Good morning. Everyone loves open source software because it’s transparent, free to use, and collaborative. But how would you feel if someone cloned and resold a project as their own? Malus.sh recreates open source projects from scratch, minus those pesky license obligations. They say it’s satire meant to expose licensing loopholes… but with 847,293 projects copied, they’re clearly laughing all the way to the bank.
AI agents are shifting to outcome based pricing. Most AI tools charge customers even if they don’t deliver results. Adobe says that’s got to stop. They’re shifting to charging only when their AI agents actually get stuff done. Salesforce, Workday, and startups like Sierra already do something similar. It’s a smart move because it completely de-risks the sales process and forces AI companies to build products that really work. After all, if your AI isn’t delivering, neither is your business. (The Information)
SpaceX inks deal to buy Cursor for $60 billion. Everyone knows Grok can’t code, and xAI won’t beat OpenAI or Anthropic without a serious upgrade. Musk’s trillion dollar SpaceX IPO dream hinges entirely on having an AI that’s smart enough to launch into space. Right now he’s not even close. That’s why Musk’s betting everything on Cursor’s coding tech, and even throwing in a bizarre $10 billion “thanks anyway” fee if he backs out. Kinda ironic that the one guy capable of launching a Von Neumann probe doesn’t have an AI that can upgrade itself. (The Verge)
Templafy’s Document Agents follow your brand and compliance rules, turning AI from a risk into a reliable teammate. No rogue slides. Every presentation is accurate, polished, and ready for the outside world – automatically.
- Trusted by KPMG, Adobe, CISCO (and 4M+ users)
- 30% of document time saved
- Works inside PowerPoint, Google Workspace, and Salesforce
Anthropic’s Pentagon ban might be ending soon. Anthropic landed on the Pentagon blacklist after refusing to let Uncle Sam use its AI models for spying on citizens or autonomous weapons. The feds labeled Anthropic a security risk, pressuring them to fall in line. CEO Dario Amodei fought back in court and recently met with White House officials to resolve the standoff. Trump now says Anthropic is “shaping up,” conveniently just as the Pentagon needs Mythos, Anthropic’s powerful cybersecurity AI. Getting banned gave Anthropic serious street cred. Let’s see if getting unbanned ruins the vibe. (Reuters)
ChatGPT Images now has internet access. OpenAI upgraded its image tool with “thinking capabilities,” letting it search the web in real time and serve up fresh visuals. Images 2.0 can finally spell, meaning no more menus offering “margartas.” It also generates multiple consistent images from a single prompt for seamless campaigns. Just what the internet needed, unlimited stock photos of smiling millennials eating perfectly spelled avocado toast. (ChatGPT)
XPONENTIAL | MAY 11-14 | Detroit, MI
Once you see the signals shaping autonomy, act on them. At XPONENTIAL 2026, connect with the founders, investors, and operators turning those signals into deals, partnerships, and deployed systems across every domain.
Meta is harvesting employee keystrokes for AI. Employees just found out that software installed on their computers tracks mouse movements, clicks, keystrokes – even screenshots. It’s all to train AI agents to handle everyday tasks like navigating dropdown menus and keyboard shortcuts. Meta swears none of this will affect performance reviews or be used beyond training AI models. Still, turning your entire workday into an AI surveillance project feels slightly dystopian. At least they’re not collecting faces and voices like XAI. (The Register)
AI agents are learning to rewrite themselves. Researchers built a system that lets agents upgrade their instructions and tools in real time. If an instruction falls short, the agent rewrites it. If a tool is too slow, the agent improves it. Useful changes stick. So instead of repeating the same steps each time, agents improve after deployment. On tough reasoning tasks, accuracy jumped from about 20% to 33%. On messy real world tasks, it climbed from 79 to 89%. Companies are now closer than ever to shipping software that evolves while it works. Have fun explaining that to the security team. (arXiv)
fun stats
🕹️ 22,000. Front line missions by Ukraine’s ground robots in the past 3 months. President Zelensky plans to swap out a third of infantry troops with robots and drones this year because “robots don’t bleed.”
📈 $1.4 billion. What Elon Musk spent buying SpaceX stock from employees last year to lock in control before the IPO. He stands to pocket another 60 million shares if he keeps investors happy.
🪝 $3 to $5. Bid range for OpenAI’s new cost-per-click ads in ChatGPT, about 3x cheaper than Google’s search ads.