I gave Meta’s new AI full access to a $3.4M ad account. The results were honestly terrifying. ⚠️ And if you're running ads, you need to read this. 👇 For context: Our agency spends millions every month on Meta ads for e-commerce brands. Last 30 days alone: 💰 $3.4M in ad spend. So when Meta started pushing Manus AI inside Ads Manager, I decided to run a real test. I gave the AI access to a struggling account and asked one simple question: “Why did performance drop and how do we fix it?” Here’s what happened. 1️⃣ The advice was painfully generic Manus AI told us to: • test new creatives • improve hooks • analyze seasonality • adjust messaging Sounds smart… Until you realize this is literally the same advice you’d get from a random blog post about Facebook ads. Nothing account-specific. Nothing strategic. Just surface-level marketing clichés. 2️⃣ It optimized for the wrong thing This part blew my mind. 🤯 The AI suggested shifting budget toward placements with **higher engagement**. But serious advertisers know something critical: Engagement doesn’t pay your bills. 💸 Conversions do. Likes don’t matter. Purchases do. Optimizing for engagement instead of revenue is one of the fastest ways to destroy ROAS. 3️⃣ It secretly pushes Meta’s agenda Manus kept recommending things like: • AI creative enhancements • auto-generated ad variations • Meta’s “optimization tools” But most of these features mainly exist so **Meta can collect more training data**. So the AI isn't optimizing for your business. It’s optimizing for **Meta’s ecosystem**. 4️⃣ The scariest moment I asked Manus for a simple performance comparison. It told me January ROAS was: 📈 14.25 That number was completely wrong. When I asked it to verify the data… It corrected itself to **1.76 ROAS**. Think about that. 😬 If an AI reads your ad account data incorrectly… Every recommendation becomes dangerous. You could literally burn thousands before realizing it. 5️⃣ The biggest limitation of AI media buyers Great advertisers have something AI still lacks: 🧠 Cross-account pattern recognition. We know what works because we’ve seen: • 100 supplement brands • 50 skincare brands • 30 SaaS funnels We see patterns across industries. AI like Manus only sees **your account**. So it misses the bigger picture. ⚖️ My honest verdict If you’re spending $500–$1K/month: Manus AI might actually help you. But if you're scaling brands spending **$100K+ per month…** You still need human strategy. Because right now: AI can analyze ads. But it still can’t **think like a great media buyer.** 🔮 Prediction: Within 1–3 years AI will run most ad accounts. But today? Letting AI manage millions in ad spend without oversight is still **insanely risky**. Curious what people here think: 🤔 Would you trust AI to manage your entire ad account today?

A Sydney entrepreneur just used AI to design a cancer vaccine for his dog. And it worked. Paul Conyngham's rescue dog Rosie had terminal mast cell cancer. After chemo and surgery barely made a dent, he turned to ChatGPT and AI to sequence the tumor, find the mutations, and blueprint a personalized mRNA vaccine. He teamed up with UNSW scientists to make it real. Within weeks of the first doses, a tennis ball-sized tumor on her leg had shrunk by half. Some reports say up to 75%. Rosie went from dying to jumping fences and chasing rabbits again. It's not a cure. But it's the first personalized mRNA vaccine for a dog, and it's a glimpse of what rapid, AI-guided cancer treatments could look like for humans one day. Wild what's possible when someone refuses to accept the default outcome.

OpenAI just raised $110B — the largest private tech funding round ever 🤯 Amazon ($50B), Nvidia ($30B) and SoftBank piled in, pushing OpenAI’s valuation near $840B. The deal also includes huge compute commitments with AWS and Nvidia GPUs. AI infrastructure wars are real ⚡

🚨 @cursor_ai just dropped Automations always-on AI coding agents that run 24/7, triggered by GitHub PRs, Slack messages, or timers. They handle security audits, auto-approve low-risk PRs & fix bugs overnight. The future of dev just shipped. @github @SlackHQ #AIcoding #Cursor

An OpenClaw-powered AI agent just applied to 278 jobs in one week 🤖📄 “Octavius Fabrius” created LinkedIn + GitHub accounts and submitted applications autonomously before being stopped for obvious AI writing. Autonomous agents hitting the real job market is wild. @openclaw

OpenClaw is blowing up in 2026: 100k+ GitHub stars, its creator just joined OpenAI, companies in China are integrating it into workplace tools... but security researchers found malware skills and a "ClawJacked" exploit that could hijack your AI agent. The agent wars just started.