
Anthropic just became a public company, and you're still the customer. The Claude maker hit the Nasdaq this week with shares opening at $111.50, a 33% jump over its $83 IPO price, pushing its market cap past $2 trillion. Bigger than Meta. Bigger than Tesla. The nearly 100-page risk filing is now public, buyers of Anthropic stock get zero voting rights, and there's $220 billion sitting on the balance sheet to fund future R&D. What changes for you on Monday morning? Nothing. Claude still works the same way. But that IPO cash is what pays for every model improvement coming down the road. If you've built any workflow on Claude, the company behind it is now subject to public shareholder pressure, quarterly earnings calls, and the scrutiny that comes with that. Worth knowing. Full story here.
OpenAI dropped prices this week, and if you're running any volume through the API, it's worth pulling up your last invoice. GPT-6.2 Astra now costs $8 per million input tokens and $40 per million output tokens. That's 20% cheaper than GPT-6 Astra. The smaller Sol model holds at $2/$10. Both are live in ChatGPT, Codex, and the API. Independent benchmarking shop Artificial Analysis already ranks Astra second overall, behind Claude Fable 5.2. If your team runs heavy automated tasks, customer service bots, or any kind of API pipeline, re-run the math this week. A 20% cut on a real workload is real money. Details from OpenAI.
Free trials sound like a growth move until someone farms your compute at scale. Claire Vo, founder of ChatPRD, lost tens of thousands of dollars over the past year to trial fraud. Bad actors weren't just signing up for free accounts. They were reselling promo codes, wrapping her product in relay layers, and running a custom MCP (a tool that lets AI models call outside services, basically a 24/7 worker) as a compute drain that never stopped. Her fixes: Stripe Radar to flag suspicious payment patterns, Vercel's firewall to block bot traffic, automated watchers monitoring usage anomalies, and killing free trials entirely. If you're building or already running any AI-powered product, her write-up is the most practical fraud-prevention playbook I've seen. Read it before you launch your next free tier. Her full breakdown here.
Tools Worth a Look
Codemany: Speeds up GitHub Actions workflows up to 3x faster. Drop-in for teams already using CI/CD pipelines for testing, deployment, or automation. Pricing not listed.
More Headlines
Google built a "mirror dimension": an isolated synthetic internet with thousands of fake websites, so its AI browser agent can practice multi-step tasks like retail checkouts without touching real accounts or placing real orders. A useful preview of how AI agents will be safety-tested before they handle your customers' purchases.
Google added AI-ready instructions for AdMob so developers can use Claude Code, Codex, or Cursor to build and audit ad integrations automatically. If you have an app monetizing with AdMob, this cuts a lot of manual back-and-forth.
Every's team let AI agents make real purchases and found agents handle refills and repeat bookings reasonably well, but bought duplicates and wrong variants when given too much latitude. Their rule: agents for replenishment, human approval for everything else.
Tau Robotics demoed a humanoid robot frying and assembling a burger, expanding beyond its original home-cleaning focus. Worth watching if you're in food service.
Raycast Primary Keyboard for iOS is now available as your default iOS keyboard.
-Scott
SmartOwner is published (almost) daily by the team at DigitalTreehouse. Want AI consulting or automations for your business? Reply to this email.
