Good morning, welcome to this week’s Lore Brief, your 3-minute brief of the most important moves in AI and tech.
This issue is brought to you by Factory, the fastest way to ship software with autonomous engineering agents.
Moonshot AI unveils Kimi K3, its 2.8 trillion parameter frontier model → The model has 1 million token context and native multimodal support. New architectural innovations deliver major efficiency gains for agentic coding, with open weights scheduled for July 27. Read more here →
Codex moves past 9 million users with repeated limit resets → Strong growth has pushed Codex and ChatGPT Work past 9 million users in just one week after the release of GPT-5.6. The team issued multiple weekly limit resets this week to handle surging demand. Read more here →
Thinking Machines releases its first model → The startup introduced Inkling, which reasons efficiently across text, image, and audio with full weights now available. It offers tunable cost-performance tradeoffs and ranks among the strongest open models on audio benchmarks. Read more here →
Micro-drones for eradicating mosquitoes:
GPT-5.6 makes /goal command obsolete:
GPT-5.6 reaches IQ of 136:
xAI open-sources Grok Build, its coding agent and terminal interface → xAI has released the complete source code for Grok Build, its agentic coding tool and TUI, on GitHub. The release includes the full agent loop, code editing and execution tools, terminal UI components, and an extensible system for skills, plugins, and subagents, all designed for fully local-first operation with user-controlled inference. Read more here →
Anthropic's Boris Cherny outlines the 4 steps of AI team adoption → Many companies see one engineer 10x their output with Claude while the rest of the team lags behind. The four stages typically involve breaking bottlenecks, adding guardrails like automated reviews and multi-agent tools, measuring real ROI beyond raw usage, and eventually automating maintenance so teams can focus on new building instead. Read more here →
Demis Hassabis urges new safety framework → The DeepMind CEO proposes creating a new US-led Frontier AI Standards Body to rigorously test advanced models on risks like cybersecurity and biology, encourage responsible practices among frontier labs, and help establish international standards before capabilities advance further. Read more here →
X tweaks algorithm to boost visibility for mutual follows → The platform is rolling out a small change that increases visibility of posts from people you follow back in replies. Many users are praising the update for making X feel more like the former Twitter again, with friendlier conversations and easier interest-based clustering. Read more here →
SunoAI suffers breach exposing training data → A hacker breached AI music generator Suno and publicly released details of its training dataset, which included 113,879 hours of YouTube Music, 62,117 hours from Pond5, 12,287 hours from Deezer, plus Genius lyrics and plans for a million hours of podcasts. Read more here →
