All articles
DreamLite: ByteDance's model generates images on-device in 3 seconds
ByteDance open-sources DreamLite — a lightweight 0.39B diffusion model that is the first known on-device model combining text-to-image generation and text-guided image editing in a single network. On iPhone 17 Pro, inference time for a 1024×1024 image is approximately 3 seconds with no cloud connection.
Persona AI and Under Armour Explore Performance Materials for Humanoid Robots
Persona AI, a maker of industrial humanoid robots, announced an early-stage R&D collaboration with Under Armour on May 11, 2026. The goal is to study how advanced performance textiles can improve the durability and resilience of humanoids operating in welding shops and other extreme industrial environments.
Thinking Machines debuts: TML-Interaction-Small merges intelligence with real-time responsiveness
Mira Murati's startup has released its first foundational model — TML-Interaction-Small. The 276B-parameter MoE model is trained from scratch to handle audio, video, and text in 200-millisecond micro-turns, without any external dialog-management components.
Proteins without secrets: a roadmap for explainable biological AI
A European research team published in Nature Machine Intelligence a roadmap for explainable protein AI — a taxonomy of XAI methods spanning four analysis layers and five interpretability roles, with emphasis on biosafety in generative protein design.
RLDX-1: a foundation model for robot hands with memory and force sensing
RLWRLD has released RLDX-1 — a foundation model built from the ground up for high-DoF robot hands. The model integrates context memory, force sensing, and motion processing into a single architecture, achieving up to 90% success rates on tasks requiring physical signals and task history.
OpenAI Launches DeployCo: $4B to Deploy AI Across Enterprise
OpenAI announced on May 11, 2026 the launch of the OpenAI Deployment Company (DeployCo) — a dedicated AI deployment firm for enterprises, backed by over $4 billion in initial investment and the acquisition of consulting firm Tomoro.
NVIDIA Maps Robotics' Future on the LLM Evolution Path
NVIDIA's Jim Fan at AI Ascent 2026 outlined a detailed robotics development roadmap based on replicating four stages of large language model evolution — from pre-training to autonomous research.
Cloudflare cuts 1,100 jobs citing AI efficiency as Q1 revenue hits record $639.8M
Cloudflare announced its first-ever mass layoff — cutting 1,100 employees (~20% of staff) — while reporting record Q1 2026 revenue of $639.8M. CEO Matthew Prince cites a 600% surge in internal AI usage as the direct driver.
Anthropic Mythos Helped Firefox Fix 423 Bugs in One Month
Mozilla disclosed that Anthropic's Mythos model uncovered hundreds of high-severity security bugs in Firefox — including sandbox vulnerabilities that had lurked in the codebase for over a decade. In April 2026 alone, the browser shipped 423 bug fixes, compared to just 31 in the same month a year earlier. Mozilla distinguished engineer Brian Grinstead confirms that Mythos finds sandbox bugs at a volume human researchers never matched.
OpenAI Launches GPT-Realtime-2: Voice Intelligence with GPT-5-Class Reasoning
On May 7, 2026, OpenAI released three new audio models in its Realtime API: GPT-Realtime-2 with GPT-5-class reasoning, GPT-Realtime-Translate supporting over 70 input and 13 output languages, and GPT-Realtime-Whisper for streaming speech-to-text. This is the first generation of voice models capable of simultaneous reasoning, translation, and transcription in real time.
Genesis AI unveils GENE-26.5: a full-stack approach to robot dexterity
Genesis AI announced its GENE-26.5 foundation model and a proprietary robotic hand designed to overcome the data bottleneck in robotic manipulation. The startup, backed by $105 million, plans to reveal its first general-purpose robot.
Google DeepMind Invests in EVE Online Studio, Rebranded as Fenris Creations
On May 6, 2026, CCP Games rebranded as Fenris Creations and completed a $120M buyout from Pearl Abyss. Google DeepMind took a minority stake and entered a research partnership to train and evaluate AI models using an offline version of EVE Online.