😺 Amazon’s 80% AI usage mandate = 6-hour outage.

PLUS: New NVIDIA, Google models & Duolingo for everything  ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌

March 12, 2026   |   Listen Online   |   Read Online  Sign Up · Advertise Welcome, humans ICYMI: We interviewed Danny Wu, Canva's Head of AI Products, on how they're building a "Creative Operating System" from 24 billion AI uses. Plus we're LIVE today at 1pm ET breaking down the agent wars: Microsoft, Google, Anthropic, and OpenAI all shipped this week. Special shout out to the sponsor of today's new video, Cohesity! Check them out.  Watch now: YouTube | Spotify | Apple Podcasts LATER TODAY: we're going LIVE today at 10pm PT | 12pm CT | 1pm ET.  Every major AI company is shipping agents this week (Microsoft, Google, Anthropic, OpenAI), and we're breaking it all down with demos, hot takes, and beginner-friendly tips. Join us: YouTube | LinkedIn | X Here's what happened in AI today: 🙀 Amazon's AI-generated code caused a string of outages, including a 6-hour retail site crash. 📰 The U.S. Senate approved GPT, Gemini, and Copilot for official use by Senate aides. 📰 Anthropic launched The Anthropic Institute to study AI's societal impacts. 🎓 How to actually use AI coding tools without breaking everything (hint: think like an architect, not a speedrunner). 🍪 NVIDIA's Nemotron 3 Super just dropped as a 120B open-weight model you can run locally with 1M context. Want to reach 675,000 AI-hungry readers? Click here to advertise with us. 🙀 Amazon's AI Code Is Breaking Amazon If you've ever shipped code on a Friday and immediately regretted it, imagine doing that with AI-generated code... across one of the largest e-commerce platforms on Earth. Amazon just held a mandatory engineering meeting after a string of outages hit its retail website and app, including a six-hour crash last week that left customers unable to check out, see prices, or access their accounts. An internal briefing note described the incidents as having a "high blast radius" and being related to "Gen-AI assisted changes." Here's what we know: Senior VP Dave Treadwell acknowledged that "best practices and safeguards" around AI coding tools aren't fully established yet Junior and mid-level engineers now need senior sign-off on any AI-assisted code changes AWS separately suffered a 13-hour outage in December after its Kiro AI tool deleted and recreated an entire coding environment Amazon has dashboards tracking whether engineers hit minimum daily AI usage targets Amazon disputes that AI wrote the bad code; they say it's a "user error" and "protocol" problem Meanwhile, on Reddit, current and former Amazon engineers paint a grimmer picture. One described "on-calls using AIs to fight each other's AIs in a proxy war of blame." Another said delivering projects matters more than whether projects actually work. Sounds healthy. Amazon also recently laid off 16,000 workers in January, mandated 80% AI tool usage targets (again, dumb; why is "usage" the metric and not "make our products better?"). Oh, and they're spending $200B on capex this year. So do the math and we have fewer engineers + more AI-generated code = more mandatory WTF did we just break meetings.  Why this matters: As The Primeagen and AI researcher Demetri Spanos discussed this week: the models are smart, but they're not THAT smart. Actually, the practices around how they are used is the real problem. Most of the excitement since December 2025 comes from maturation of agent loops and team workflows, not raw model improvements alone. AI can write code, but the code it writes is often much more verbose than it needs to be, and when some poor human (likely you) has to go in and read it, unlike other large well known codebases around the world, no one knows wtf this one says… including you.  So how do you set up your processes to catch the code that AI does write wrong? And how do you set up your systems to get AI to write code efficiently. As Dylan Patel said in a recent Matt Berman interview, some of his team is spending something like $5K a day on Claude Code tokens. How many tokens do you think that guy is spitting out?  Amazon will figure this out. But they're learning the lesson every company using AI for production code will eventually learn: the speed you gain means nothing if you can't trust the output. FROM OUR PARTNERS Ever wondered what Reddit's AI SOC roadmap actually looks like? Watch this exclusive session with the Reddit security team to explore how one of the internet's largest platforms is building an AI SOC at scale. In this session, you'll learn: Reddit's overall approach to AI Key challenges that led Reddit to build an AI SOC The governance policies they're putting in place to manage AI safely Their forward-looking AI roadmap for the SOC A live walkthrough showing how you can build your own AI-powered SOC Watch here 🎓 AI Skill of the Day: The Architect Approach to AI Coding Here's another insight from Demetri Spanos: Most people using AI coding tools are trying to go 10x faster. That's the wrong goal. In the history of software engineering, a 10-20% productivity improvement is already enormous. So here's the real question: can you get 10% better using AI? If so, you're already winning. The mistake most people make is asking AI to generate an entire project at once. Instead, think like an architect. Here's a practical framework from AI researcher Demetri Spanos: The prompt (this is as much a prompt for YOU the engineer to think through as it is for the AI): I'm building [describe your project]. Before writing any code, help me: 1. Break this into separate modules (aim for 10-20 components) 2. Define how each module talks to the others (APIs, data flow) 3. Estimate the approximate size of each module (ballpark lines of code) 4. Identify which modules can be built independently Then generate each module one at a time, separately. Each module should be 200-2,000 lines. Do not generate the next module until I've reviewed the current one. Why this works: AI models write too many lines of code when given open-ended prompts. By constraining each generation to a single module with clear boundaries, you get code that's reviewable, testable, and actually maintainable. You also catch errors per-module instead of debugging a 10,000-line blob. Also, by going slower, you can actually keep track of what you've built when something does inevitably break. The key insight: Know roughly what the components should be, ballpark the size of the code, evaluate each module separately, then assemble. Maintain quality and increase it by 10%. That's the whole game. Now, for non-AI coders, here's one for you: Lance Martin built a Claude Code skill that makes it natively understand Claude API features like prompt caching, adaptive thinking, effort control, tools, and more, with ready implementations across eight languages (GitHub). Want more tips like this? Check out our AI Skill of the Day Digest for this month. Have a specific skill you want to learn? Request it here.  🍪 Treats to Try *Asterisk = from our partners (only the first one!). Advertise to 675K+ readers here! *AI+ Renaissance is the headline AI startup conference of the year, held in San Francisco. On March 15, the Sunday before Nvidia's GTC, AI+ is bringing together 2,000 founders, researchers, builders, and investors to chart the future of the AI industry. Ticket sales end soon. Get yours now. NVIDIA Nemotron 3 Super is a 120B open-weight model (12B active parameters) with 1M context and native multi-token prediction for faster inference; now available on Ollama, LM Studio, and Cloudflare Workers AI—free to run locally. OpenAI plans to integrate Sora video generation directly into ChatGPT to push weekly active users toward 1 billion after the standalone Sora app dropped from #1 to #165 on the App Store (thread). Wondering turns any topic into a guided learning path with bite-size visual lessons, active practice, and long-term mastery tracking, from the ex-head of NotebookLM (App Store)—free to try. Proof is a document editor where you and AI agents co-write in real time, with agents leaving comments and suggesting edits you accept or reject, and every character tracked to who wrote it (code)—free to try. 📰 Around the Horn "Should" a.k.a "we'll see :)" The U.S. Senate approved ChatGPT, Gemini, and Copilot for official use by Senate aides and voted 99-1 to let states continue developing their own AI regulations. Ford launched "Ford Pro AI" to analyze over 1 billion daily data points from 840,000 commercial vehicle subscribers for fleet optimization, route planning, and predictive maintenance. Perplexity announced Personal Computer, an always-on Mac mini AI operating system that gives its assistant persistent local access to your files, apps, and sessions. Anthropic researchers found that 10 out of 16 leading chatbots helped plan violent attack scenarios when prompted; only Claude refused 100% of queries. Want absolutely EVERYTHING that happened in AI this week? Click here! FROM OUR PARTNERS Allow us to reintroduce… Slackbot (now GA) Think of Slackbot as your always-on teammate. It understands your conversations, files, and workflows to deliver what you need, right when you need it without any setup.  Instead of hunting for information, Slackbot synthesizes what you need — respecting your permissions and using only what you can already see.  Watch now Thursday Trivia One is AI, and one is real. Which is which? Vote below! A. B. Which is AI, and which is real? Which is AI, and which is real? The answer is below, but place your vote to see how your guess everyone else (no cheating now!) A is AI, and B is real. B is AI, and A is real. A Cat's Commentary More from the TA Family Microsoft Agent Flaw Enables Remote Code Execution via AI Agents — eSecurity Planet Perplexity Comet Browser Bug Leaks Local Files via AI Prompt Injection — eSecurity Planet Trivia Answer: A is AI, and B is real; fair warning, after clicking the video from B, I started getting a lot of Norwegian dance videos in my TikTok algo, and I'm… not mad about it??   That's all for now. What'd you think of today's email? 🐾🐾🐾🐾🐾 Like a hit of catnip 🐾🐾🐾 Good, not great 🐾 It sucked P.P.S: Love the newsletter, but only want to get it once per week? Don't unsubscribe—update your preferences here. Advertise with us and get in front of over 650,000 AI enthusiasts Update your email preferences or unsubscribe here © 2026 TechnologyAdvice, LLC 3343 Perimeter Hill Dr. Ste. #215 Nashville, Tennessee 37211, United States Terms of Service