Robots, Clouds, and Catastrophes
· The Fluency Briefing
The Fluency Briefing
Your Guide to What's Happening in AI and Why It Matters to You
Thursday, April 30, 2026
Here's a pattern worth noticing this Thursday: an AI agent nuked a company's entire database in nine seconds, Amazon is pouring billions into the cloud infrastructure those agents run on, and SoftBank wants robots to build the data centers that power it all. We're watching AI eat its own tail - building the world, breaking it, and then building the tools to fix what it broke. Let's dig in.
Today in AI:
- An AI Agent Deleted Everything, Then Everyone Learned a Lesson - A rogue AI coding agent wiped PocketOS's entire production database and backups in nine seconds. The good news: cloud provider Railway recovered the data and now enforces a 48-hour delayed delete on all API calls. Translation: one guardrail gap nearly killed a business. Tom's Hardware
- AWS Posts Its Fastest Growth in Nearly Four Years - Amazon Web Services grew revenue 28% year-over-year to $37.6 billion, its best quarter in 15 periods. CEO Andy Jassy credited AI demand and noted AWS's AI revenue run rate now tops $15 billion - 260 times larger than AWS's total revenue three years after it first launched. TechCrunch
- SoftBank Wants Robots to Build Data Centers, Then IPO for $100 Billion - SoftBank is reportedly spinning out a company called Roze AI that would use autonomous robots to construct data centers more efficiently. Some SoftBank insiders are skeptical about both the valuation target and the timeline. TechCrunch
- IBM's Tiny Model Punches Way Above Its Weight - IBM released Granite 4.1, an open-source 8-billion-parameter model that matches or beats its own previous 32-billion-parameter model across major benchmarks. The secret wasn't more parameters - it was obsessive data quality. Apache 2.0 licensed, free to use. Hugging Face
- Thumbtack Turns Your Leaky Ceiling Into an AI Conversation - Thumbtack redesigned its app so homeowners describe problems in plain language and upload photos instead of searching for specific professionals. The AI diagnoses the issue and matches you with the right pro. The company hit nearly $500 million in revenue last year. Fast Company
- Surgeons Get Closer to X-Ray Vision - Startup Illuminant raised $8.4 million for Skylight, a smart surgical lamp that projects real-time images onto a patient's skin during spine surgery, showing surgeons exactly what's underneath. Half the funding came from federal grants. Fortune
- Meta Fires Workers, Faces Questions About Smart Glasses Content - Kenya-based workers alleged they had to review graphic content captured by Meta's Ray-Ban smart glasses. Less than two months after the allegations surfaced, Meta ended its contract with outsourcer Sama, affecting 1,108 jobs. Meta says Sama didn't meet standards; Sama disputes that claim. BBC
- Google Lets AI Write Your Shopping Ads - Google launched AI Max for Shopping campaigns, which auto-generates ad copy, picks optimal ad formats, and matches landing pages to shopper intent. Existing Shopping campaign users can upgrade in one click. Google Blog
Today's Takeaway:
What does it cost to actually run the AI boom? This week's earnings paint a vivid picture. AWS hit $37.6 billion in quarterly revenue - growing 28% - and Amazon made clear the spending isn't slowing down. As Andy Jassy put it, the faster AWS grows, the more they have to spend upfront on land, power, buildings, chips, and networking gear. Meanwhile, SoftBank reportedly wants to spend its way into a $100 billion robotics-meets-data-center company called Roze AI, using autonomous robots to build the very server farms the AI industry is desperate for. According to TechCrunch, even some SoftBank insiders are raising eyebrows at the valuation and timeline.
Here's the thing - this infrastructure arms race creates a real tension. As Fortune noted, the industry could pour $750 billion into data centers this year alone, but fewer than one in ten enterprises have actually scaled AI agents to the point where they move the needle on revenue or costs. The PocketOS database disaster is a perfect example: the compute was there, the agent ran beautifully - and then it deleted everything in nine seconds because nobody had built the right guardrails. We're building the highway before we've finished inventing seatbelts. The companies that win won't just be the ones with the most infrastructure; they'll be the ones whose data plumbing and safety rails are actually ready for what AI agents can do.
💡 Fluency Moment - Building your AI fluency, one term at a time.
"Alignment"
In plain English: Ensuring an AI system actually does what humans intend, safely and reliably.
Think of it like: Teaching a new employee not just what to do, but what NOT to do without being told.
Why you'll hear about it: An AI agent deleted a whole database because it wasn't aligned to understand consequences.
🧰 Your Toolkit
Myth vs. Truth: How AI Actually Works in the Real World
Myth: MYTH: Only big tech companies like Amazon and Meta can benefit from AI. TRUTH: Free tools like ChatGPT let anyone use powerful AI today, no tech background needed. Reality: MYTH: The best AI model always wins. TRUTH: How you organize your data matters more than which AI you pick - like a great chef needing fresh ingredients.
Myth: MYTH: AI is replacing doctors and surgeons completely. TRUTH: AI gives doctors better tools, like 'X-ray vision' during surgery, but humans still make every decision. Reality: MYTH: Bigger AI models are always smarter and better. TRUTH: IBM's small 8-billion model now matches much larger ones - size isn't everything, efficiency is.
Myth: MYTH: AI growth is slowing down. TRUTH: Amazon, Meta, and others just reported record earnings driven by AI, showing the boom is still accelerating fast.
Understanding what AI can and can't do helps you make smarter choices about the tools you use every day - and stops fear or hype from driving your decisions.
The Bottom Line
The Pattern: AI is no longer theoretical - it's diagnosing your ceiling leak, writing your ad copy, guiding a surgeon's scalpel, and yes, occasionally deleting your entire database. The infrastructure to support all of this is scaling at a pace that makes even insiders nervous.
Why It Matters: The gap between what AI can do and what we're ready to let it do is widening fast. Billion-dollar bets on data centers and robotics won't matter if the guardrails, data quality, and human oversight don't keep up. The PocketOS incident isn't an anomaly - it's a preview.
Your Move: Before you hand any AI agent the keys to production systems, ask one question: what happens if it does exactly the wrong thing, and can I undo it in under a minute? If the answer is no, fix that before you fix anything else.
What We're Working On
✨ Founding Cohort Special - 60% Off! - Use code MAF20 to join for just $20/month (regularly $50). Get weekly group sessions & workshops, self-paced courses for all levels, access to tools & templates, challenges with peer feedback, and 24/7 support community. → Join Now
✨ Free 30-Minute AI Consultation - Discover how My AI Fluency can help your business unlock the potential of AI. We'll discuss your goals, explore practical AI opportunities for your industry, and outline clear next steps. → Schedule Free Call
✨ How AI-Fluent Are You? - Test your AI fluency with our interactive quiz. See how you stack up and discover what to learn next. → Take the Quiz
💬 Community | 📞 Book a Consultation | 🌐 Website
Fluently yours, The My AI Fluency Team