500 companies.

The 10th ABAW Workshop and Competition at CVPR 2026 advances multimodal human-centered AI by introducing new challenges including emotional mimicry intensity estimation, ambivalence/hesitancy recognition, and fine-grained violence detection, alongside traditional affect estimation and recognition tasks. The competition leverages large-scale in-the-wild datasets, and the paper track covers a broad range of topics from pose estimation to fairness and robustness.

ICG is a novel framework that integrates MLLM-based prompting with personalized preference alignment to generate high-quality, contextually relevant cover images. It extracts semantic features via meta tokens, refines them with user embeddings, and injects personalized context into diffusion models. A multi-reward learning strategy combines public rewards with a personalized preference model, eliminating the need for labeled supervision. Experiments show improvements in image quality, semantic fidelity, and personalization, boosting user appeal and recommendation accuracy.

This paper presents $E^3$-Agent, an executable and evolving agent for resource management of edge AIGC. It separates a fast-path router from a slow-path LLM meta-controller, learns online from execution feedback, and adapts to unknown time-varying service-time mappings. Evaluation shows 65%-73% latency reduction over static baselines and effective stutter suppression.

A paper argues that with generative AI dissolving the human capacity to write correct code as the binding constraint, software work reorganizes around two pillars: Mixer Mode (humans operating multiple judgment axes continuously like a sound engineer) and Meta-Software (software that observes, validates, and governs other software). The two pillars are inseparable, drawing a parallel to the historical transition from artisanal to mass production.

South Africa holds 88% of global platinum-group metals, hosts Africa's largest data center market, and sits at the center of a US-China AI infrastructure contest. Yet its draft AI policy, withdrawn after hallucinated references, fails to leverage these advantages for favorable terms. The article examines South Africa's structural leverage, three possible AI infrastructure futures (Chinese, US, local open-weight), and the need for binding governance provisions.

Unionized staff at The New York Times' Tech Guild accuse management of refusing to disclose AI usage plans and using internal AI tools to monitor performance, leading to unfair labor practice charges. The dispute highlights broader industry tensions over AI in newsrooms.

ACM CAIS 2026 registration is full; join the waitlist. The conference runs May 26–29, 2026 in San Jose, featuring keynotes, 63 research papers, and 46 system demos, with a partnership with the AI Engineer World's Fair.

This article details how to deploy a fully local voice conversation pipeline for the Reachy Mini robot, eliminating the need for cloud servers or API keys. It uses a cascaded approach combining VAD, STT, LLM, and TTS, with recommended defaults: llama.cpp with Gemma 4, Silero VAD, Parakeet-TDT 0.6B v3 STT, and Qwen3-TTS. Various LLM options are provided, including local MLX, Transformers, vLLM, or remote Responses API.

Save your money, your privacy, and the planet. This installable AI offers several benefits you won't find with more traditional models like ChatGPT.

Local models offer privacy, cost savings, control, and availability. While not as capable as frontier models, they are improving. This post explains how to set up local models in Zed using LM Studio, Ollama, or llama.cpp, and offers tips for effective use.

This systematic review of 139 studies proposes a unified framework and meta-analysis. Results show multimodal fusion improves accuracy by 5.28% on average, multiview fusion boosts accuracy by 4.67% and F1 by 3.08%, but only a minority of studies used statistical tests, raising reproducibility concerns.

The software industry is undergoing an unprecedented metamorphosis. From simple statistical completion of early coding assistants, through conversational chatbots and the failure of multi-agent systems, we have arrived at the era of the Agentic Loop. This comprehensive guide analyzes the entire evolution, from the Completion paradigm to the revolutionary Ralph Loop that is redefining how we write code.

Curlo is a privacy-first macOS app for searching, previewing, and organizing large sound libraries. It lets you find SFX or music by describing what you want to hear, with fully local offline search, metadata management, AI auto-tagging, and integration with DAWs.

Researchers from UMD, Google, Meta, and other institutions use AutoTTS to let a coding agent independently discover control algorithms for AI reasoning. The algorithm it found cuts compute by about 70 percent compared to standard self-consistency while matching its accuracy. The whole search cost $40 and took 160 minutes.

Meta cuts 350 jobs in Ireland amid global layoffs, sparking concerns over AI-driven displacement. Amazon, Oracle, Block also cite AI for job cuts. Experts warn of 'AI washing' as some layoffs may be unrelated. Official data shows 10.7% drop in Irish tech employment. ESRI report warns 200,000 roles could be displaced by AI.

Meta launched an internal AI leaderboard called 'Claudeonomics' to track employee token usage, but shut it down after data leaked. The trend of tracking AI usage is growing, with Nvidia's Jensen Huang proposing AI tokens as part of compensation.

We chat with Daytona's CEO about their insane 74% MoM Growth, 850K Daily Runs, Bare Metal Sandboxes, RL Evals, and the New Agent Cloud

New UC San Diego research shows that with persona prompts, advanced LLMs like GPT-4.5 are judged human more often than real humans in a three-party Turing test. LLaMa-3.1 also performed at human levels. The study raises important questions about online trust and the nature of humanlikeness.

The proliferation of LLM adversarial testing tools has made it difficult for operators to navigate the landscape. Autonomous agents that select attacks, compose transforms, and execute tests are emerging as a solution. Dreadnode's agent executed 674 attacks against Meta's Llama Scout in about three hours with an 85% success rate. However, limitations include constrained coverage, alignment issues with the orchestrating agent, and a lack of formal comparison with human experts.

3M Users, 100K Signups/Week, Own-Metal Data Centers, $200K+ Coding Agent Spend, and the Death of PRs

Tribal announced a $10 million seed round led by Team8 to expand its metadata-native AI agent platform, which maps enterprise systems' metadata layers to deliver production-ready solutions with governance. The platform integrates with Salesforce, SAP, and other major systems, aiming to turn AI experiments into repeatable business impact.

We're about to find out if systems like SynthID and C2PA can effectively combat deepfakes. Google expands SynthID verification to Chrome and Search, also checking C2PA metadata. OpenAI adds SynthID to its images. Meta will use C2PA on Instagram. But challenges remain: metadata easily stripped, open-source models not adopting, and Google's dual role as AI creator and solution provider.

The author's latest demo of Google's Android XR reference glasses and Project Aura shows Gemini-powered capabilities like adding events to calendar, editing photos, and cross-app integration, suggesting ambient AI is the future of wearables.

Exclusive: Employment tribunal claim says worker lost his job after distributing leaflets throughout London office

Researchers introduce EgoTraj, an egocentric multimodal dataset recorded with Meta Quest Pro, featuring 75 navigation sequences in real-world urban environments with synchronized RGB video, head poses, eye gaze, and scene annotations. The dataset aims to advance trajectory prediction for humanoid robotics, wearable sensing, and assistive navigation, and benchmarks multiple state-of-the-art methods.

As deep research agents increasingly automate complex information-seeking tasks, reliable evaluation becomes critical. LLM-as-judge is used to assess these agents, but its reliability is poorly understood. The REFLECT benchmark introduces a fine-grained meta-evaluation, revealing that current LLM judges are unreliable, with best models achieving accuracies below 55% across reasoning, tool-use, and report-quality failures, especially poor on evidence verification. The study offers actionable guidance for building more reliable evaluation pipelines.

Meta is forcing over 7,000 workers to move to new teams, including teams focused on AI cloud infrastructure and an internal AI agent codenamed Hatch, as it pivots to AI. Last month, 1,000 engineers were reassigned to Applied AI, with employees told transfers are mandatory.

OpenAI now uses C2PA metadata and SynthID watermarks to tag AI-generated images, making them harder to strip. A public verification tool is also rolling out.

A quiet day before Google I/O lets us amplify a notable blogpost by Vlad Feinberg on job preparation for pretraining roles at frontier AI labs, alongside major model releases (Cursor Composer 2.5, Qwen3.7), local inference speedups via MTP in llama.cpp, and research on MoEs, RL, and agent evaluation.

Meta begins another round of layoffs cutting 8,000 jobs while ramping up AI investments. Employee morale is low, with internal resistance to AI data tracking tools.

Meta committed $145B to AI infrastructure and began laying off 8,000 people in the same week. Standard Chartered cut over 7,000 jobs, explicitly replacing 'lower-value human capital' with AI investment. Pope Leo XIV announced his first AI encyclical co-authored with Anthropic's Christopher Olah, to be released May 25.

This study introduces intelligent frameworks that use Large Language Models (LLMs) to improve task scheduling for construction robots. The system utilizes a Natural Language Processing interface for communication and adapts in real-time to unexpected site conditions. It concurrently uses two LLM agents: a generator (GPT-4) and a supervisor (Gemma 3/Llama 4/Mistral 7b) to provide precise task schedules. Evaluation results highlight the crucial role of LLMs in construction robotic tasks.

This paper proposes Deep Pre-Alignment (DPA), a novel architecture that replaces the standard ViT encoder with a small VLM as perceiver, ensuring visual features are deeply aligned with the text space of the target large language model. On the 4B parameter scale, DPA outperforms baselines by 1.9 points across 8 multimodal benchmarks, with gains widening to 3.0 points at the 32B scale. DPA also achieves a 32.9% reduction in language capability forgetting and demonstrates consistent gains across different LLM families like Qwen3 and LLaMA 3.2.

The paper reveals that quantizing LLMs to lower precision (e.g., 3-bit) causes 6-21% of previously unbiased items to exhibit new stereotypical behaviors, while models' willingness to answer 'unknown' drops by 17.4%. Standard metrics like perplexity barely change, indicating aggregate metrics miss fairness-critical degradation.

Instruction-tuned language models appear fair in high-stakes decisions but harbor biased internal representations. Using mortgage underwriting as a test case, this study shows that while outputs show no bias, internal layers encode and amplify demographic associations. Activation steering and cross-layer interventions reveal that reintroducing suppressed information at critical layers can flip decisions almost entirely. The latent bias is asymmetric, direction-dependent, and exploitable via adversarial prompts or fine-tuning. The authors argue that output-only audits are insufficient and propose a dual-layer testing framework combining output evaluation with representational analysis for AI governance.

Nous Research has published Lighthouse Attention, a training-only selection-based hierarchical attention mechanism that wraps around standard scaled dot-product attention during pretraining and is removed afterward. Unlike prior methods that pool only keys and values, Lighthouse pools Q, K, and V symmetrically across a multi-resolution pyramid, reducing the attention call from O(N·S·d) to O(S²·d) and running stock FlashAttention on a small dense sub-sequence. Tested on a 530M Llama-3-style model at 98K context, it achieves a 1.40–1.69× end-to-end wall-clock speedup against a cuDNN SDPA baseline with matching or lower final training loss.

Annota is a local-first, end-to-end encrypted note-taking app with a powerful editor, AI assistant, and absolute privacy. It supports local Ollama or BYOK AI. All features are free for local use, with subscriptions for cloud sync and multi-device support.

LocalVibe is a pure-Rust local AI coding assistant optimized for Apple Silicon. It provides chat with quantized LLMs via Metal, on-device ONNX embeddings, LanceDB vector search, and a TUI interface. It includes MCP server for Claude Code, an OpenAI-compatible HTTP server, and various tool integrations.

An open-source benchmark suite that runs a full GPU/CPU benchmark with one command, covering Ollama LLM inference and XGBoost training, and generates an interactive HTML report.

ArXiv, a popular platform for preprint academic research, is taking a new step to attempt to reduce the volume of papers that include AI slop. If a paper has "incontrovertible evidence that the authors did not check the results of LLM generation," such as hallucinated references or "meta-comments" left by an LLM, authors will be banned from ArXiv for a year, according to Thomas Dietterich, ArXiv's section chair of its computer science section. Future ArXiv submissions will also have to be accepted at "a reputable peer-reviewed venue."

A new AI framework can detect microscopic internal defects in metal 3D printed components that are invisible to the naked eye but compromise structural integrity, potentially overcoming a key barrier to widespread adoption of metal additive manufacturing.

Meta is installing tracking software on US-based employees' computers to capture mouse movements, clicks, and keystrokes for AI training. The tool, MCI, also takes occasional screenshots. Meanwhile, Meta plans to lay off 10% of its workforce globally starting May 20.

Tech companies are increasingly targeting middle managers in AI-driven restructurings, with Coinbase, Amazon, Block, and Meta cutting layers. Workers fear loss of mentorship and career advancement.

AI ethics is the practice of ensuring that AI systems are developed and used in ways that are fair, transparent, accountable, and safe. This article explores the core questions, challenges, regulatory landscape, and practical steps for responsible AI.