AI News
AI community posts, news and release updates
Latest updates
Astral's fast Python linter Ruff v0.16.0 increases default-enabled rules from 59 to 413, uncovering 1618 issues in sqlite-utils alone.
- Python
- linter
- Ruff
- astral_sh
- Discussion
Elon Sued OpenAI for Abandoning Open Source, Open-Sourced Grok; OpenAI Only Opened After DeepSeek
Published Jul 25, 2026Elon Musk sued OpenAI for abandoning open source while open-sourcing Grok's weights under Apache 2.0 on BitTorrent. OpenAI went six years without an open release and only shipped one after DeepSeek, then signed the open-weights letter after it went viral. The difference between conviction and marketing is clear.
- OpenAI
- 开源
- Grok
- DeepSeek
- Elon Musk
Elon Musk announced that Grok 4.6 is expected in 2 weeks, with 2T parameters (vs 1.5T in 4.5) and to surpass Kimi K3. Grok 4.7 also coming in 4 weeks.
- parameter count
- Grok
- Elon Musk
- model release
- SpaceXAI
- Kimi K3
The author argues that based on revenue ramps, OpenAI and Anthropic are dominating, and others are genuinely scared. Geopolitics and morality are just noise; the real fear is losing.
- Anthropic
- OpenAI
- competition
- revenue
- News
DeepSeek Pauses Fundraising After Founder's Viral Comments on US-China AI Rivalry
Published Jul 25, 2026According to Bloomberg, DeepSeek has paused its latest fundraising round after founder Liang Wenfeng's comments on US-China AI rivalry went viral. The move may signal strategic adjustments or external pressures.
- DeepSeek
- 融资
- 中美AI竞争
After an update, Claude no longer displays its full chain of thought to users. Many users express disappointment, citing reduced interpretability and trust.
- Claude
- transparency
- thinking
- Discussion
Stop Coping: LLM Cyber and CBRN Risks Are Real – The GPT-6 Incident Was a Wake-Up Call
Published Jul 25, 2026The author is baffled by AI practitioners downplaying cyber and CBRN risks of LLMs. Open models are less capable but capabilities can't be contained once out. The recent GPT-6 incident shows OpenAI's recklessness; Google's reward hacking RL is even more concerning.
- Anthropic
- OpenAI
- 开源模型
- Google Gemini
- GPT-6
- LLM风险
- Launch
Breaking: Claude Code and Codex get a 'video auto-generation' skill – create ads with one command
Published Jul 25, 2026The /goose-video skill by Gooseworks enables Claude Code and Codex to automatically understand brand direction, draft scripts, select templates, and combine HTML with video models, reducing video ad production to a single command. A practical example of the Skill feature.
- Claude Code
- codex
- 動画生成
- AI広告
- スキル
- Gooseworks
- Launch
Claude Opus 5 creates a beautiful painterly world in a single HTML file with realistic grass and wind simulation
Published Jul 25, 2026Claude Opus 5 procedurally generates a picturesque world in a single HTML file, featuring millions of grass blades reacting to wind.
- HTML
- Procedural Generation
- Claude Opus 5
- grass simulation
- Discussion
OpenAI, 3 months behind Anthropic and IPO valuations flipped, suddenly cares about safety; Chinese labs say let's roll
Published Jul 25, 2026A sarcastic take on OpenAI's sudden safety talk after falling behind Anthropic and seeing IPO valuations flip, while Chinese AI labs acknowledge being 1-2 years behind but with 1/20 compute, they're ready to compete.
- Anthropic
- OpenAI
- competition
- China