Writing
Notes on AI, characters, and work systems from the point where implementation meets operations. This is writing about what is being built now, not a warehouse of old posts.
Build agents. Speed up how outside inquiries and requests come in, and how estimates go out. As AI makes individual ability matter more, I'll keep removing bottlenecks. Communication is part of that work.
Read articleHow would I build the new decision model Jev into my own criteria? An AI host kept asking, and I got stuck. I got stuck because the question did not match how we actually use models.
Read articleIn 2008 I prototyped a memo tool that ran on both PC and mobile, and one question stayed open: which device supports thinking? Eighteen years later the answer was not about devices at all.
Read articleComparing Trac and XPlanner in 2007 led to a simple conclusion: adding hierarchy to tasks does not add execution. That judgment still shapes how an AI planning layer should be built.
Read articleExecuted, sent, and confirmed are different states. The more AI enters work, the more evidence and human handoffs become part of the product.
Read articleA character should be more than an image or a profile. It can become an interface between people and services.
Read articlePutting AI in a chat window does not change work by itself. Requests, decisions, ownership, and history need a system.
Read articleDo not underestimate AI. Do not pretend a demo or a strong claim has already proved value either.
Read article