On where token pricing lands as inference costs keep falling. Worth sitting with alongside everything we are building on per-token billing.
Reading
Things I've read that stuck with me: articles, posts, and talks worth sharing, along with a thought or two of my own.
tokens too cheap to meter jyn.devAccelerating GPT-5.6 Sol Ultrafast with OpenAI CerebrasI keep thinking about how little of today's software UI survives if inference becomes practically instant. Loading states, waiting screens, even the search box all assume the answer takes a second to arrive, and once it doesn't, core flows feel closer to direct manipulation.
AI is removing the middle class of software engineering florianherrengt.comThis is going to be a real problem. I've already watched smart engineers get trapped in an AI-slop feedback loop.
Gateway API v1.6: TCPRoute and UDPRoute Graduate to Standard Kubernetes BlogGlad to see this finally reach GA. I first ran into Gateway API years ago while implementing an early alpha version in our Kubernetes cluster.