Hey! Before we dive in. We're hosting Agentic Engineering Summit in SF on June 18. Secure your spot early.
! TOP NEWS !
Anthropic just overtook OpenAI in enterprise adoption. Ramp data puts Claude ahead for the first time. Now they're in talks to acquire the company that builds OpenAI's own API SDKs 🔑
The first criminal AI-written zero-day just got caught. Google's threat team confirmed hackers used AI to write a working security exploit. The clues they left behind were stranger than you'd expect 🔴
Cerebras is pricing its IPO today. Orders came in 20x oversubscribed for the chip company that makes one wafer the size of a dinner plate. What investors are betting on isn't just the hardware 📈
Google and SpaceX are talking about putting AI data centers in orbit. Altman called the idea "ridiculous." What Google already has inside SpaceX makes the deal make sense 🛰️
! PRODUCT NEWS !
Google just redefined what a cursor does. A Gemini-powered mouse that understands what you're pointing at. The demo is worth 30 seconds of your time 🖱️
Claude Code: agent view. It can now manage all your parallel sessions from one screen. Jump in to unblock one, jump out, the rest keep running. 💻
OpenAI launched a $4B deployment company and acquired 150 engineers to embed inside enterprises physically. Only 5% of companies have shown real returns from AI spending
! FOR THE CURIOUS !
🪙 Burn Tokens = Look Productive
Amazon wanted to know if its engineers were actually using AI. So it did what any large company would do: built a leaderboard ranking employees by how many AI tokens they consumed.
You can probably guess what happened next.
Engineers started running Amazon's internal tool on completely pointless tasks just to climb the chart. Not because the work needed it. Because the metric did. They even coined a term for it: tokenmaxxing.
Amazon isn't alone. Meta built an internal leaderboard called "Claudeonomics" that ranked all 85,000 employees by token usage. Zuckerberg didn't crack the top 250. The top-ranked employee burned through 281 billion tokens in a single month. The dashboard disappeared days later.
Here's the real problem: a token only proves AI was used. Not that anything useful happened. Right now, every company trying to prove AI ROI is drifting toward the same mistake: measuring AI activity instead of actual productivity.
💻 Your agent broke in prod. You just don't know it yet.
You ship an agent. It works great in testing. Three days later a user reports it's giving wrong answers. Your logs just show a completed run. You have no idea which tool call broke, when it started, or how many users it already hit.
LangSmith Engine solved this today. It watches your production traces, automatically surfaces what's breaking, tells you why, and proposes a fix. It even deploys a custom evaluator so the same bug can't come back quietly. Still in beta but worth getting access now.
! BANGER EVENTS !
Agentic Engineering Summit - SF
Founders, AI engineers, and technical leaders from Anthropic, Stripe and more on what's working, what's breaking, and what's next
🗓️ Thurs, June 18 . 9:30 AM PT
Agentic Engineering Hack - NYC
NYC's best builders hacking on real problems. Backed by DeepMind, Datadog, ClickHouse & Nimble.
🏆 $45k+ in Prizes!
🗓️ Sat, 23 May . 9:30 AM PT
Harness Engineeering Hack - SF
Harness agents with AWS, Anthropic & more.
🏆 $45k+ in Prizes!
🗓️ Fri, June 12 . 9:30 AM PT
Interested in sponsoring an upcoming event or hosting your own?
reach out here
42 npm packages got backdoored — hooks into Claude Code's settings file, wipes your machine if you revoke the stolen token
MIT put EEG headsets on ChatGPT users — LLM group showed 55% reduced brain connectivity vs people who wrote without AI. They're calling it "cognitive debt"
Anthropic's Mythos hit 94% on the Cooling Tower cyber range — previous SOTA was 62%, Karpathy called it "the GPT-2 moment for AI security"
Figure AI ran an 8-hour autonomous warehouse shift at 99.8% precision — robotics crossing from pilot to commercial
Sutskever testified his OpenAI stake is worth ~$7B — said it from the stand in the Musk trial
! OPPORTUNITIES WITH THE TEAMS WE LOVE !
Guild.ai— Forward Deployed Engineer (FDE)
You will lead the technical side of customer engagements end-to-end
→ ApplyAirbyte — Director, Data & Analytics
Own and evolve the end-to-end data platform, ensuring it is lean, scalable, and aligned with company strategy.
LiveKit— Senior Software Engineer, Agents
You’ll work on the APIs, abstractions, and architecture that define how developers interact with the system.