! TOP NEWS !
Meta is recording employee keystrokes and screenshots to train AI. There's no option to opt out, 8,000 employees set to exit next month are getting logged first š
SpaceX can buy Cursor for $60B or pay $10B to walk away. Cursor gets access to xAI's Colossus supercomputer. A year ago it was valued at $9.9B š
LeCun told Dario Amodei to stop making predictions about job losses. The full exchange got personal š„
Vercel got breached through a third-party AI tool. Customer API keys were accessed. If you deploy there, rotate your env vars now š
Unauthorized users breached Anthropic's Claude Mythos on launch day. Raising questions about control of the AI model.
! PRODUCT NEWS !
Moonshot open-sourced Kimi K2.6, a 1T-parameter coding model. How it stacks up against Opus and GPT-5.4 is what's getting attention š¤
OpenAI just launched Workspace Agents. Persistent AI (cute) teammates that connect to your Slack, Linear, and Drive with write access. They don't wait to be asked šØāš»
OpenAI also shipped GPT-Image-2. It crushed the Arena leaderboard with a 1,512 ELO score The gap over the next best model is wild šØ
! FOR THE CURIOUS !
šø You're the frog. AI is the water
Researchers from UCLA, MIT, Oxford, and Carnegie Mellon gave 1,222 people an AI assistant, then pulled the plug mid-test. Their performance dropped below the people who never used AI at all.
They're calling it the "boiling frog" effect. Like a frog in slowly heating water, the cognitive dependency sets in so gradually you don't notice until it's too late. Each autocompleted function, each drafted email, each summarized doc feels costless to you but your brain is quietly outsourcing the hard parts and forgetting how to do them. And it only took 10 minutes of AI use.
So if you're reading this newsletter with your favorite AI open in the next tab... maybe close it for a sec. Just to check.
š”ļø AI-generated code is fast. It's also full of holes.
Replit just shipped a Security Agent that sits between your AI-generated code and production. It combines static analysis with AI reasoning to catch the stuff that both LLMs and traditional scanners miss on their own.
If you're auto-generating production code with Cursor, Claude Code, or Codex, bolting on "AI-only" security isn't cutting it anymore. This is the hybrid approach that's been missing.
! BANGER EVENTS !
Ship to Prod Hackathon - SF
Build real AI agents with AWS, Wundergraph & more.
š $45k+ in Prizes!
šļø Fri, 24 April . 9:30 AM PT
AI Agents in Prod - Ship Demo Day - SF
Showcase your project or attend workshops. Backed by Google Deepmind, Eleven Labs & more.
šļø Thurs, 23 April . 9:30 AM PT
tokens& Founder Dinner- SF
An exclusive dinner night with some of the coolest founders/investors. Limited spots
šļø Friday, April 24 . 8:00 PM PT
Weighted Moves: AI & Chess Mixer- SF
Chess meets decentralized AI. A night at the oldest chess club in the US, with AI-paired wine and networking
šļø Thurs, April 30 . 7:00 PM PT
Interested in sponsoring an upcoming event or hosting your own?
reach out here
OpenAI turned on cost-per-click ads inside ChatGPT ā $3-5 per click
Amazon is pouring another $5B into Anthropic ā $100B AWS commitment over five years
Claude Code's new /ultrareview sends a fleet of bug-hunting agents before you merge ā Pro and Max get 3 free through 5/5
Google says 75% of its new code is now AI-generated ā up from 50% last fall
! OPPORTUNITIES WITH THE TEAMS WE LOVE !
Bland AI ā Technical Account Manager (SF)
Own the post-sales journey, ensuring customers launch quickly and achieve measurable business results with AI phone agents.
ā ApplyRedis ā Developer Advocate
Create outstanding technical content that encourages developers to build with Redis, engage with the developer community, and advocate for the best developer experience.
ā ApplyBland AI ā Forward Deployed Engineer (SF)
Build and deploy AI voice agents directly with enterprise customers, from first call to production.
ā Apply