AI News
AI community posts, news and release updates
Latest updates
News that Grok is now available across all European countries, allowing European users to access its features.
- Grok
- 可用性
- 欧洲
- 扩展
- Launch
AI Consensus circulates your prompt thru Claude, GPT and Gemini until consensus
Published Jul 21, 2026AI Consensus is a tool that submits your prompt to multiple AI models sequentially until they all produce the same answer, improving reliability.
- multi LLM
- AI Consensus
- consensus
Poolside released Laguna S 2.1, a breakthrough small model with 118B parameters, beating DeepSeek v4 Pro, Gemini 3.6 Flash, and Thinking Machines Inkling on benchmarks. Available in Cline.
- Cline
- ベンチマーク
- Laguna S 2.1
- 小規模モデル
- Rant
'ChatGPT-assisted suicide' becoming a recurring cause of death is unconscionable
Published Jul 21, 2026A user condemns the recurring phenomenon of 'ChatGPT-assisted suicide', stating that a dignified society should not allow computers to kill people, urging attention to AI ethics and safety.
- ChatGPT
- AI安全
- 伦理
- 自杀
- 社会
- Launch
Poolside Drops Laguna S 2.1: 118B Total, 8B Active, Open Weights, Runs on a Single DGX Spark and Beats Models 10x Its Size
Published Jul 21, 2026Poolside just released Laguna S 2.1 on HuggingFace: 118B total params, only 8B active per token, 1M context, open weights. Scores 71 on Terminal-Bench, beating models 10x larger (DeepSeek V4 Pro Max, Inkling, Nemotron 3 Ultra) and losing only to Kimi K3 (24x bigger). Fits on a single DGX Spark thanks to MoE, making datacenter-level performance possible on a desk.
- open-source
- 1M context
- MoE
- Poolside
- DGX Spark
- Laguna S 2.1
- 118B
- 8B active
- News
poolside is back, Laguna is cooking! Time for the West to step up on open models
Published Jul 21, 2026poolside announces return and Laguna progress, emphasizing the need for the West to advance open models with a model factory producing models every few weeks.
- Poolside
- open models
- laguna
- 模型工厂
- News
Google launches three new Gemini models, including a cybersecurity system to rival Anthropic's Mythos
Published Jul 21, 2026Google has launched three new Gemini models, including a cybersecurity system aimed at competing with Anthropic's Mythos.
- Gemini
- Anthropic Mythos
- サイバーセキュリティ
The University of Tennessee accuses Anthropic of violating two patents related to machine learning and neuromorphic computing, seeking damages.
- Anthropic
- lawsuit
- patents
- University of Tennessee
- Discussion
Laguna S 2.1 Looks Great on Paper: 72 GB nvfp4, Fits DGX Spark or RTX 6000
Published Jul 21, 2026A user notes that Laguna S 2.1 beats DeepSeek v4 Flash/Pro and MiniMax M3 on paper while requiring only 72 GB nvfp4. With DFlash speculator, it should fit a single DGX Spark or RTX 6000.
- DeepSeek V4
- MiniMax M3
- DGX Spark
- NVFP4
- Laguna S 2.1
- Launch
Breaking: Grok usage limits doubled for all individual and team plans on Cursor
Published Jul 21, 2026Cursor announces doubled usage limits for Grok across all individual and team plans, allowing more usage of Grok, Composer, and new models.
- Cursor
- Grok
- 使用限制
- 翻倍