Daily AI Digest
2026-08-20
Source:橘鸦 AI 早报 · 22 items
Codex team says multilayer safeguards are now live to reduce the risk of accidentally deleting user files
开发生态
Read original →
OpenAI says a technical issue invalidated Daybreak Blue access for some users, requiring re-verification
开发生态
Read original →
OpenCode Go launches Muse Spark 1.2 and offers eight times the usage allowance for Hy3
开发生态
Read original →
ZCode adds off-peak quota resets to GLM Coding Plan, with users reporting a reset card in the console
开发生态
Read original →
Google AI Studio Build adds two-way synchronization with GitHub
开发生态
Read original →
Replit launches Free Mode powered by GPT-5.6 Luna
开发生态
Read original →
ClaudeDevs announces three updates to Claude Managed Agents
开发生态
Read original →
谷歌 launches student offer: Google AI Pro or Plus free for one year
产品应用
Read original →
Google introduces multiple AI learning tools for Search
产品应用
Read original →
MiniMax launches the MiniMax Design creative tool
产品应用
Read original →
Meta launches a beta version of the Meta AI app for Mac
产品应用
Read original →
OpenAI to expand ChatGPT Ads to 31 European markets next week
产品应用
Read original →
小红书 dots studio announces limited-time free API access to the dots3-note preview model
模型发布
Read original →
Unsloth releases Qwen3.8-27B Dynamic v3.0 quantization
模型发布
Read original →
蚂蚁百灵 open-sources six Base Model checkpoints for Ling-3.0-tiny and Ling-3.0-flash
模型发布
Read original →
Liquid AI releases QAD Q4_0 checkpoints for four LFM2.5 models
模型发布
Read original →
Ornith releases the Ornith-1.5 model family as open source
模型发布
Read original →
OpenAI announces continued availability of Zero Data Retention and plans for new security mechanisms
行业动态
Read original →
Stripe announces agreement to acquire OpenRouter
行业动态
Read original →
Anthropic's second-quarter revenue reportedly surpasses OpenAI for the first time
行业动态
Read original →
字节 Seed reportedly establishes four top-level departments for foundation models
行业动态
Read original →
唐杰: Parameters are not the only lever for Scaling; GLM-5.3 is a post-training controlled experiment
技术与洞察
Read original →