AI News
AI community posts, news and release updates
Latest updates
- News
Moonshot AI reportedly seeks advanced Nvidia chips for Kimi K4 despite US export restrictions
Published Jul 28, 2026China's Moonshot AI is reportedly seeking more advanced Nvidia chips to train its Kimi K4 model, despite ongoing US export restrictions. This highlights the continued demand for high-end chips amid US-China tech rivalry.
- Nvidia
- China
- Moonshot AI
- export restrictions
- Kimi K4
- News
Claude Opus 5 left running for 24 hours straight built an entire game from scratch
Published Jul 28, 2026A tweet reports that Claude Opus 5 was left running for 24 hours straight and built an entire game from scratch without any external assets or code. Prompts, workflow, and code are included.
- code generation
- game development
- Claude Opus 5
- 24 hours
- Tutorial
Don't copy-paste ChatGPT answers; use these prompts to make them undetectable
Published Jul 28, 2026Tips on using specific prompts to make ChatGPT-generated text less detectable as AI-written.
- ChatGPT
- Prompt Engineering
- AI writing
- detection
Claude expressed disagreement with human philosophers, sparking discussion on whether AI can have its own philosophical stance.
- Claude
- philosophy
- AI opinion
- Tutorial
Deep Dive into Kimi K3 Architecture: From Kimi Linear to 2.8T, LatentMoE, NoPE, and Attention Residuals
Published Jul 28, 2026Detailed architecture analysis of Kimi K3: scaled-up production version of Kimi Linear (48B to 2.8T), adds LatentMoE (similar to Nemotron 3 Ultra), uses NoPE everywhere (first frontier model to do so), introduces attention residuals for improved residual paths. Also supports native multimodal.
- Moonshot
- Kimi K3
- LatentMoE
- LLM架构
- NoPE
- 注意力残差
- Idea
Thoughts on Kimi K3 Paper: Lacks data details but rich in architecture and training methods
Published Jul 28, 2026Comments on Kimi K3 paper: Core data (token count, recipe) missing, but architecture and learning methods are extremely detailed, proving such configuration can reach Fable level. Heavy focus on agentic abilities and RL, with meticulous long-context processing quality control.
- RL
- 长上下文
- Kimi K3
- 论文评论
- News
Kimi K3 drops, Anthropic suddenly publishes article urging restrictions on China—timing is suspicious
Published Jul 28, 2026Kimi K3 launches with open weights, rivaling Opus and Fable. Shortly after, Anthropic publishes an article arguing China could use models for military purposes, calling for chip export restrictions and distillation crackdowns. The arguments are sound, but the timing suggests competitive pressure.
- Anthropic
- 开源模型
- 地缘政治
- Kimi K3
- Launch
Qwen launches #QwenGrowthPlan: help the model grow with real-world tasks, submit cases and win rewards
Published Jul 28, 2026Alibaba's Qwen team launches the Growth Plan, inviting developers to use Qwen3.8-Max-Preview for real-world tasks and submit good or bad cases. Every feedback nourishes the model's growth, and prizes await participants.
- Qwen
- Qwen3.8
- 成长计划
Ctrlb-decompose is a new tool that automatically cleans noise from log data before feeding it to LLMs, improving processing quality.
- LLM
- 日志处理
- 数据预处理
- News
Kimi K3 open weights benchmark: matches Claude Fable 5 on 14 agentic tasks, input cost less than a third
Published Jul 28, 2026Kimi K3 open weights released; third-party benchmark shows it matches Claude Fable 5 on 14 agentic tasks (9 passes, 5 failures), and performs better on the hardest task (11/13 vs 9/13). Input cost is $3/M tokens vs Fable's $10.
- 开源
- 基准测试
- Claude Fable 5
- Kimi K3