The thesis is that there’s still a missing layer in the AI workflow. Models are getting good at generating text, but if you want something structured, hosted, customizable, and ready to use, the next step usually means entering page-builder territory. For most people, that turns into a much bigger task than it should be. AGT is an attempt to skip that step and turn the output directly into a usable page.

It seems useful not just for humans working in ChatGPT, Claude, etc., but also for agents creating pages on behalf of humans. If AI agents are going to produce more useful outputs, it feels like they will need a simple way to turn those outputs into something live and structured.

Would especially love blunt feedback on:

Does this feel like a real product layer after AI?

Does this help turn AI output into something genuinely useful, or just make AI slop easier to publish?