The AI-native transformation is reshaping corporate cost structures. Airbnb announced a significant increase in AI token spending and calls itself an AI-native company [0]; meanwhile, an OpenAI researcher's coding agent has reached a daily spend of $601 and roughly doubles every month [1], while Meta and Microsoft have begun cutting employees' use of Claude [12], showing a divergence in industry AI investment. On the security front, the MCP agent communication protocol has exposed structural flaws, with Google and four other institutions...
## 🔍 Core Insights
**The AI-native transformation** is reshaping corporate cost structures. **Airbnb** announced a significant increase in AI token spending and calls itself an AI-native company [0]; meanwhile, an **OpenAI** researcher's coding agent has reached a daily spend of **$601** and roughly doubles every month [1], while **Meta** and **Microsoft** have begun cutting employees' use of **Claude** [12], showing a divergence in industry AI investment. On the security front, the **MCP agent communication protocol** has exposed structural flaws, with five institutions including Google having confirmed the related vulnerabilities [15].
## 🚀 Key Updates
- **Airbnb to significantly increase AI token spending** [0]: The CEO says AI is the best thing that has ever happened to Airbnb, and the ROI is there.
- **OpenAI researcher's coding agent spends about $601 per day** [1]: An Epoch AI report says usage roughly doubles every month, annualizing to about $2 million.
- **Google Gemini will support making private calls to friends and family on users' behalf** [2][9]: The Call for Me feature is expected to expand from business scenarios to private calls.
- **Apple TV 4K (4th generation) will support Siri AI and Apple Intelligence** [3]: tvOS 27.2 code shows a new data authorization toggle.
- **MCP agent communication protocol exposes structural flaws** [15]: Five institutions including Google and JP Morgan Chase confirmed the related vulnerabilities.
- **Meta and Microsoft cut employee use of Claude** [12]: Microsoft lowered its internal Claude spending forecast by more than 33%.
- **Shanghai AI Lab's SkillGym paper** [14]: Training on verified skill execution trajectories improved Qwen3.5-35B-A3B by 19.10 percentage points on Terminal-Bench 2.1.
- **Hao AI Lab open-sources Minecraft agent Jev** [22]: It achieves a 70% win rate against human players, using prompt evolution rather than training.
## 🔗 Sources
[0] Airbnb to significantly increase AI token spending — https://aihot.news/items/t2qqwlk1v93kn10c99l5pg6f9
[1] Epoch AI report: OpenAI researcher's coding agent spends about $601 per day and roughly doubles every month — https://aihot.news/items/jyo5wkzg1w5kp2gxzx2sesbyv
[2] Google's Call for Me feature may be upgraded: Gemini will support making private calls to friends and family on users' behalf — https://aihot.news/items/rqed79ovuc47fhctqibtcyyfm
[3] tvOS 27.2 code shows Apple TV 4K (4th generation) will support Siri AI and Apple Intelligence — https://aihot.news/items/fi6ahpxthdbszqij928xydgj0
[9] Gemini Call for Me may expand to calls with friends and family, and can send a "running 15 minutes late" message on the user's behalf — https://aihot.news/items/mid9e4m377uvvg4frqyrav39s
[12] Meta and Microsoft cut employee use of Claude — https://aihot.news/items/c2nkq5vvxv1qmsjls4wd3g4je
[14] Shanghai AI Lab's SkillGym paper: Training on verified skill execution trajectories can improve Qwen3.5-35B-A3B by 19.10 percentage points on Terminal-Bench 2.1 — https://aihot.news/items/ysoc64sr9161851v9k
The AI-native transformation is reshaping corporate cost structures. Airbnb announced a significant increase in AI token spending and calls itself an AI-native company [0]; meanwhile, an OpenAI researcher's coding agent has reached a daily spend of $601 and roughly doubles every month [1], while Meta and Microsoft have begun cutting employees' use of Claude [12], showing a divergence in industry AI investment. On the security front, the MCP agent communication protocol has exposed structural flaws, with five institutions including Google having confirmed the related vulnerabilities [15].
🚀 Key Updates
- Airbnb to significantly increase AI token spending [0]: The CEO says AI is the best thing that has ever happened to Airbnb, and the ROI is there.
- OpenAI researcher's coding agent spends about $601 per day [1]: An Epoch AI report says usage roughly doubles every month, annualizing to about $2 million.
- Google Gemini will support making private calls to friends and family on users' behalf [2][9]: The Call for Me feature is expected to expand from business scenarios to private calls.
- Apple TV 4K (4th generation) will support Siri AI and Apple Intelligence [3]: tvOS 27.2 code shows a new data authorization toggle.
- MCP agent communication protocol exposes structural flaws [15]: Five institutions including Google and JP Morgan Chase confirmed the related vulnerabilities.
- Meta and Microsoft cut employee use of Claude [12]: Microsoft lowered its internal Claude spending forecast by more than 33%.
- Shanghai AI Lab's SkillGym paper [14]: Training on verified skill execution trajectories improved Qwen3.5-35B-A3B by 19.10 percentage points on Terminal-Bench 2.1.
- Hao AI Lab open-sources Minecraft agent Jev [22]: It achieves a 70% win rate against human players, using prompt evolution rather than training.
🔗 Sources
[0] Airbnb to significantly increase AI token spending — https://aihot.news/items/t2qqwlk1v93kn10c99l5pg6f9
[1] Epoch AI report: OpenAI researcher's coding agent spends about $601 per day and roughly doubles every month — https://aihot.news/items/jyo5wkzg1w5kp2gxzx2sesbyv
[2] Google's Call for Me feature may be upgraded: Gemini will support making private calls to friends and family on users' behalf — https://aihot.news/items/rqed79ovuc47fhctqibtcyyfm
[3] tvOS 27.2 code shows Apple TV 4K (4th generation) will support Siri AI and Apple Intelligence — https://aihot.news/items/fi6ahpxthdbszqij928xydgj0
[9] Gemini Call for Me may expand to calls with friends and family, and can send a "running 15 minutes late" message on the user's behalf — https://aihot.news/items/mid9e4m377uvvg4frqyrav39s
[12] Meta and Microsoft cut employee use of Claude — https://aihot.news/items/c2nkq5vvxv1qmsjls4wd3g4je
[14] Shanghai AI Lab's SkillGym paper: Training on verified skill execution trajectories can improve Qwen3.5-35B-A3B by 19.10 percentage points on Terminal-Bench 2.1 — https://aihot.news/items/ysoc64sr9161851v9k