SIDE QUESTS THAT TAUGHT ME SOMETHING
I have a long-running habit of trying new technology before the ergonomics, documentation or business case are fully ready.
DaemonCore is the current, much larger experiment: separate model intelligence from operational authority, make identity and context explicit, and retain evidence of what autonomous software was actually allowed to do.
Semantic Publishing Protocol explores publishing artefacts with machine-readable meaning, context, attribution, intent and consent instead of relying on an agent to reverse-engineer all of that from rendered pages.
The AI Starvation Loop models a feedback problem: agents consume without traffic, publisher incentives weaken, less material gets created, and the data ecosystem AI depends on becomes poorer.
Error handling, middleware, reusable UI patterns, dynamic metadata and the other awkward requirements that arrive once Power Platform stops being a demo and starts becoming software.
SharePoint Spaces prompted a more useful question than “is VR cool?”: which enterprise information becomes easier to understand when the interface is spatial, and which workloads become ridiculous when you put them in a headset?
Yes. Should I have? Also yes, because trying it exposed exactly how far immersive productivity still had to go. The browser worked, the development environment worked, and precision interaction was hard enough to make the whole exercise memorable.
The experiment is rarely “can I make the technology do this?” The useful question is “what do I learn by making it try?”