Anthropic released Claude Fable 5 and Mythos 5, sharing the same base model with different safety guardrails. Fable is for everyone, Mythos is for vetted partners only. API pricing: $10/$50 per M tokens. Pro/Max subscribers get free access until June 22.

Model capability is accelerating faster than adaptation. Fable 5 output costs 57x more than DeepSeek V4 Pro. "More capable models cost more" is becoming the norm.

Claude Pro/Max usersDevelopersAI tool evaluation

Tim Cook's last WWDC. Siri backend switches to Google Gemini, gaining cross-app context understanding and multi-step action execution. Photos Spatial Reframe lets you recompose shots. iOS 27 focuses on performance.

Apple is not building a standalone AI chat product. It weaves AI into existing apps and system interactions. Siri is "the unified entry point for all apps on your phone," diverging from both OpenAI and Google.

Apple usersConsumer tech watchersProduct observers

Anthropic filed S-1 confidentially on June 1 ($965B). OpenAI filed on June 8 ($852B). Combined pipeline exceeds $1.8 trillion. AI's two most important companies start quarterly reporting to Wall Street.

Going public means quarterly earnings, explaining growth, facing market patience. Business model divergence amplifies: OpenAI runs ads+subscriptions+API, Anthropic runs pure subscriptions+API+enterprise. Anthropic surpassing OpenAI in valuation signals market preference for B2B.

Industry watchersInvestors / analystsClaude users

LLM-driven Data Agents (Databricks Genie, Snowflake Cortex, etc.) are widely deployed. This paper reveals a new attack surface: individually harmless queries that leak sensitive data when combined. Three vulnerability layers, 14 attack techniques, all six tested systems have real gaps.

Agent security is not just about harmful content. When agents connect to real data and execution environments, the analysis process itself becomes the attack surface. Enterprise database controls still reflect pre-agent design paradigms.

Data engineers / DBAsAI Agent developersSecurity professionals

ChatGPT ads expand from US to UK, CPM ~$60, annualized ad revenue past $100M. Gemini 3.5 Flash ships at 284 tokens/sec, positioned for speed and cost. Apple Siri AI also runs on Gemini under the hood.

Behind both items: AI product business models are diverging. OpenAI monetizes with ads (Altman called it "a last resort" two years ago), Anthropic explicitly won't serve ads, Google competes on infra with speed-tier models.

ChatGPT usersAI tool evaluationIndustry watchers

The community-built Chinese culture game China Quest went from idea to live deployment to a school Multicultural Night, with feedback far exceeding expectations. Three modules shipped. Multiple contributors collaborated, including first-time Git users. This issue includes a project video.

This project proved that a group of motivated people can build something genuinely useful in a short time with AI-assisted tools, no professional dev team needed. The full loop from idea to on-site positive feedback closed successfully.

Community project fansNon-technical AI buildersSchool event organizers

A prompt template that uses familiar life contexts (cooking, shopping, driving) to turn any subject into an interactive HTML page with progress bars, instant feedback, and level-up rewards. Fill in four blanks, paste into Claude or ChatGPT, get a playable learning game in ten minutes.

The biggest barrier to learning is not difficulty, it is irrelevance. Using experiences the learner already understands to explain new concepts cuts comprehension time in half. No tech background needed. One template, unlimited reuse.

Parents with school-age kidsSelf-learnersTeachers / tutors