Happy Friday! Get ready to be roasted by your smart speaker. Amazon’s new adult-only Alexa+ upgrade promises to “help first, judge always,” giving sarcastic replies sprinkled with a few f-bombs. She’s not as raunchy as Musk’s spicy companions (yet), but she’ll definitely have some NSFW things to say about your playlist.
Everyone’s pitching enterprise-ready AI employees. Companies have outgrown chatbots and need AI capable of handling tasks independently. Yet as popular as OpenClaw is, it’s a corporate liability. In recent tests, AI agents autonomously broke into confidential systems and disabled security protections that got in their way. Now NVIDIA, Anthropic, OpenAI, Perplexity, Microsoft, and every random dev on Reddit are rushing to sell safer enterprise versions. Genspark just raised $545 million for a cloud-based Claw on Azure, claiming it can limit damage if the agent writes bad code or downloads malware. But are you ready to trust it with your data and API keys? [The Guardian, The Information]
Silicon Valley’s new flex is babysitting agents. Tech workers are competing to see how much of their grunt work they can outsource to AI without messing everything up. It’s become the ultimate productivity brag, with some comparing the dopamine rush to playing strategy games like Age of Empires. Notion’s cofounder even admits that managing AI beats managing people. Engineers are so obsessed they’re losing sleep and checking on their bots before morning coffee. The best can juggle up to 5 agents before chaos ensues, similar to some human teams. Coding is gone. We’ve officially streamlined thinking to two options: accept or reject. [WSJ]
You can’t stream the energy.
Google Cloud Next hits Vegas April 22-24. Be there for insider access to agentic systems, networking, hands-on labs, and 1:1 conversations with the experts building what’s next in AI and cloud.
Get your spot before passes sell out.
Amazon’s AI agent crashed its retail site 4x in one week. One outage lasted 6 hours, completely locking shoppers out of checkouts and accounts. Internal docs blamed an AI agent using an outdated company wiki. Amazon tried to hide the AI connection, but leaked docs exposed the mess anyway. The timing couldn’t be worse, as Amazon’s been aggressively replacing workers with AI, betting robots were ready to fly solo. Instead, chaos forced Amazon into an embarrassing reversal, putting humans back in the loop. Still no word on why their disaster recovery plan didn’t kick in. [Fortune]
Reach your audience
Get in front of 70,000+ tech decision makers with budget and technical influence. 95% US based. 66% open rate. Read daily by AI leaders and builders at growing companies.
China loved OpenClaw so much they banned it. Earlier this week, viral images showed crowds lining up across Chinese cities to install the self-hosted autonomous AI agent. Local governments in Shenzhen and Wuxi even handed out subsidies, turning China into OpenClaw’s biggest market overnight. The excitement lasted just days before Beijing abruptly pulled the plug, urgently warning government agencies and state-owned enterprises about huge cybersecurity risks. Now the same companies that rushed to install OpenClaw are cashing in by removing it just as quickly. Maybe giving sensitive data to an AI notorious for security holes wasn’t China’s smartest move. [Fast Company]
NVIDIA’s new Nemotron 3 Super redefines what’s possible with agentic AI. Until now, autonomous agents stumbled on complex tasks, losing track of goals or wasting resources on simple decisions. Nemotron 3 solves both problems, running tasks 5x faster and twice as accurately as previous versions. With a 1-million-token context window, agents can hold an entire codebase, or nearly 750k words in memory before collapsing. It’s the model powering Perplexity’s new “Computer.” Want to run it yourself? You’ll need 8 NVIDIA H100 GPUs, or better yet, NVIDIA’s latest Blackwell hardware. That’ll cost around $250,000 upfront or $22k monthly. Nemotron is open source, but realistically only companies with deep pockets can play. Wake me when I can run it on my MacBook. [Nvidia]
fun stats
🙌 7% to 10%. The “sweet spot” for using AI tools at work to maximize productivity. Only 3% hit it – most people spend less than 1% of their time using AI.
🦄 $32 billion. What Google just paid in cash for cybersecurity unicorn Wiz, its biggest acquisition ever. The deal helps Google strengthen cloud security in the AI race.
🔨 $26 billion. How much Nvidia is spending to build open source AI models to compete with OpenAI, Anthropic, and DeepSeek.