Thinking Machines Lab releases Inkling, a 975B-parameter open-weights LLM: Thinking Machines Lab has released Inkling, an open-weights model at 975B parameters — a significant milestone for the open model ecosystem approaching the 1T-parameter threshold. The open vs. closed dynamics in foundation models continue to evolve. Thinking Machines Lab · HN discussion
OpenAI loses EU trademark dispute: OpenAI lost a trademark case at the EU court, weakening its brand exclusivity in the European market. This signals growing legal friction for AI companies around IP protection. dpa international · HN discussion
BIS report: AI boom financing is shifting from equity to debt: The Bank for International Settlements published an analysis showing AI investment is increasingly debt-financed rather than equity-financed — a structural shift that raises questions about sustainability if IPO markets don’t deliver. BIS · HN discussion
Data centers have added $23B to US public electricity bills: Fortune reports that the rapid expansion of AI data centers has driven up electricity costs for the American public by an estimated $23 billion. The externalized costs of AI infrastructure are becoming visible. Fortune
xAI open-sources ‘Grok Build’ coding agent harness: xAI released Grok Build on GitHub — a fullscreen TUI-based coding agent harness with mouse interaction and extensibility. The AI coding tools race is expanding into open-source territory. GitHub/xai-org · HN discussion
Stripe and Advent make joint offer to acquire PayPal ($53B+): Stripe, together with Advent, has proposed to acquire PayPal for over $53 billion — potentially one of the largest fintech M&A deals ever. This could reshape the payments infrastructure landscape for startups. Reuters · HN discussion
Suno AI hack exposes training data scraping from YouTube, Deezer, Genius: A hacker breached Suno and revealed that the AI music generator scraped millions of tracks from YouTube Music, Deezer, Genius, Pond5, and podcast RSS feeds. Training data copyright issues resurface with concrete evidence. 404 Media · HN discussion
Microsoft and 3M announce strategic AI data center partnership: 3M and Microsoft are partnering on AI data center infrastructure — spanning cooling, materials science, and enterprise transformation. Physical infrastructure investment continues to accelerate alongside model development. Microsoft Source
Agnost AI (YC S26) launches to extract user feedback from agent conversations: YC S26 startup Agnost AI launched a tool that automatically extracts user feedback from AI agent conversations — a user research layer for the agent era. Agnost AI · HN discussion
Former Google DeepMind researcher publishes “Why I Left Google DeepMind”: A former DeepMind researcher publicly detailed their departure, citing Google’s support for the immigration enforcement supply chain. Internal ethical tensions at major AI labs remain a recurring signal. turntrout.com · HN discussion
“When A.I. Is a Member of the Family” — The New Yorker: The New Yorker explores a near future where AI assistants, companion robots, and AI caregiving become normalized. The piece suggests AI’s social acceptance may be shaped more by emotional trust than raw capability. The New Yorker
Linus Torvalds to AI critics: “fork off”: Linux creator Linus Torvalds told AI skeptics in the kernel community to fork the project if they object to AI involvement. The open-source world’s acceptance of AI-assisted development is becoming non-negotiable at the highest levels. The Register · HN discussion
Using DSLs to make LLMs reliable — Martin Fowler: Martin Fowler published a practical guide on combining Domain-Specific Languages with LLMs to produce more reliable, constrained outputs. A timely pattern for teams building enterprise AI features where hallucination control is critical. Martin Fowler · HN discussion
Open-source memory for coding agents — ‘deja-vu’: An open-source tool that provides persistent memory for coding agents, synced over SSH, gained traction on HN. For teams integrating agent capabilities into products, this is a reference approach for solving agent context retention. GitHub/vshulcz · HN discussion
AI training data copyright risk intensifies: The Suno hack provides concrete evidence of widespread scraping practices. Founders building AI products should watch the legal landscape closely — the absence of social consensus on “fair training data” is a material risk. 404 Media
AI investment shifting from equity to debt — sustainable?: The BIS report highlights that AI funding is increasingly debt-based. If IPO markets don’t deliver exits, AI startups could face a rapid funding squeeze. BIS
Meta sued for allegedly using AI to make layoff decisions: Former employees claim Meta’s layoff decisions were made by AI algorithms, not human managers. The legal and ethical boundaries of AI-driven workforce decisions are being tested. Ars Technica · HN discussion