Exploring the friction points of building real-world AI systems. Notes on agentic workflows, infra evaluation, and reverse engineering.
Production failures in computer vision are rarely simple 'wrong predictions.' They are complex conceptual drifts—blur, lighting, camera shifts. Here’s how I built a 'flight recorder' to catch them before they become incidents.
December isn’t “dead.” It’s noisy, uneven, and full of timing signals. Here’s what I observed by tracking company hiring momentum daily—and how to prepare for January after the holidays.
I stopped tracking individual job listings and started tracking hiring behavior—momentum, freezes, job lifespan, and timing. Here’s what changed and why it’s actually useful.