AI Agents Expand Across Work, Evaluation, and the Physical World
에이전트 | Fri Jun 26 2026 00:00:00 GMT+0000 (Coordinated Universal Time) | 5 sources
The agent ecosystem is expanding broadly, from OpenAI Codex handling long-running tasks to Patronus building agent evaluation environments and General Intuition training robots through games.
Analysis
[OpenAI] saw Codex emerge as the company's primary internal AI work tool [1]
- Codex accounts for 99.8% of OpenAI's weekly output tokens
- Non-developer users grew 137x
- Tasks taking over 8 hours are now delegated to it
[Patronus AI] raised $50M Series B to build a 'digital world' for stress-testing AI agents [3]
- Led by Greenfield Partners
- with $70M cumulative funding
- Provides simulation environments replicating websites and internal systems
- Revenue grew 15x over the past year
[General Intuition] reached a $2.3B valuation with agent models trained on video game data [4]
- Raised $320M in new funding
- with $454M cumulative
- Fortnite-playing agents and quadruped robots share the same model
- Only 8 minutes of data needed for fine-tuning on real robots
[Notion] discontinued Notion Mail as most users manage email through AI agents [2]
- Email client based on Skiff acquisition is being sunset
- Auto label rules automatically migrate to Custom Agents
- HIPAA users must transition by June 30
- 2026
[FIFA & Lenovo] provided custom AI agents for teams competing in the 2026 World Cup [5]
- Collects approximately 150 million data points per match
- IMU sensors inside the ball record movement 500 times per second
- England used AI for penalty kick analysis
- reducing time from 5 days to 5 hours
Sources
- [1] How agents are transforming work - OpenAI Blog
- [2] Notion killing Skiff-influenced email app since most users use AI agents instead - Ars Technica AI
- [3] Patronus AI lands $50M to build ‘digital worlds’ that stress-test AI agents - TechCrunch AI
- [4] General Intuition’s $2.3B bet that video games can train AI agents for the real world - TechCrunch AI
- [5] World Cup Teams Are in a Race for AI Dominance - Wired AI