OpenAI Lost $12.3 Billion in One Quarter. Again.

Today's issue starts with a number: $12.3 billion. That's what OpenAI lost in three months, while the AI company most of your customers have never heard of quietly moved into the black. If you've assumed the AI company with the loudest headlines is also the safest bet, today's numbers might change your mind. We've also got a Microsoft security patch you actually need to check on, and a study on what AI homework help is doing to kids' real grades.

5,000+ professionals use Skill Leap to get ahead with AI. Right now, two of their best courses are completely free - Claude 101 and the 14-Day AI Boot Camp.

Claude 101 covers prompting frameworks, Artifacts, file analysis, and real-world workflows in 19 lessons.

The Boot Camp covers ChatGPT, Gemini, Midjourney, and prompt engineering in 16 lessons. Downloadable workbooks. LinkedIn certificate.

Zero cost, no credit card, no catch.

Get 2 Free AI Courses

What happened: Anthropic reported $11.6 billion in second quarter revenue, more than double its first quarter, and posted its first operating profit of $559 million, according to qz.com. OpenAI's Q2 revenue came in at $6.7 billion, up 18% from Q1, but its operating loss widened to $12.3 billion from $9.3 billion the quarter before, according to the Wall Street Journal. It's the first quarter Anthropic's revenue has topped OpenAI's.

Why it matters to you: Anthropic built its business selling to companies that pay and renew, while OpenAI has spent heavily supporting hundreds of millions of free ChatGPT users who may never convert to paying customers. If you've picked an AI vendor based on brand recognition alone, this is a reminder that buzz and financial stability aren't the same thing.

What to do about it: Before you build a workflow around any single AI tool, check whether the company behind it is actually making money, not just making headlines.

What happened: Security researchers found a flaw nicknamed "CoSnitch" that let a single malicious link trigger Microsoft Copilot into leaking data from a victim's connected accounts, with no additional click required beyond opening the link, according to The Hacker News. Microsoft has now shipped a fix, but CSO Online reports the company knew about the vulnerability for almost eight months before patching it.

Why it matters to you: If your business uses Microsoft Copilot connected to Outlook, OneDrive, or Teams, this is exactly the kind of flaw that could have quietly pulled data out of your accounts without anyone clicking "download" or "send."

What to do about it: Make sure Microsoft 365 and Copilot are fully updated, and review what your AI tools are actually connected to. If a hijacked assistant could reach everything, that's not convenience, that's exposure.

What happened: Stripe is acquiring OpenRouter, the service that lets developers switch between more than 400 AI models through one connection, in a deal reportedly worth more than $7 billion, according to TechCrunch. OpenRouter currently routes more than 10 trillion tokens a day across the industry.

Why it matters to you: A lot of the AI tools you already use, including chatbots, writing assistants, and automation platforms, may be quietly built on top of OpenRouter without you ever seeing its name. When the plumbing behind an industry gets bought for billions, pricing and availability for everyone downstream usually shifts.

What happened: Anthropic published research showing its Claude models ran protein design experiments largely unsupervised, producing molecules that bound to their intended targets on 14 out of 15 attempts, at success rates above the industry average, according to Anthropic. Twist Bioscience and Adaptyv Bio built and tested the actual molecules in their labs; Claude handled the design work.

Why it matters to you: This isn't a chatbot answering questions anymore, it's AI running a real step of the drug discovery process with minimal human hand holding. You don't need to do anything about this today, but it's a preview of how fast "AI drafts an email" is turning into "AI runs entire technical workflows."

What happened: College students in the US can now claim a full year of Google's AI Pro plan, normally $19.99 a month, for free, including extra Gemini access, 5TB of storage, and new study tools, according to Google. Students outside the US get a smaller free plan in more than 140 countries. The offer runs through December 31, then switches to full price automatically.

Why it matters to you: If you've got a kid, employee, or intern in college, this is worth roughly $240 in free software over the next year, no catch beyond eventually needing to cancel or start paying.

What to do about it: If someone in your life qualifies, have them sign up before December 31 and set a reminder to cancel or downgrade before the free year runs out and billing kicks in automatically.

What happened: Cerebras unveiled its CS-4 system, which the company says delivers up to 30 times more tokens per second per user than standard GPU-based systems, combining three of its wafer-scale processors into one rack, according to Cerebras. The company is positioning it as a direct challenge to Nvidia's dominance in AI inference hardware.

Why it matters to you: Faster, cheaper AI computing eventually shows up as lower prices or faster response times in the tools you already pay for. It won't happen overnight, but this is the kind of hardware race that tends to trickle down to your monthly software bill.

What happened: A study of roughly 26,000 to 27,000 Chinese students aged 12 to 18 found AI users raised their homework scores by 18% and cut assignment time nearly in half, but scored 20% lower on exams six months later, according to the South China Morning Post. The drop was concentrated among students who used AI to rush through work; those who spent about as much time on assignments as non-users saw little decline.

Why it matters to you: Whether it's your kids or your employees, using AI to finish something faster isn't the same as using it to actually learn something. The tool isn't the problem, how it gets used is.

What to do about it: If people in your life are leaning on AI for schoolwork or training, encourage them to use it to check their understanding after they've tried the work themselves, not to skip the work entirely.

What happened: Unitree, the Chinese humanoid robot maker known for backflipping robots, raised roughly $904 million in its Shanghai IPO and saw shares surge as much as 542% intraday before closing up 460%, according to CNBC.

Why it matters to you: This is the AI rollercoaster in one headline: a single day of trading added billions in paper value to a robotics company, and nobody knows yet whether that valuation will hold. If you're watching AI stocks or funds as part of your own investing, days like this are exactly why "AI" and "safe bet" don't always belong in the same sentence.

Zoom out and today's issue is really one story wearing eight different hats: money is finally sorting out which AI companies are built to last and which ones were just built to look impressive. Anthropic turned a profit while OpenAI's losses grew. A security hole sat open for eight months because nobody made fixing it urgent. A robotics stock jumped 460% in a single day for reasons that will probably look silly in six months. None of this means panic or swear off AI, it means treat every AI headline the way you'd treat a stock tip from a stranger: interesting, maybe useful, but worth checking before you act on it.

If today's issue saved you from reading eight different newsletters, forward it to one person who needs it. New here? Subscribe free and get tomorrow's issue too.

Talk tomorrow,
Mark Shilensky

Follow me personally on Social Media:
Facebook: https://www.facebook.com/MarkShilenskyPage
Instagram: https://www.instagram.com/markashilensky/
Threads: https://www.threads.com/@markashilensky
LinkedIn: https://www.linkedin.com/in/shilensky/
X: https://x.com/markshilensky
TikTok: https://www.tiktok.com/@mark.shilensky
BlueSky: https://bsky.app/profile/markshilensky.bsky.social