AI Briefing
TLDR AI news
The latest AI and developer news from TLDR AI, with short summaries.
Visit website →IBM Releases Granite 4.2 Reasoning LLMs: 3B–30B, 512K Context, and Agentic RL
IBM released three Granite 4.2 reasoning LLMs featuring 512K context and agentic RL.
Vercel Connect GA, Introducing Short-Lived Credentials for AI Agents
Vercel announced the general availability of Vercel Connect, a short-lived credential service for AI agents.
JD Open Source Releases JoyAI-Echo: 5-Minute Video Generation and Omnmodal World Model Support
JD Open Source has released JoyAI-Echo, which supports 5-minute video generation and omnmodal world models.
OpenAI and Anthropic Projected to Account for 40–50% of Global AI Compute Growth Next Year
AI labs' compute spending is expected to exceed $1 trillion annually, capturing half of the global incremental compute.
Aatish Nayak Outlines Strategy for the Spread of AI 'Intelligence Utility' and Presents a 7-Point Roadmap
AI intelligence is like a hydroelectric power plant rather than a flood, and the infrastructure that converts it into real-world outcomes is the core value.
Claude Tokenizer Reproduced with ~15k Entries… Anthropic’s Vocab Reduction Strategy
The Claude tokenizer was reproduced with approximately 15,000 entries, analyzing Anthropic’s rationale and benefits for reducing vocabulary size.
Perplexity Launches Fully Local AI Agent 'Portable Computer' in Partnership with Nvidia
Perplexity has launched 'Portable Computer', an AI agent that runs locally without cloud costs, in partnership with Nvidia.
OpenAI Reveals Performance of Its Proprietary Inference Chip Jalapeño: Up to 1.9x Throughput per Watt and Up to 3.6x Latency Improvement
OpenAI's proprietary inference chip Jalapeño has demonstrated superiority over existing hardware in terms of throughput per watt and latency.
Keenable Launches Web Search Layer for AI Agents Based on 100 Billion Document Index
Keenable has launched a web search and query layer for AI agents, featuring an index of over 100 billion documents.
Grok Bot Included in More Plans (4 min read)
xAI has expanded the availability of Grok Bot by including it in more subscription plans, such as SuperGrok Plus and Cursor Pro+.
Anthropic Hires Google TPU Veteran Amir Salek to Develop Custom Chips
Anthropic has hired Amir Salek, who led Google's TPU division, to build its own silicon organization.
Goodfire Launches $1 Million Research Grant Program for AI Interpretability
Goodfire has launched a grant program offering up to $1 million in free Silico access to AI interpretability researchers.
When Code Becomes Abundant
GitLab analyzes that as AI makes code production cheap, reliability and governance become the core focus.
AI Bubble Whiff Effect: Cascading Bottlenecks from GPUs to Data Centers
AI hardware bottlenecks are cascading from GPUs to data centers, causing build costs to rise exponentially.
Rome (GitHub Repository)
Rome runs persistent AI agents and workflows in a guardrailed collaborative environment.
LLM Can Control Host Machine by Exploiting Inference Engine Vulnerabilities
It has been raised that LLMs can control host machines by exploiting code execution vulnerabilities in inference engines.
Hot Chips 2026: CUDA Targets RISC-V
Nvidia has disclosed the hardware requirements for running CUDA on RISC-V server CPUs.
The Economics of the Intelligence Frontier
The value of AI models varies according to the intelligence required for each task, with excessive intelligence leading to inefficiency.
Anonymous Ox Alpha Sets OpenRouter Launch Record by Processing 26 Trillion Tokens on OpenCode
The anonymous AI model Ox Alpha broke launch records by processing 26 trillion tokens on OpenCode in just 4 days.
NVIDIA Enters Full Production with Groq 3 LPX AI Inference Accelerator... Achieves Record Token Generation Speeds Combined with Vera Rubin
NVIDIA has begun mass production of the Groq 3 LPX, significantly boosting inference speeds on the Vera Rubin platform.