AI Briefing
TLDR AI news
The latest AI and developer news from TLDR AI, with short summaries.
Visit website →Recent Trends in LLM Architecture: KV Sharing, mHC, Compressed Attention
Open-weight LLMs including Gemma 4 and DeepSeek V4 are lowering long-context costs through KV sharing and compressed attention.
OpenAI Quietly Acquires Voice-Cloning Startup Weights.gg, Then Disbands the Team
OpenAI acquired Weights.gg and shut down its unauthorized celebrity voice catalog.
Portability is a myth: why the best AI stack can't be hardware-agnostic
The article argues that portability in AI kernels is a myth, and top performance comes from hardware-specific DSLs.
ChatGPT's New Personal Finance Feature
OpenAI has piloted a ChatGPT personal finance feature for US Pro users.
Runway, which started out supporting film production, now competes with Google in AI
Runway is challenging Google by expanding from video generation into world models.
Microsoft Cancels Claude Code License, Shifts Developers to GitHub Copilot CLI — Financial Motives a Key Factor
Microsoft will end use of Claude Code by the end of June and switch to its own GitHub Copilot CLI.
Toto 2.0: Time Series Forecasting Enters the Scaling Era
Datadog released Toto 2.0, ranging from 4m to 2.5B parameters, demonstrating clear scaling effects.
OpenSquilla Unveils Open-Source AI Agent That Cuts Token Costs
OpenSquilla has unveiled an open-source AI agent runtime that reduces token costs.
Nvidia's Jensen Huang teams up with UK startup Ineffable Intelligence to open AI's next frontier
Nvidia is co-developing reinforcement learning AI with UK startup Ineffable Intelligence.
Igor Babuschkin Pursues Up to $1 Billion in Funding for River AI
xAI co-founder Igor Babuschkin is pursuing up to $1 billion in funding for River AI.
Elon Musk's SpaceXAI Continues to See Talent Exodus After Merger
More than 50 research and engineering staff have left SpaceXAI since the merger.
Microsoft quietly searches for OpenAI alternatives
Microsoft is scouting alternative startups to reduce its dependence on OpenAI.
Raindrop Workshop GitHub Repository
Raindrop Workshop provides a local debugger that helps with agent tracing and evaluation.
Codex makes automating and customizing coding tasks easier
Codex adds Hooks and scoped tokens.
How We Built Secure, Scalable Agent Sandbox Infrastructure
Browser Use put agents in sandboxes and relayed them through a control plane for security and scalability.
OpenAI Considers Legal Action Against Apple
OpenAI has considered legal action against Apple over the underwhelming ChatGPT integration.
Cloud Agent Development Environment
Cursor unveiled development environments and security controls for Cloud Agent.
Work with Codex from anywhere
Codex has come to the ChatGPT mobile app, letting you check and approve remote tasks from your phone.
Grok Build Early Beta Released
Grok Build has been released in early beta for SuperGrok Heavy subscribers.
2028: Two Scenarios for Global AI Leadership
Anthropic warned that if the US fails to maintain its compute advantage, it could cede AI leadership to China by 2028.