Daily Brief
2026-08-30 · Morning · 8 items
-
Google Search tests defaulting to AI mode, pushing blue links further down
Some queries now load longer AI answers by default, pushing traditional results even lower on the page.IT之家（RSS）
-
Honor Magic8 AI Color Tracking upgraded with a palette-style control
Cloud updates now let users independently adjust tone and color levels, all triggered by a single voice command.IT之家（RSS）
-
Tencent releases and open-sources Tencent Hy4 Preview
With 770B total and 49B active parameters, it offers 1M+ token context for coding, office work, and research.Hacker News 热门（buzzing.cc 中文翻译）
-
OpenAI Resets Codex and ChatGPT Work Usage Limits, Patches Multiple Quota Bugs
Some bugs silently consumed 15%–70% of users' weekly quota; after fixes, the same allowance now lasts 10%–50% longer.X：Testing Catalog (@testingcatalog)
-
OpenAI Resets Codex Quotas and Fixes Multiple Usage-Draining Bugs
Worst cases burned through 70% of the weekly allowance; after fixes, the same quota now lasts 10%–50% longer.X：Kim (@kimmonismus)
-
Codex and ChatGPT Work Quotas Reset, Usable Time Up 10%–50%
Some users were losing up to 70% of their weekly allowance to silent bugs; the same quota now lasts 10%–50% longer.X：Tibo (@thsottiaux)
-
Cursor Responds to OpenAI Cutoff: 5% Share Is Inflated
OpenAI models drive only about 5% of Cursor traffic; weaker models burn more tokens, skewing the figure.X：Tibo (@thsottiaux)
-
Anthropic has Claude train models itself to curb alignment failuresRECAP
The method still works on models 4.7x larger than the training target without hurting general capabilities.Anthropic：Research（发表成果 · 网页）