Project Glasswing: Securing critical software for the AI era

Key Takeaway: Anthropic unveiled Mythos, a highly powerful AI model for cybersecurity, but its tightly controlled release suggests the move may be as much about protecting its competitive edge and enterprise business as protecting the internet.

More Insights:

• Anthropic launched Mythos under Project Glasswing, limiting access to select partners for defensive cybersecurity work.

• The company says Mythos found thousands of zero-day vulnerabilities, including many long-hidden critical flaws.

• Mythos was previously leaked under the codename “Capybara,” where Anthropic described it as its most powerful model yet.

• Critics question whether the restricted rollout is really about safety, or about blocking distillation and locking in enterprise customers.

• The strategy reflects a bigger AI industry shift: frontier labs are increasingly gating top models behind closed partnerships instead of broad public release.

Why it matters: This is bigger than one model launch—it signals a future where the most advanced AI may no longer be openly available, but reserved for governments and major corporations. If that trend continues, the real divide in AI may not be intelligence, but access.

Your Kafka Bill Is an Architecture Problem

More than 80% of Kafka costs aren't hardware – they're interzone networking fees. WarpStream BYOC eliminates them entirely by replacing stateful brokers with stateless agents that write directly to your own object storage. 

No disks, no inter-AZ replication, no partition fees. Goldsky cut TCO by 90%+. Your existing clients keep working – just point them at a new URL. 

Learn how it works, then sign up for free. Get $400 in credits that never expire. No credit card required to start.

Learn More

The OpenAI Foundation is working to finalize over $100M in grants to support and accelerate Alzheimer's research

Key Takeaway: OpenAI Foundation is launching a major AI-driven Alzheimer’s research effort—starting with over $100 million in grants—to speed up drug discovery, biomarker development, open data creation, and testing of new and existing treatments.

More Insights:

• Alzheimer’s is framed as one of medicine’s toughest unsolved problems, with rising impact on patients, families, and healthcare systems.

• OpenAI Foundation plans to support six research institutions and continue making Alzheimer’s grants through 2026 and beyond.

• The strategy focuses on five areas: mapping disease causes, designing AI-assisted drugs, building open datasets, improving biomarkers, and testing off-patent treatments.

• The article argues AI is especially useful because Alzheimer’s is driven by many interacting biological and environmental factors over time.

• The broader goal is not just to fight Alzheimer’s, but to prove AI can unlock breakthroughs across other complex diseases too.

Why it matters: This matters because Alzheimer’s has stubbornly resisted traditional science for decades, and this effort suggests a shift from studying the disease in pieces to confronting its full complexity at once. If AI can help crack a problem this difficult, it could redefine what humanity expects from medicine itself.

Databricks co-founder wins prestigious ACM award, says ‘AGI is here already’

Key Takeaway: Databricks co-founder Matei Zaharia has won the 2026 ACM Prize in Computing for his groundbreaking work on Spark, and he says the future of AI depends on treating machines as powerful new systems—not as imperfect humans.

More Insights:

• Zaharia received the 2026 ACM Prize in Computing and is donating the $250,000 award to charity.

• His PhD work at UC Berkeley led to Apache Spark, which helped transform big data and launched Databricks.

• Databricks has grown into a massive cloud and AI company, reaching a $134 billion valuation and a $5.4 billion revenue run rate.

• Zaharia believes “AGI is here already,” but says people misunderstand AI by judging it with human standards.

• He warns that AI agents can be highly useful yet dangerous, especially when trusted like human assistants with sensitive access.

• He is especially excited about AI’s future in research, engineering, science, and reliable information discovery.

Why it matters: This matters because Zaharia’s view pushes a deeper question: maybe the biggest AI breakthrough is not building machines that think like humans, but learning how to use non-human intelligence wisely. The real challenge may be less about whether AI is “smart enough” and more about whether society is mature enough to direct its power safely.

Other stuff

• ChatGPT finally offers $100/month Pro plan

• The 70-Person AI Image Startup Taking on Silicon Valley's Giants

• A gap in understanding AI is growing

• OpenAI Tells Investors It Has Computing Advantage Over Anthropic

• Scientists invented a fake disease. AI told people it was real

Are you tracking agent views on your docs?

AI agents already outnumber human visitors to your docs — now you can track them.

See your AI traffic!

Unclassified 🌀 

• Join a live workshop to create, secure & deploy AI agents using OpenClaw + Docker Model Runner (Run locally). Reserve your spot

Hope you enjoyed today's newsletter

⚡️ Join over 300,000 people using the Superpower ChatGPT extension on Chrome and Firefox.