join.dev
AI news · 10 stories
Anthropic proposes an equity-backed public-benefit fund · Meta removes disputed AI-scraping language · AI-stock selloff puts earnings and spending under scrutiny
Anthropic announced on October 18 that it plans to commit 5% of its shares to a new public-benefit fund, subject to stockholder approval. The fund could support nonprofits and direct distributions, including responses to AI-driven labor disruption.
Read at Anthropic
Meta removed disputed commercial-terms language after concerns about AI training rights. It says crawling practices remain unchanged and website owners can still signal preferences through robots.txt.
Read at The Verge
Reuters reports that recent declines in AI-linked shares have sharpened attention on upcoming earnings from Meta and chip companies. Investors are looking for evidence that profits can justify the scale and financing costs of AI infrastructure spending.
Read at Reuters
GitHub’s published schedule retires Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini and Grok 4.5 from Copilot on October 19. Teams relying on these models need to move workflows to supported alternatives and check administrator model policies.
Read at GitHub
Anthropic’s October 16 workplace study reports a falling share of junior technical hires and rising output as employees adopt AI tools. The company identifies attention and judgment as emerging bottlenecks; its internal observations do not establish that every workplace will follow the same pattern.
Read at Anthropic
Schwab says it blocks outside AI agents attempting to move customer money and plans controlled MCP connections for balance checks and research. The approach separates information access from transaction authority as financial firms confront agent security and liability concerns.
Read at The Next WebNVIDIA and AI4Bharat announced an approximately 2.8-trillion-token synthetic corpus spanning 22 Indian languages, with release described as coming soon. An open generation recipe provides translation and quality-checking workflows that developers can adapt to expand Indic model training data.
Read at NVIDIA
Google’s October 16 update describes recently launched short video overviews, interactive learning overviews and expanded quiz formats in Gemini Notebook. The features turn uploaded coursework into structured study material and feedback, giving learners more ways to review their sources.
Read at Google
Hugging Face described normalization optimizations on October 15 that achieve up to roughly 50 times faster BERT tokenization than version 0.23 in its tests. The changes were still at pull-request stage, targeting preprocessing overhead in embedding, retrieval and classification pipelines.
Read at Hugging Face
GitHub’s October 14 HydraFusion update adds a Max option, a refreshed model pool and workflows that vary reasoning effort between drafting and review. Both options remain research previews across VS Code, the Copilot app and CLI, coordinating multiple models into one response.
Read at GitHub