llm 0.31

The verbosity parameter is the detail worth pausing on: newer OpenAI models have begun returning structured or verbose responses by default, and without a control surface like this, CLI-based workflows can break or become noisy in ways that are hard to pipe downstream. Willison is essentially patching around a behavioral shift in the models themselves, not just adding a feature.

This update is largely disconnected from the funding, hardware, and policy stories that have dominated recent Modelwire coverage. The closest thread is the broader developer tooling wave: the App Store piece from TechCrunch in mid-April noted that AI is lowering barriers to software creation, and Willison's llm tool sits squarely in that current, giving non-framework developers a scriptable path to the latest models without standing up a full SDK integration. GPT-5.5 support arriving this quickly in a third-party CLI also signals how fast OpenAI's model surface is expanding, which creates real maintenance pressure for open-source wrapper projects like this one.

Watch whether Willison adds similar verbosity controls for non-OpenAI providers in the next one or two releases. If he does, it suggests the verbose-by-default behavior is spreading across model families; if it stays OpenAI-specific, it's a quirk of GPT-5.5's response formatting rather than an industry-wide shift.