why this newsletter exists
for people building AI agents who care about getting them right
Most content on AI reliability, alignment, and safety lives at one of two extremes.
On one side: dense academic papers, formal mathematics, philosophical frameworks that assume a PhD. Rigorous, but written for people who don't ship. On the other: LinkedIn-style hot takes, "5 ways AI will change everything," breathless predictions with no substance.
There's a gap in the middle. For people building AI agents who want intellectual rigor without needing a degree. People who prefer practical frameworks but still want the thinking to hold up.
That's where I am.
I'm a founding engineer at a growth-stage AI startup, shipping agents and AI systems in production. The systems I build affect real people's decisions: their careers, their hiring, their livelihoods. According to the EU AI Act, some of them qualify as high-risk. Reliability, alignment, and safety aren't optional — they're the job.
This newsletter is where I work through that problem. Some posts will translate research into builder language. Some will be production lessons from running agents at scale. Some will be experiments testing these concepts in practice.
If you're building AI agents and trying to get them right, this is for you.
We're all working prototypes here.