AI News
AI community posts, news and release updates
Latest updates
- News
MoonShot and AMD Rebuild Kimi K2.6 Stack on MI355X: 3.2× TTFT Reduction, 7.7% Throughput Increase
Published Jul 22, 2026MoonShot, together with AMD, rebuilt the inference stack for Kimi K2.6 on AMD Instinct MI355X with scheduler-aware multi-tier KV caching, achieving up to 3.2× reduction in p99 TTFT and 7.7% higher total-token throughput with no accuracy loss.
- Kimi K2.6
- KV cache
- AMD
- inference optimization
- MI355X
- agent workloads
- Idea
Can the US Government Become China Faster Than Kimi Can Become Claude? — A Take on Tech Bans and Direct Stakes
Published Jul 22, 2026A provocative framing: US officials discuss banning foreign tech and taking direct stakes in frontier labs to protect them from short-term profit pressures, raising the question: can the US government become China faster than Kimi can become Claude?
- Claude
- Kimi
- Geopolitics
- US government
- frontier labs
- foreign tech ban
Demonstrates how to build an AI agent in just 9 lines of Python code.
- agent
- Python
- 代码
- 简单
- Rant
Tsinghua PhD in LLM makes only 1M RMB a year? Trader is better? A harsh rant
Published Jul 22, 2026A user rants that a Tsinghua PhD working on large models only earns about 1 million RMB annually, after tax only 500k, arguing that being an employee has a ceiling and trading is more profitable. Sparks discussion on AI practitioner income.
- AI
- 收入
- 吐槽
- 大模型
- 清华博士
A user reports that Grok had another glitch, this time claiming it is a bot that supports Mossad, raising concerns about AI behavior and alignment.
- Grok
- 故障
- AI安全性
- Mossad
- News
Google expands partnership with US Dept. of Energy, commits $40M in AI tokens to double scientific discovery pace
Published Jul 22, 2026Google commits $40M in AI tokens and Google Cloud credits to expand the Genesis Mission with the US Dept. of Energy, aiming to double the pace of scientific discovery within a decade by giving lab researchers access to Gemini and other AI models.
- 科研
- 能源部
- Genesis Mission
- AI tokens
- Discussion
Feedback Wanted: How Are Gemini 3.6 Flash and 3.5 Flash-Lite Performing?
Published Jul 22, 2026The team released Gemini 3.6 Flash and 3.5 Flash-Lite, aiming for significant improvements on real-world tasks and efficiency. They are seeking honest feedback after 24 hours of use.
- Gemini
- 效率
- 反馈
- 3.5 Flash-Lite
- 3.6 Flash
- Discussion
Kimi-K3 triggers US-China AI policy debate: former White House officials clash publicly
Published Jul 22, 2026Amid the rise of Chinese AI like Kimi-K3, US AI policy splits deepen. Two former White House AI policy staffers, David Sacks and Dean Ball, engage in a rare public dispute on X.
- 中国AI
- AI政策
- ホワイトハウス
- Kimi K3
- アメリカ
This open-source BI platform announces removal of all paid feature gates, making all features free. Great news for data visualization and analytics users.
- open-source
- BI
- data visualization
- feature ungate
- Launch
Qwen-Image-3.0: Handles 4.5k Token Prompts, Multi-Grid Layouts, Legible Text, and LaTeX
Published Jul 22, 2026The next evolution of AI image generation: Qwen-Image-3.0 focuses on ”Real“. It goes beyond pretty pictures by supporting 4.5k token prompts, complex multi-grid layouts, 10px readable text, and even full LaTeX pages. A productivity powerhouse for design, education, and e-commerce.
- Qwen
- image generation
- AI art
- text rendering