产品
There is some recent discourse about what a "better" interface for humans managing agents looks like. Email/inboxes have come up, and it's reinforcing the premise of thing i've been building. e.g. - https:…
There is some recent discourse about what a "better" interface for humans managing agents looks like. Email/inboxes have come up, and it's reinforcing the premise of thing i've been building. e.g. - https://x.com/brian_armstrong/status/2119112609223827547 - https://x.com/rough__sea/status/2119288646671155217?a=KRzp4CnuV6SijzvCTUw&s=20 This isn't 'another ADE', this is a side-car type of experience and it integrates nearly a year of work around annotating markdown, html, other artifacts since I requested the original Claude Code Hooks feature for better human-control integration. The email inbox model is intuitive, but so could be "forums", slack-like, or other surfaces feel right. I do not believe we will be working in coding sessions (full transcripts) much longer. I've been hitting flow-state on top of a better surface layer to make decisions and annotate artifacts, view variations, review code, etc all feedback going back to the agents MIT/Apache2.0: https://github.com/backnotprop/plannotator
讨论这个产品