Cut Over the Model Path or Don't Ship: A Fail-Closed Inference Checklist

Last time I wrote about a Terminal-watching daemon that auto-clicks Computer Use approval dialogs....

I started Strokeline with one dumb, simple idea: What if you could describe an animation the same...

Persistent memory turns an agent from a re-explained tool into one that carries context forward — but...

Free capacity is cheap only when nobody is waiting on it. The moment a job sits in a shared queue,...

What Google Summer of Code, LFX, and Symmetry Autumn of Code taught me, and how that habit turned into building my own AI infrastructure tools.

A backend engineer opened a pull request at sixteen forty after an afternoon spent steering a coding...

I’ve been experimenting with AI integration in software applications for some time, not as an...

You type a sentence into ChatGPT and it looks like normal English. The model doesn't see it that...

My last two posts gave rates with a self-selected denominator. This one fixes that: a generation-sliced sample of active repos (6.2%) and a uniform sample of the whole repo population (1.0%), plus the ecosystem numbers that surprised me.

I got the rejection email on a Tuesday. I've been rejected before — everyone has. This one broke...