AI News
AI community posts, news and release updates
Latest updates
- Launch
Show HN: RL Bandits Select Context for Coding Agent – Grok 4.5 Outperforms Fable 5
Published Jul 28, 2026This project uses RL bandits to dynamically select context for coding agents, with experiments showing Grok 4.5 outperforms Fable 5 in fixing code errors.
- coding agent
- RL
- Fable 5
- Grok 4.5
- bandits
- context selection
- Discussion
Oh god, is this how it starts? Already planning to convert TypeScript to JavaScript with Claude. Can I still make fun of vibe coders?
Published Jul 28, 2026A developer expresses excitement about Claude's new capabilities, planning to convert TypeScript to JavaScript, and jokes about becoming a 'vibe coder'. Sparks discussion on AI-assisted programming.
- Claude
- vibe coding
- TypeScript
- JavaScript
Lilian Weng, a prominent AI researcher formerly at OpenAI, has resigned from Thinking Machines Lab, potentially impacting the lab's future direction.
- Thinking Machines Lab
- Lilian Weng
- resignation
A tweet claiming that the founder of DeepSeek confessed that China's AI lags significantly behind the US.
- DeepSeek
- China
- America
- AI gap
- Idea
Second-generation bookseller investigates: bulk book purchases by AI companies might be algorithmic arbitrage, not training data
Published Jul 28, 2026A second-generation bookseller provides a detailed analysis of recent rumors about AI companies bulk-purchasing rare books. He finds that the orders are mostly for obscure but non-precious books, shipped to Amazon FBA prep services, suggesting algorithmic arbitrage rather than training data. He warns about the quieter crisis of disappearing bibliographic metadata and describes his project to build better records.
- Anthropic
- AI training
- books
- metadata
- bibliographic data
Microsoft fine-tuned Alibaba's Qwen 27B model to perform computer tasks like clicking and navigation, advancing AI agents for desktop automation.
- Microsoft
- Qwen
- fine-tune
- Computer Use
Modal offers Kimi K3 inference API with $30/month free compute credits, covering ~10M input or 2M output K3 tokens. One-click OpenCode setup, compatible with OpenAI tools. Also covers other models and raw GPU. No surprise bills, open to everyone.
- free API
- free compute
- Kimi K3
- Modal
- $30 credits
Reports reveal that Gemini is on the list of tools banned for Google engineers, ironically a Google product itself, raising questions about internal AI tool safety and reliability.
- Gemini
- banned tools
- engineers
A user reports that reverting to the Fable model gave worse performance than Opus 4.6-8, caused major issues in Hermes Agent, and they will not use it again.
- Claude
- Anthropic
- Hermes Agent
- Fable
Nvidia has formed the Open Secure AI Alliance, but major AI players OpenAI, Google, and Anthropic are notably absent, raising questions about their commitment to collaborative AI safety.
- Anthropic
- AI safety
- OpenAI
- Nvidia
- Open Secure AI Alliance