Welcome back. In today’s “sci fi is now real” news, researchers say we finally have enough computer power to simulate a human brain. They’re firing up one of the world’s fastest supercomputers to model billions of neurons and synapses, equal to the cerebral cortex. The hope is to get new insights into memory, learning, and why we hit snooze every morning. Surely this will answer everything.
AI is killing scientific curiosity one paper at a time. Researchers who rely on AI crank out papers faster, score more citations, and speed through academic ranks. But there’s a cost. AI-driven research sticks to predictable, safe topics, steering clear of messy, unexplored territory. Sociologist James Evans says AI is quietly shrinking science’s intellectual diversity, triggering a feedback loop where scientists chase easy wins instead of real discoveries. Academic incentives reward speed and quantity, not originality, turning cutting edge research into a race toward conformity. (Spectrum IEEE)
Musk turned on the world’s first gigawatt AI data center, and immediately got busted for breaking environmental rules. xAI powered its massive Colossus 2 facility using portable methane gas turbines, dodging pollution controls and grid limitations. These turbines pump out pollutants linked to asthma and cancer, impacting nearby communities. Now the EPA has slammed that loophole shut, clarifying that even AI data centers must comply with the Clean Air Act. Colossus is stuck in limbo until Musk finds a legal way to keep the lights on. Moving fast is easy when you ignore the rules. (The Guardian)
Most companies get stuck tinkering with prompts and wonder why their agents fail to deliver reliable results. This guide from You.com breaks down the evolution of agent management, revealing the 5 stages for building a successful Al agent and why most organizations haven’t gotten there yet.
You’ll learn:
- Why prompts alone aren’t enough and how context and metadata unlock reliable agent automation
- 4 essential ways to calculate ROI, plus when and how to use each metric
- Real-world challenges at each stage of agent management and how to avoid them
If you’re ready to go beyond the prompt, this is the playbook for you.
Sequoia just blew up VC’s biggest unwritten rule. Investors usually steer clear of backing direct competitors, preferring to pick one winner. Sequoia already has major stakes in OpenAI and xAI. And OpenAI warned investors they’d lose insider access if they put money into rivals. But Sequoia’s diving headfirst into Anthropic’s massive $25 billion round anyway. Apparently, losing access to OpenAI’s secrets isn’t scary enough to walk away from Anthropic’s IPO payday. (TechCrunch)
Side note: This weekend I finally got why desktop AI agents matter. I spent days syncing our CRM to BigQuery, bouncing endlessly between browser tabs, terminal windows, database queries, and AI conversations. My goal was to have Gemini analyze sponsors, subscriber engagement, and metrics for this newsletter. After the 100th tab switch, something clicked. Suddenly tools like Anthropic’s Claude Cowork made sense. At first I didn’t get the hype, but after drowning in manual tasks, the potential was obvious. As a former CISO, I’m still hesitant to hand over control of my desktop to an agent. But after this weekend, I’m seriously considering reclaiming some of my precious days off. (Adam Wildheart)
November’s election could depend on how much politicians pay your AI. Political ad buyers are starting to shift spending from traditional ads toward influencing AI agents. Experts predict politicians will soon funnel billions into algorithms that subtly shape voter opinions. It’s currently unregulated, so voters might never realize their personal AI assistant has become a paid political influencer. While it’s hard to believe frontier AI’s like Grok would get behind this, the shift is clear. TV ads are out, and elections as a subscription service are in. (MediaPost)
Forget tea, Europe’s throwing Silicon Valley’s AI into the harbor. With US-Europe tensions high, relying on American tech looks less like convenience and more like geopolitical suicide. Belgium’s cybersecurity chief warned they already “lost the internet” to America and fears AI is next. Inspired by China’s DeepSeek, European labs are racing to build their own AI without big GPUs, hoping openness and collaboration can help close the gap. Pushing your allies into becoming competitors probably isn’t the smartest strategy for AI world domination. (Wired)
fun stats
🪦 5. Staff who quit Mira Murati’s AI startup Thinking Machines Lab last week. The exodus included two co founders, with 4 out of 5 joining OpenAI.
🧰 $1.7 billion. Additional capital VC firm Andreessen Horowitz (a16z) is committing to AI infrastructure projects. Sharpen your pitch decks: they’re specifically targeting AI back-end and developer tool startups.
🧵 141.5 million. Daily mobile users on Meta’s Threads. That’s more than Elon Musk’s X gets, but Threads still lags for web visits.
👛 79%. IBM surveyed execs who think AI will significantly contribute to their revenue by 2030, up from 40% today.