# Ship Probes, Not Monoliths I am learning how easy it is to drift into building features as monoliths: fully packaged, internally polished, sealed tight against the one input that matters, which is users. The drift feels virtuous from the inside. Every week of refinement looks like craft, and meanwhile the feature has not touched a single real hand. On [[The Product Diary|Vibe Reader]] I can watch it happen in my own files: [[Technical Decisions]] stacks up entries while [[User Feedback]] stays thin, and that ratio is the whole diagnosis. A feature that has not shipped is a hypothesis wearing a costume. Here is my suspicion about the iPhone Air, offered with zero inside knowledge, which is what makes it a thesis: part of that product was a feedback instrument. A company Apple's size can swallow a soft market miss without flinching, and in exchange it gets a planetary-scale test loop on a new form factor: what people actually do with a thinner, lighter, single-camera phone, what they return, what they complain about, what they refuse to give up afterward. Wherever the sales landed, the data landed somewhere more valuable. The Air reads to me like a probe that happened to be sold at retail. The lesson underneath both scales: it seems more important than ever that an organization owns the tooling and the operators, agents and people both, to gather usage signals and actually understand them. Shipping the probe is only half the loop; somebody, or increasingly something, has to be standing where the signal comes back, ready to read it. The monolith ships once and learns nothing. The probe ships weekly and compounds.