Welcome back! It's recap day, with this week's five biggest stories and five tools to try this weekend. ChatGPT read a letter nobody could crack for 217 years, and Google's new model caught up with it. OpenAI also launched 20+ things in the middle of a rough safety week, while Anthropic lined up what could be the biggest IPO ever.
In today's Generative AI Newsletter:
ChatGPT: What did Napoleon's secret letter say?
Google: Did Gemini 4 really catch up with ChatGPT?
OpenAI: What went wrong after its agents escaped?
ChatGPT Pro: Why does $200 get you half as much now?
Anthropic: How do you lose $42 billion and still ask for $2 trillion?
ChatGPT cracked a 217-year-old secret letter
For 217 years, a coded letter to one of Napoleon's generals sat on the list of ciphers nobody could read. OpenAI's GPT-6 Astra cracked it in about six hours.
Carter Church, an AI engineer who says he'd never broken a cipher before, gave it one scanned page from a 1969 French army journal.
It read all 1,300 handwritten symbols straight off the image.
A partial key already covered about a third of the letter, and Astra worked out the other two-thirds.
The letter is from March 1809, two weeks before Austria invaded Bavaria. It tells General Marmont where every army stood and not to be scared of a few troops or a gathering of rabble.
Church says the man who keeps a long-running catalog of unsolved ciphers has now marked it solved. Five symbols are still uncertain, but they don't change what it says.
Google's Gemini 4 caught up with ChatGPT
Google announced Gemini 4 Argon, its first new flagship model in almost a year. Artificial Analysis, the independent benchmark site, gave it 53, the same as GPT-6 Astra. Claude Opus 5.5 still leads at 58.
It costs a fifth of Astra's price, and on one test it made things up 15% of the time versus Astra's 51%.
You can't use it yet. It's going to the US government and a small group of cyber defenders first, with no date for everyone else.
OpenAI paused training, then fired three safety researchers
OpenAI stopped training its most powerful models for the second time in three months. One of its agents broke out of a test setup with no internet, and the automatic off switch failed. Someone had to stop it by hand two and a half hours later.
It also cancelled GPT-6.1 Astra, its next model, because it wasn't always honest about what it had done.
Then it fired three safety researchers for mishandling sensitive information. The Wall Street Journal reports they shared it with an outside AI safety group.
OpenAI launched 20+ things and cut its $200 plan in half
At DevDay, OpenAI announced more than 20 things. The big one is dots, AI agents that keep working on their own computer after you close ChatGPT.
GPT-6.1 Sol is almost as good as GPT-6 Astra at a fifth of the price. There's also a new $500 Pro plan with 25 times the usage of Plus.
The $200 plan keeps its price but gets half the usage from October 30, so each Plus-worth of usage now costs $20 instead of $10.
Anthropic lost $42 billion and still wants $2 trillion
Anthropic's IPO filing leaked this week. It lost $42 billion in 2025 on $4.6 billion in revenue, though about $34 billion of that is an accounting charge, not cash.
Now Bloomberg reports it wants to go public as soon as mid-November at a valuation of up to $2 trillion. That would be the biggest IPO ever.
Revenue grew 12x last year. A select group of big investors meets the team in San Francisco on October 14.
Top 5 tools of the week
Jobright: Finds jobs you actually qualify for and tailors your resume to each one. It's free to start. For anyone tired of filling out the same application 50 times.
Napkin: Turns your text into diagrams and charts without a single prompt. Paste a paragraph, click Autospark and export it as a PNG or PDF. For anyone who needs a visual and can't design.
AI Plant Doctor: Tells you what's wrong with your plant from one photo, then gives you a step-by-step plan to fix it. Free for 3 scans a week. For anyone who's killed a houseplant and never knew why.
Superhuman: Puts AI inside the apps you already use, fixing your writing as you type and drafting email replies in your voice. It's from the company behind Grammarly. For anyone who lives in email and docs.
TreePad Deep Researcher: Runs your question through the best of 15 research engines, and every claim in the answer links to the exact line in its source. You pay per query, and new accounts get $3 free. For anyone who's been burned by a made-up AI citation.
Learn more about AI from the experts building it
📸 Follow us on Instagram for fast, visual AI updates in 30 seconds.
📺 Watch us on YouTube to hear insights directly from leading AI voices, builders, and innovators.
🐦 Follow us on X for breaking AI news and real-time industry updates.
🧠 Learn how to build your next AI application with practical resources and expert guidance.
🎓 Start learning with free AI courses from GenAI Academy.
💰 Invest in the GTM infrastructure of the AI economy.
