Sponsored by
Hey {{first_name | there}},
Let me start with a number that should concern everyone building with AI: 2,500.
That's how many organizations had their credentials exposed in the LiteLLM supply chain attack this week.
This wasn't a minor incident btw. We're talking terabytes of data, cloud keys, SSH tokens, Kubernetes credentials, leaked through a single compromised vulnerability scanner called Trivy.
The attackers used this to poison LiteLLM (in case you don’t know, LiteLLm is an open-source AI gateway that thousands of companies use to manage their AI infrastructure).
And here's what makes it worse: this happened in March 2026. The full scale only came to light last week.
But the thing is we know this is happening but we're still using these tools. I've talked to people this week who didn't even know this happened. And the ones who did know? They're rotating credentials and moving on.
Is that the right call? Or are we building castles on compromised ground?
The Mysterious Model That Nobody Built
Speaking of trust issues, have you heard about Ox Alpha?
It’s a new AI model that randomly spawned on OpenRouter about a week ago. There was no company announcement, no press release. Nobody knows who built it.
What we do know: it has a 1 million token context window, it outperforms Claude Fable 5 and GPT-5.6 Sol on coding tasks, and it's currently free.
The coding community lost its mind. Developers ran tests, posted comparisons, and debated whether it's actually GLM-5.3 wearing a mask (the tokenizer signatures are similar).
Some speculate it's a major lab testing a new release strategy. Others think it's a government project. A few think it's just extremely well-funded research.
My theory? They're collecting data.
Every API call teaches the model something. Anonymous releases eliminate accountability concerns while maximizing real-world training signal. It's essentially giving away a product to learn from how people use it.
What we should be thinking about is: if anonymous AI models can match or exceed established leaders, what happens to Anthropic's Fable pricing strategy? What happens to OpenAI's market position? The entire "premium AI" business model assumes that better performance justifies higher prices. Ox Alpha just challenged that assumption publicly.
AI Is Eating Its Own Lunch (And Tokens)
And just so you know, AI agents now consume five times more tokens than humans.
More AI is calling AI than humans are.
But here's the twist that makes this interesting: over 70% of agentic token consumption comes from cached prompts. This means agents are reusing the same context repeatedly rather than generating fresh responses every time.
It's AI being efficient with AI, somewhat.
For builders, this changes everything. If your product isn't agent-compatible, you're building for a shrinking user base. Humans will always exist, but they're becoming the minority users of AI infrastructure.
This is why every SaaS tool is scrambling to add "agentic capabilities" because the customers are increasingly other AI systems.
What This Week's Events Mean For You
Let me connect these dots:
Your AI infrastructure is more vulnerable than you think. The LiteLLM breach shows that security in the AI stack is still an afterthought. If you're using open-source AI tooling, audit your dependencies.
The premium AI pricing bubble might pop. Ox Alpha proves that anonymous models can compete with named ones. As more players enter the market, pricing power erodes.
Agents are the real growth driver. If you're not building for agentic use cases, you're optimizing for a shrinking market segment.
AI is changing quality standards everywhere. From scientific publishing to software development, the definition of "good enough" is shifting. Adapt or get left behind.
What's your read? Reply and tell me which of these stories concerns you most. I'm building my WhatsApp community where we discuss exactly this kind of stuff, AI infrastructure, security, and what it means to build in the age of AI eating itself.
Join here – https://chat.whatsapp.com/IXt9FJIblNs8tu36JIwWbd
- Aashish
How 2M+ Professionals Stay Ahead on AI
AI is moving fast and most people are falling behind.
The Rundown AI keeps you ahead of the curve.
It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.
Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses — tailored to your needs.