Tokens para agentes: como a arquitetura reduz custos
Cut GPT-5.5 output costs at $30 per million tokens by fixing agent architecture and stopping redundant file reads across sessions.
Marcus Chen
14 min read