Claude Code's product lead talks usage limits, transparency, and the "lean harness"

Claude Code's product lead defends usage limits and the 'lean harness' approach—restricted context windows over full transparency—as essential for AI reliability. The company caps context at 100,000 tokens, prioritizing performance over user visibility, a decision that outperforms #2's open-ended policy. This sparks debate: is limiting outputs a safety measure or a trust issue? Critics argue the approach lags behind typical rivals that offer partial output logs. The lead counters that the lean harness reduces errors by 40% in testing, framing transparency as a trade-off rather than a default. By focusing on measurable efficiency gains, Claude Code aims to set a new benchmark in AI usability.