Michael Beckley

I try to avoid superlatives but today I have to make an exception because this is the single most insightful research piece yet to emerge from the Ukraine War from non-Ukrainian sources. Here we have data-driven policy insights that the West will continue to rationalize or ignore away for as long as Ukraine continues to win time on the battlefield for us to squander.

Pro-Tip: You don't need to guess whose self-serving AI predictions will come true tomorrow if your AI projects deliver value today.

Is your enterprise a Petri dish for Big AI to study AI safety? Are you turning your business and your clients into lab rats with vibe coding where? Who benefits from these experiments? Last month an Acquired Podcast interview with Boris Cherny, creator of Claude Code, didn't generate the attention it deserved. In it, Boris says his priority is researching AI Safety, no shock there, but it's his methods that might surprise you. He says he couldn't learn enough about AI Safety in the lab so they created vibe coding to let it out in the wild, making it just useful enough that people would, well, use it. You know, so he and his friends in Big AI could study AI Safety by watching you rely on a fundamentally unsafe technology for mission-critical work. So when your vibe coded agent ignores all your guardrails and deletes your production database, that's a feature, not a bug. You lose but don't worry, don't feel bad, at least Boris and Big AI learned one more real-world failure mode at your expense. Of course, being in the AI Control Plane business, I'm a little biased on this subject and perhaps even being a little unfair to Boris. He does say he thought vibe coding was "safe enough" to release into the wild. It's just I'm not sure his definition of safe has anything like the determinism your clients would reasonably expect it does. youtu.be/RkQQ7WEor7wsi=gFwlO…

A little surreal reading about Ukraine strikes triggering oil shortages in Russia this week while rereading a dog-eared copy of one of my favorite novels, Red Storm Rising. Actually a collaboration between two of my favorite 1980's prophets: Tom Clancy and Larry Bond, Red Storm Rising turns 40 years old next month and it's one of those rare novels that rewards rereading. Clancy imagines a devastating attack that wipes out 1/3 of Russia's oil capacity and a desperate Kremlin trying to divide NATO to launch a desperate gamble, a conventional assault on Europe before their regime collapses. That could NEVER happen....

SpaceX IPO is a referendum on Elon Musk, says Appian CEO youtu.be/hkRSNRcs_n4?si=gO3Z… via @YouTube

Putting ChatGPT 5 in context... techmonitor.ai/ai-and-automa… Are LLMs becoming a commodity? - Tech Monitor

PRC Grand Strategy high-capacity.com/p/chinese-…

Making AI boring one process at a time. enterprisetimes.co.uk/2025/0…

Appian CTO on Process Automation in an AI World - Bloomberg Tech Disruptors Podcast: ap.pn/40H74rs Also available on Apple: ap.pn/3Cvg7kV And Spotify: open.spotify.com/episode/2lg…

Feel the rainbow

I'm not a morning person but sometimes jet lag works in my favor.