Tom MacWrightreviewing software engineer portfolios that are (1) vibecoded website (2) github projects that are 100% edited by or with claude
i don't know what i'm supposed to get from this
Dan MacTough@tmcw Theoretically, I would be trying to demonstrate that I am taking advantage of the latest tools to both experiment quickly and build with exhausting rigor. More modestly, maybe I would just be trying to show that I'm exploring the "cutting edge." In your position, I would spend a little effort trying to figure out if that's what I'm looking at, but if it really just looks like nothing but lazy slop, well... that's who they are
@danmactough to steelman the first idea, how am I supposed to see rigor from a github repo where every commit message is llm-generated and there are no issues? how can i differentiate between someone doing a good job intelligently prompting or harnessing or whatever the fuck it is this week (sorry) versus someone telling the llm to be really rigorous?
@tmcw Well, if the premise is that there's no way to tell if the person is demonstrating rigor (commit msgs are slop and no issues), then you're kind of begging the question of how to evaluate it. But "all in" on AI _can_ bear signs of rigor. What if there were decent issues that the llm used? What if the repos demonstrated a variety of ways to try to get the llm to build the same thing? What if there were a repo for a project for a novel tool that addressed a pain point in llm-assisted dev?