Happy Friday! Meta would like to know how you’re feeling. It would also like to hear you sigh, watch you take your meds, and track all the times you get hangry for skipping breakfast. Meta filed a patent for an AI wearable that tracks your emotional state. It will record your voice and surroundings throughout the day, then build a timeline showing how your mood changed. All perfectly normal behavior from a company built on targeted advertising.
🏖️A quick programming note: Summertime is here and we’re about to slow down. Taking a break to touch grass and recharge, see you back here in a few weeks.
Everyone said AI would get more expensive. Instead, the price of coding intelligence is collapsing. Meta priced its new coding model at roughly a quarter of what Anthropic and OpenAI charge. Musk priced Grok 4.5 at $2 per million input tokens while making it 4x more efficient than Claude on coding tasks. Even OpenAI’s newest model uses 54% fewer tokens for agentic coding. Coding agents burn through mountains of tokens as they loop through files, test fixes, and keep working until the job is done. Lower token prices and fewer tokens per task dramatically increase how many agents a company can afford to run. Anthropic and OpenAI are chasing better margins for investors. Zuckerberg and Musk can afford to chase market share. And they’re all competing with China. (Quartz, VentureBeat)
ChatGPT is moving from answers to outcomes. OpenAI just dropped ChatGPT Work, an AI agent that can use your files and apps to build whatever you ask for. Ask it to prepare for a sales meeting and it can dig through the account history, build the presentation, and keep updating it as new info arrives. It can also break big projects into smaller jobs, then launch agents to handle them at the same time. Anthropic got there first with Claude Cowork. Now both companies are racing to build software that learns how your business works, then starts running the software for you. Teach it a process once, and it can launch agents to repeat the work at scale. SaaS companies better build some fat APIs, because they’re about to learn what it feels like to be a tool. (OpenAI)
Stop hallucinations. Power your agents & chatbots with real-time data from the Brave Search API:
- Specialized endpoints built for LLMs
- 40B page index, no Google scraping
- $5 per 1,000 calls, plus free monthly credits
- RAG pipelines, Claude MCP, OpenClaw, Hermes, and more
👀 closer look
The EU is forcing Google to open Android to AI rivals. Google built Gemini deep into Android, giving it access to what’s on your screen and letting it take actions across your apps. In 17 days, Google will have to give rival assistants the same access. People could choose Claude as their default assistant, then use it to send emails or order dinner without opening another app. Google says that would create privacy and security risks. Apple lost a similar case this week, closing Google’s last path to block the Android rules before they take effect. Europe is about to make “bring your own AI” the rule for the world’s most popular mobile operating system. (Tech Times)
Anthropic just invited Ben Bernanke into its inner circle. The former Fed Chair now sits on a trust that can appoint and remove most of Anthropic’s board. Bernanke spent his career studying depressions and steering the US through the 2008 financial crisis. Now he’s helping govern a company that believes AI could have the biggest economic impact of any tech in modern history. Anthropic is also nearing a $1 trillion valuation and possible IPO while fighting Washington over how its models can be used. The trust is supposed to keep Anthropic accountable to the public. Adding one of the world’s most respected crisis managers also makes the company look safer to investors, regulators, and governments. Anthropic is preparing for a world where AI labs wield enough economic power to need their own central banker. (Seattle Times)
fun stats
🤖 41%. Longform LinkedIn content likely written entirely by AI. Nearly half of all posts on X are AI generated too, as AI content floods social platforms.
🤙 25. Degrees of freedom in the humanoid robot Neo’s hands, putting them close to human dexterity.
👨🦰 6%. Drop in oversight when managers reviewed work from a named AI “teammate” instead of an unnamed AI agent. 1 in 3 companies now treats AI like an employee, yet it weakens review quality and trust.