AIが期間内の動向を整理
The move to put AI to work advances across Claude, GPT-5, and Google Cloud
June 17 featured clear moves from trying AI to embedding it into work. Anthropic shared usage analysis for Claude Code, OpenAI 공개ed a pre-release evaluation method for new models, and Google Cloud announced conversational extensions for data analysis.
SOURCE CHECK
Primary Sources 7
Primary Sources
Key Points
- 1Anthropic showed that Claude Code is increasingly used with a split of planning by humans and execution by AI
- 2OpenAI released Deployment Simulation to estimate new-model behavior using past conversations
- 3Google Cloud expanded conversational agents for BigQuery, Looker, and related tools to make analysis easier in natural language
- 4Claude Corps was also announced to train people who can use Claude in real work, advancing both adoption and training
Anthropic: Claude Code is moving toward “humans plan, AI executes”
Anthropic analyzed 400,000 Claude Code sessions and found that usage is increasingly centered on a split where people decide the plan and the agent handles execution. It also showed that people with stronger domain knowledge succeed more often, which suggests AI use is not only about coding skill.
OpenAI: pre-release evaluation is getting closer to real operations
OpenAI’s Deployment Simulation uses past conversation data to simulate candidate models and predict the rate of undesirable behavior. With anonymized data, it aims to produce signals closer to real usage than traditional evaluations. This is useful for companies that want to assess risks before adopting a model.
Google Cloud: data analysis becomes easier through conversation
Google Cloud expanded conversational agents for BigQuery and related products so users can ask questions in natural language and move from analysis to root-cause checks and dashboard summaries. That makes data easier to use not only for analysts but also for business teams.
Adoption and training are advancing together
Anthropic also announced Claude Corps, a fellowship program to train early-career talent to use Claude in practical settings. The message is clear: organizations need not only AI tools, but also people who can use them effectively in daily work.
Related News Items
Anthropic / guide
Domain knowledge helps intermediate users succeed with Claude Code
OpenAI / feature
Easier to predict model behavior using real deployment data beforehand
Google / feature
Google makes data analysis easier through conversation
Anthropic / press_release
Train more people to put Claude into real work