AI Briefing
TLDR AI news
The latest AI and developer news from TLDR AI, with short summaries.
Visit website →Introducing Kiro Crew
Amazon has open-sourced Kiro Crew, an AI agent system that autonomously coordinates multiple development tasks.
DiffusionGemma Technical Report
DiffusionGemma accelerates text generation by iteratively refining 256 tokens in parallel.
Anatomy of ChatGPT Work: An Agent for 1 Billion Users (18-min read)
ChatGPT Work is the blueprint for the general-purpose knowledge work agent that OpenAI is aiming for.
What Codex Actually Sends to the Model (10-Minute Read)
Codex sent a request of approximately 9,435 tokens to the model for an input of just 16 characters.
Unified API for AI Model Routing
Google Cloud API Gateway dynamically routes OpenAI-compatible requests to multiple LLMs.
Anthropic Reportedly Signs $10 Billion Cloud Deal with Volta (3 min read)
Anthropic is reportedly signing a $10 billion, six-year cloud contract with Volta.
Adding Human Sense to AI Models
Design Arena raised a KRW 7.9 billion seed round by using human taste data to evaluate AI models.
Google Developing Plugins for Gemini Enterprise
Google is developing Plugins and a Notifications area for Gemini Enterprise.
Google Workspace Plugin
Cursor can now directly read, modify, and manage Google Workspace apps.
White House to Review New Model Testing Framework with AI Companies on Tuesday
The White House will review a cybersecurity testing framework for advanced AI models with AI companies.
GPT-5.6 Sol uses twice the tokens of GPT-5.5
GPT-5.6 Sol's per-session token usage increased 2.25x compared to GPT-5.5.
Fast Gemma's Proven Inference Optimization Recipe (7 min read)
The VIDRAFT team boosted Gemma inference speed to 510.58 TPS on an A10G while maintaining quality.
From RLVR to RLSVR (GitHub Repository)
SpyRL implements open-ended LLM self-improvement through game-rule-based rewards.
MirrorCode (8-minute read)
MirrorCode evaluates whether AI can autonomously reimplement entire programs over long periods.
OpenAI's Unreleased Model Astra Solves 10 Open Math Problems
OpenAI's unreleased model Astra has produced new results on 10 open math problems.
One Agent for Every Environment: How Kiro Built Its Agent Harness
Kiro unified the separate agents of IDE, CLI, and web into a single harness.
How OpenAI Built GPT-Live
OpenAI built the architecture for GPT-Live in six months to enable real-time voice conversations.
Mind Lab Tests Continual Learning with Macaron-V1
Mind Lab released the LoRA-based Macaron-V1, demonstrating the potential of continual learning.
Anthropic Pays the Highest Salaries in the AI Industry. The CEO Now Realizes People Might Choose It for the Money
As AI researchers switch labs in pursuit of top salaries, Anthropic's mission-driven hiring strategy is being put to the test.
Former OpenAI Executive Fidji Simo Battles POTS, Develops Treatment Using AI and 3,500 Blood Samples
Fidji Simo founded ChronicleBio to find a treatment for POTS using AI and blood data.