Send us a text

🌍 INAI • The Open AI Hub

The Intelligence Atlas → the world’s most comprehensive, open hub of AI knowledge. 2 Million+ tools, models, agents, tutorials & daily news—free for all, updated every day.

https://github.com/inai-sandy/inAI-wiki

AI News Daily — Dec 23, 2025 (2500 chars)

Top Highlights: Google launched Interactions API with Deep Research for stateful, background AI workflows. ZhipuAI's GLM-4.7 excels at coding benchmarks while Google expanded Gemini 3 with faster Flash and 1M-token context. OpenAI introduced a ChatGPT app marketplace with tone controls, evolving from chatbot to platform. Amazon/AWS unveiled "frontier agents," Nova Forge, and $15B data center investment. Governments tightened rules: India banned ChatGPT for officials, UK targets deepfake apps, Firefox added AI "kill switch".

New Tools: Anthropic/MATS released Bloom for behavioral evals; Google's Interactions API enables long-running workflows; Qwen open-sourced Image Layered for Photoshop-style edits; Paperedge offers free collaborative research management; Maxio's MCP Governance Layer secures finance system access; Teikametrics' Marketplace Strategist optimizes cross-platform ads.

LLM Updates: GLM-4.7 achieves near-top SWE-bench scores; Gemini 3 Flash accelerates multimodal tasks with ~90% accuracy at 1M tokens; Baidu's ERNIE-5.0 tops Chinese leaderboards; MiniMax M2.1 targets agentic coding; Claude Opus 4.5 balances speed and reasoning; ChatGPT adds marketplace and personalization; vLLM upgrades reduce infra costs.

Research: LLaDA 2.0 scales diffusion models to 100B params; Medmarks provides largest medical LLM benchmark; MoReBench reveals moral reasoning gaps; multi-agent study finds coordination overhead; radiology AI detects liver disease from X-rays (>80% accuracy)and cuts kidney analysis time 30%+.

Industry: OpenAI explores $100B raise with SoftBank interest; Anthropic partners with U.S. DOE on Genesis Mission; M&A: Zendesk buys Unleash, HCLSoftware acquires Wobby, Anysphere acquires Graphite.

Showcases: GLM-4.7 generates Space Invaders on Apple Silicon; GPT-5.2-Codex builds 3D simulator; robotics demos show fast iteration; 13-year-old builds FallGuard fall detector.

Discussions: Hassabis vs LeCun debate on AGI; reasoning emergence tied to training conditions; agents' "body" is the browser; verification trumps training time.

Support the show

🌍 INAI • The Open AI Hub

The Intelligence Atlas → the world’s most comprehensive, open hub of AI knowledge. 2 Million+ tools, models, agents, tutorials & daily news—free for all, updated every day.

https://github.com/inai-sandy/inAI-wiki