_____________
User Story
As a reader logged in to Hacker News on a locally running web browser, I want a process running on my device that polls for "Show HN" posts and automatically replies to them — as me — with the results of an LLM-analyzed critique of the posted projects discovered by the process.
Acceptance Criteria
• A brutally frank critique of the posted "Show HN" project is given by a state-of-the-art LLM
• A count of existing projects functionally identical to the post being critiqued is displayed
• A list of authoritative learning resources on whatever the LLM determined the author is probably trying to accomplish is provided
• …???…
_____________
FWIW: Even with the well-documented initial inertia-reducing powers of today's coding agents, it's super, super unlikely that I'll ever get around to implementing this idea myself.
I'd be totally cool with somebody else taking a swing at it, though.
As to feedback:
> A count of existing projects functionally identical to the post being critiqued is displayed.
I (and probably others like me) have little bespoke programs that do something similar to this. Not to the extent of physically going to the external site though, it's more around looking at semantic similarity.
> …LLM agents posing as
> people on HN…
I suppose I could have made it clearer that the reply would explicitly make readers aware that the analysis was done by a Robo — i.e., an LLM.