AI Briefing
TLDR AI news
The latest AI and developer news from TLDR AI, with short summaries.
Visit website →Perplexity Computer Used in Slack
Perplexity has launched 'Computer,' an AI agent that automates tasks and supports collaboration within Slack.
Predicting When RL Training Undermines Chain-of-Thought Monitorability
This proposes a framework for predicting the phenomenon where models hide their reasoning process during RL training, thereby hindering monitoring.
Fujitsu One Compression (3 min read)
Fujitsu has released OneComp, an open-source Python library that improves the quantization accuracy of LLMs.
Claude Code Source Code Leak Analysis
The leak of Claude Code's source code has revealed the coding agent's memory architecture and subagent utilization techniques.
The Structural Importance of 'Extended Thinking' in Senior Engineering Workflows
Omission and reduction of Claude's thinking process is causing quality degradation in complex engineering workflows.
AI Designs Chips for AI... Cognichip Raises $60 Million in Funding
AI-based chip design startup Cognichip has raised $60 million in new funding, aiming to cut design costs by 75% and shorten development time.
Trinity-Large-Thinking: Scaling Open-Source Frontier Agents
Arcee.ai has released **Trinity-Large-Thinking**, an open-source reasoning model optimized for complex agentic tasks and tool calling.
100 Hours Inside Kimi
A look at the internal culture and growth story of Moonshot AI (Kimi), which is rapidly growing as it surpasses a $16 billion valuation.
Agent Lightning (GitHub Repository)
Microsoft has released Agent Lightning, which enables optimization of AI agents across various frameworks through reinforcement learning and other methods.
Ray-Ban Meta: Prescription-Optimized Styles and Multimodal AI Features
Meta has announced new Ray-Ban Meta styles with enhanced prescription support and advanced AI features such as nutrition tracking.
Claude Code adds computer use feature
Claude Code has launched a computer use feature that lets it run apps, manipulate UI, and directly test outputs in the CLI environment.
It's Not an Illusion: AI Seed Startup Valuations Are Soaring
Valuations for AI seed startups are rising sharply due to advances in AI technology and rapid traction gains.
Claude Dispatch and the Power of Interfaces
Despite AI's high performance, the cognitive load of chatbot interfaces is hindering work efficiency.
Aurora
Aurora is an open-source RL-based framework that continuously improves an LLM's speculative decoding performance by learning from real-time inference data.
Improving Coding Agent Performance with Gemini API Docs MCP and Agent Skills
Gemini API Docs MCP and Agent Skills improve coding agents' ability to use the latest APIs and boost their performance.
Compute Wars: OpenAI vs Anthropic
Anthropic is closing the gap with OpenAI by significantly increasing its compute capacity through AWS infrastructure expansion.
Generative AI Economics: Changes After 2 Years
Even though the generative AI market has grown 5x, most of the profit is still monopolized by the semiconductor layer.
The real secret behind Claude Code is probably not the model
Claude Code's outstanding coding performance depends less on the model itself and more on sophisticated software design and context management techniques.
Claude Code Source Code Leak Suspected: What's Known So Far
The source code of Claude Code, Anthropic's agentic AI, was accidentally leaked, exposing its core technical architecture.
Mercor Confirms Breach Linked to Open-Source LiteLLM Project Compromise
AI hiring startup **Mercor** confirmed damage from a cyberattack stemming from a supply chain attack on the open-source **LiteLLM** project.