video1.0<iframe src="https://www.loom.com/embed/12e0c5124c684f8cafeee6bd290788ca" frameborder="0" width="1658" height="1243" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12431658Loomhttps://www.loom.com12431658https://cdn.loom.com/sessions/thumbnails/12e0c5124c684f8cafeee6bd290788ca-04b70b997f50cf7a.gif198.677This Loom explains a Fairtrade-style approach to long-running, unreliable context in an AI-assisted gardening system. The plan is an explicit loop of about 600 words that only sees new updates, rewrites with reasons for every change, and archives the rest while staying current. It uses a live location in Lancashire County plus weather reports and alert feeds for local pests to wake the gardener immediately, running a Liquid AI LFM 2.5 model on a laptop. Logging includes timestamps and records actions such as covering peppers with a row cover and harvesting ripe tomatoes from bed 6, with everything retained in “rotary memory” for later troubleshooting like identifying why carrots failed due to sandy soil that drained too quickly.