Halo NeuroAI has released Sopro V2 Turbo, a 120M-parameter open-source voice cloning model that natively supports European Portuguese and enables real-time streaming on laptop CPUs.
OpenAI Codex adds a 'persistent' option to the reasoning-effort protocol and supports it in the TUI and SDK.
According to Stripe data, new startups in the AI era are dispersing to cities with populations under 1 million, yet frontier tech sectors continue to maintain the agglomeration effects of major metropolitan areas.
The LEAP panel forecasts a surge in data center investment and a slowdown in semiconductor growth, while predicting continued revenue growth for OpenAI and Anthropic.
OpenAI's GPT-5.6 discount policy increased token usage for the Terra and Luna models by 5.6x and 13.8x respectively, eroding competitors' market share.
The SGLang Diffusion team accelerated the MiniMax-H3 video generation model by up to 6.24x on eight NVIDIA H200 GPUs using Cache-DiT and SubBlock sparse attention.
Hugging Face has released Monsoon, a speech recognition evaluation dataset for Hindi and Indian English, to support Global South languages.
Vercel CLI has expanded to allow dashboard functions such as DNS records, domain renewals, and project state control to be performed from the terminal.
Claude resolved 26–96% of safety gaps across 10 types of alignment failures, outperforming human researchers.
You can now instantly create and deploy eve agents directly from the Vercel dashboard.
Vercel AI Gateway supports the open-source MoE model Hy4 Preview with 770B parameters.
Apple has released Agent Seer, a tool that automatically generates AI agent evaluation scenarios without manual effort, using only MCP specifications.
LLM belief updates diverge from Bayesian principles, and non-Bayesian heuristics actually demonstrated better performance.
It has been confirmed that OpenAI requires a minimum annual contract of $50,000 to use the Zero Data Retention (ZDR) feature.
A new llama.cpp branch featuring Metal and n-gram SSD offload optimizations has been released, improving Qwen-3.8-Flash-Next inference performance on Apple Silicon.
Free educational materials released for building RAG, agents, and evaluation systems using raw APIs without frameworks like LangChain, based on Colab.
py-evoFE v0.3.0, an open-source Python library that automatically generates optimal features for tabular data using genetic algorithms, has been released.
Ollama's KV cache quantization settings are applied globally, which can cause performance degradation for specific models.
UC Berkeley is launching a two-semester master's program specializing in AI and machine learning.
Experiential, an open-source AI gateway that optimizes model routing by analyzing user traffic, has been released.
Scale AI has released HarnessOpt-Bench, a benchmark for measuring the self-improvement capabilities of AI agents.
Anthropic is expanding its AI for Science program by offering 10,000 scientists worldwide free or $15/month Claude subscriptions.
AI Briefing