Three stories a day, from someone who builds with this stuff

Hundreds of AI announcements land every week. Almost none of them change what you should do tomorrow. This site exists to name the three that do.

The story

GenAI Under the Hood started as a YouTube channel about what actually happens inside language models: how they run, where they break, and what they cost. The daily brief grew out of the notes I was already keeping to decide what was worth a video.

I'm Curtis. I've spent nineteen years in technology, most of it in education and the public sector, and I currently design AI architectures for large public-sector organizations on AWS. I'm also finishing a PhD in computer science on prompt engineering and model output quality, I hold nine AWS certifications, and I speak regularly at re:Invent, re:Inforce and AWS Summits. The brief is the filter I use myself, published.

What I believe

Three beats thirty. A ranked shortlist from someone who has read everything is worth more than a feed of everything.

Say what it changes. A headline without a consequence is trivia. Every item here gets one line on what it means for the person reading it.

Reader first. Sponsors are labeled, affiliate links never change what makes the cut, and the brief stays free.

The two sites

This one is the news. runllmslocally.com is the lab: benchmarks of open models on consumer GPUs, plus a hands-on series on AI security. When a story here has a "run it locally?" tag, that's where it points.