Pika's new Founder Starter Kit plugs a full brand-to-video pipeline directly into Claude via MCP, replacing an entire marketing stack with four chat commands
xAI opens grok-build-0.1 to all developers via public API -- a purpose-built agentic coding model at $1/M tokens with always-on reasoning and a 256K context window
Android Studio's Agent Mode now ships with Firebase skills pre-installed, letting the AI agent set up Firestore, Auth, and security rules with zero extra configuration
OpenAI's Rosalind Biodefense gives vetted government labs and developers free access to GPT-Rosalind, its specialized biology reasoning model, to build pandemic and biosecurity tools — but dual-use risks loom large.
Alibaba's Qwen team unveils a single 5B model that controls 11 robot platforms across manipulation, navigation, and trajectory prediction with one set of weights.
A new theory proves that predicting your own latent representations requires exponentially less data than token-level objectives like masked language modeling or diffusion.
Google's encoder-free Gemma 4 12B drops vision and audio encoders, fitting multimodal reasoning into a 6.6GB quantized file that runs on a 16GB laptop.
Tencent's Hy-MT2-7B arrives as a GGUF via Unsloth, bringing purpose-built 33-language translation that beats DeepSeek and commercial APIs to local hardware
Sakana AI teams up with a private OSINT outfit run by two of Japan's best-known security analysts to scale satellite and open-source intelligence work.
The LocalAI team applies APEX, a MoE-aware quantization method, to Liquid AI's LFM2.5-8B-A1B, squeezing a 128K-context reasoning model down to 3.6–6.3 GB for fully local inference.
Google's Magenta RealTime 2 is a free, open-weights 2.4B model that generates live stereo music from MIDI, text, and audio with just 200ms of latency — on a MacBook.