0xSero looks back on a year of Claude Code usage, discussing how changes in models affected the technical and administrative tasks he delegates. His usage figures and model comparisons describe his own experience rather than a controlled benchmark.
He explains metric-driven research loops that make and evaluate small changesAn AI feedback loop occurs when AI outputs or their consequences become inputs that influence later model or agent behavior., and demonstrates a manager coordinating sub-agentsA multi-agent system contains multiple AI agents that interact, coordinate, divide work, or influence one another while pursuing tasks. working on inference performanceInference is the use of a trained model to produce an output from supplied input.. He also shows an example of gathering public-library information and turning it into a table and map.
0xSero discusses iterating on designs in an interactive document and writing precise reusable skillsContext engineering designs how relevant instructions, evidence, state and tool information are assembled for a model's task.. He cautions that stale or unrequested skills can affect behavior and describes reviewing and rebuilding them for recurring tasks.
Watch on YouTube
