Import AI 458 Asks What AI Miracles 2025 Will Bring
Jack Clark's latest newsletter frames the coming year as a test of how quickly AI breakthroughs reshape expectations.
AI & ML
Frontier models, research results, agents, and developer tooling.
Anthropic co-founder tells IMF the singularity has begun
Jack Clark used a major IMF lecture to argue that AI is now driving a singularity-like acceleration in science and technology.
MiniMax M3.1 Flash: A 428B MoE Model With Low-Cost API Access
The new agentic model combines a 1M-token context with sparse activation and aggressive pricing for broad developer access.
Google's ERA Aims to Turn Empirical Research Into Computational Discovery
Google Research's ERA tool, grounded in a Nature publication, is positioned as a step toward accelerating computational discovery.
NVIDIA and EBU Expand Sovereign AI Framework for European Media
The European Broadcasting Union is turning its sovereign AI principles into deployable services, while Adobe Frame.io uses NVIDIA GPUs to speed video search for producers like UFA.
Google DeepMind Launches Asia Pacific Accelerator for Environmental Risks
The new program signals DeepMind's push to apply AI to environmental challenges in the region.
OlmoEarth v1.1: More Efficient Earth Observation Models
Allen AI releases an updated family of Earth observation models designed for improved efficiency.
AI Clones Push Indie Devs to Hide Early Game Demos
Indie developers are delaying demos and guarding prototypes as AI tools make it easy to clone gameplay quickly.
Cisco and OpenAI Partner on Codex for Enterprise Engineering
The collaboration aims to scale AI-native development and automate defect remediation at Cisco.
Google’s zero-trust aggregation aims to keep analytics private
Google Research outlines a privacy-preserving approach to analytics that reduces reliance on any single trusted party.
OpenAI Launches Rosalind Biodefense for Government and Vetted Developers
The initiative broadens access to GPT-Rosalind, applying frontier AI to biodefense and public health preparedness.
Why AI Agent Terminology Matters: Harness vs. Scaffold
A Hugging Face post argues that precise terms like harness and scaffold are essential for building and discussing AI agents.
OpenAI Brings Frontier Models and Codex to AWS
Enterprises can now use OpenAI's frontier models and Codex through AWS, with familiar controls and procurement workflows.
OpenAI Lays Out Third-Party Eval and Governance Plans
OpenAI published two frameworks that tie external AI testing to broader safety and regulatory alignment for frontier models.
Google Research Touts 'New Era of Discovery' at I/O 2026
A single blog post from Google Research frames its I/O 2026 presence as the start of a new phase in AI-driven science, though it offers no concrete announcements.
Researcher claims neural networks are far more interpretable than assumed
A new blog post argues that the tangled inner workings of large language models may be largely legible, sparking both excitement and skepticism.
DeepMind's Co-Scientist Speeds Up Aging Research
Google DeepMind's AI tool helps biologists find novel rejuvenation factors and link scattered findings in aging research.
Endava uses OpenAI Codex to cut requirements analysis from weeks to hours
A case study shows how the consulting firm is using Codex to speed software delivery and shift toward an agentic workflow.