I've been migrating my many cloud projects into a single virtual private server (VPS) in an effort to reduce costs. Just for fun, I've been using ChatGPT to, um, assist. It has often been useful but has also led me astray a number of times. It's all recorded in hours and hours of video here (don't watch it; it's excruciating). I mention this because while working for such an extended time with the AI I had exactly the same experience as documented in Mike Kentz' blog entry The AI Will Wait. He writes, "The AI will wait. It has no preferences. It experiences nothing resembling impatience. The urgency we feel is entirely generated by us." But also, because the AI forgets as you fill its context window, "AI work often becomes anamorphic: each sprint adds segments fast, but without a maintenance molt, older instars get left behind and disappear." I find I have to summarize the good bits and regenerate the chat window on a regular basis. That way, the AI isn't led astray by its own errors.
Today: Total: [Direct link] [Share]
