AI Briefing#foundation-models
The latest AI and developer news about #foundation-models, with the original source and a short summary.
The latest AI and developer news about #foundation-models, with the original source and a short summary.
The framework now supports third-party server models like Claude and Gemini via a unified Swift API, and introduces Dynamic Profiles for adaptive agent sessions.
NVIDIA releases GR00T 2, a foundation world-action model that predicts future video and robot actions, enabling zero-shot transfer from human demos and real-time control of Unitree G1 humanoids via the GR00T 2 Turbo variant.
Three Spotify Principal Product Managers explain how their roles differ across the research, model, and experience layers of the AI stack.
Anthropic's S-1 filing reveals $1.7B in Q3 2026 personnel costs, with stock-based compensation representing the largest share of the expense.
The method corrects posterior probabilities at test time without retraining, maintaining performance on balanced data while fixing recall collapse in imbalanced scenarios.
The largest 1B-parameter model achieves 44.71% average zero-shot QA accuracy, surpassing the 39.79% of the corresponding GPT baseline.
The new architecture internalizes memory via Gated DeltaNet-2 and Mixture-of-Storage, outperforming parametric baselines while preserving general capabilities.
The NeurIPS 2026 paper provides a library of synthetic time series generators and augmentations for training a DeltaProduct based linear RNN.
The open-source release includes 5M to 209M parameter checkpoints supporting zero-shot forecasting with up to 32,768 historical time steps.
The collaboration aims to train a single pretrained Large Physics Model on billions of simulated LHC collisions to streamline particle reconstruction and analysis.
The new Pi API allows developers to upload data and fine-tune π0 and π0.5 models on managed infrastructure, enabling hosted inference for diverse robotic applications.
Google DeepMind, Isomorphic Labs, and Meta are collectively investing $300 million to support the creation of multi-modal datasets for predictive models of life.
OpenAI has released an open-weight model that erases PII such as names and account numbers.
MiniMax has unveiled MMX-CLI, its official multimodal CLI for agents.
Forge trains, aligns, and evaluates frontier-level AI models using a company's internal knowledge.
Mistral AI joins the NVIDIA Nemotron Coalition to co-develop open frontier models.
AWS Elemental Inference automatically converts live video into mobile and short-form formats.
Amazon Science's most-viewed publications of 2025 were led by Nova, safety, and robotics.
A roundup of Amazon Science's 10 most-viewed blog posts of 2025, organized by topic.
Nova Forge lets enterprises directly train frontier models on their own data.
Upstage was selected for CB Insights' 2025 AI 100, proving Korea's AI competitiveness.