OpenAI’s Atlas browser blends browsing with in-app AI support, offering page summaries, automated tasks, and memory for macOS users initially.

By 2025, AI-written articles make up over half of new online content, but human-crafted pieces remain preferred and more visible in search results.

Perplexity offers direct, cited answers with conversational context, saving time and tabs, but Google remains best for technical searches or shopping.

Ollama 0.12.6-rc0 adds experimental Vulkan support, enabling AMD/Intel GPU acceleration for LLMs. Currently source-build only, expanding beyond CUDA.

Sora’s rapid success faces backlash as deepfake videos of public figures raise serious legal and ethical concerns.

DeepSeek’s R1 cost ~$294K to train with 512 H800 GPUs and 80h. Inference costs are ~$0.55 input / $2.19 output per million tokens.

DeepSeek-V3 majorly upgrades from V2 with a massive 236B parameter count, a vastly larger 128K context window, and superior performance, all offered for free.

Alibaba is developing a new AI chip to reduce China's reliance on foreign technology like NVIDIA's, amid ongoing export controls and supply chain challenges.

Microsoft's open-source VibeVoice-1.5B AI can generate synthetic speech, threatening to disrupt the audiobook industry by reducing reliance on human narrators.

Gemini 2.5 Flash Image, known as nano-banana, delivers clear edits, creative blends, style shifts, and logical scene reasoning with simple prompts.

Nano Banana, a mysterious AI model spotted on LMArena, wows with realistic image edits and compositing. Photoshop isn’t obsolete yet it’s under pressure.

Ollama's new app for macOS and Windows offers a user-friendly interface to download and interact with local AI models, supporting files and images.

The Moe architecture enables running massive models like GPT-OSS-120B on consumer hardware by activating only a small fraction of its parameters at a time.

Qwen-Image-Edit adds precise bilingual text, semantic, and appearance editing via dual encoding accessible via Chat, API, Hugging Face, and open-source

NextStep-1 uses a 14B autoregressive Transformer with a flow-matching head, delivering detailed image generation and strong editing benchmarks.

A GitHub user leaked ChatGPT-5’s system prompt soon after release, raising concerns that disclosure may allow manipulation or security bypasses.

OpenAI's GPT-5, with “PhD-level knowledge,” brings faster replies, coding skills, app integrations, and AI personalities to 700M users.

Genie 3 is Google DeepMind’s real‑time world model: AI generates interactive 3D environments from text prompts, sustaining minutes of consistent exploration.

Qwen-Image is a new 20B open-source model that integrates flawless bilingual text rendering with powerful image editing & generation capabilities.

China's Zhipu AI releases GLM-4.5 models, outperforming Claude 4 Sonnet in benchmarks, signaling China's open-source AI leadership.