The same effect can be observed if you've ever been a software developer where you're told what solution to build without any of the context of the problem you're solving. "We need an FTP server, quick, ops, get on that." leading into "Oh, it turns out the customer didn't need that to receive our emails" leading to a bunch of very puzzled devops.
Funny, I always try to lead with context when handing off product/design work, but I've often experienced devs saying 'please just tell me what to do'. No such issue with AI coding tho, it works great there.
From what I've heard and seen , "Please just tell me what to do" is the attitude that developers pick up in dysfunctional cultures, usually large companies with incredibly many software developers producing incredibly little and low quality software.
don’t lead with context. lead with the problem, then the possible solution(s), then add context but only relevant context.
when people start by talking about seemingly irrelevant context stuff, like what jane from accounting said about the Foobinator page last week and how that meant there was meeting yesterday about … …, it just annoys devs cos it seems to us that the person is talking about completely irrelevant information / random business crap.
if you’re doing something like the above, it would make sense they’re cutting you off and trying to get to the point.
or they’ve just stopped caring but want their paycheck… ymmv
Yea the context I usually lead with is the problem and the intended or likely trajectory of the feature, not the noise like you say. I think different folks just like different things.
Yeah lol. That definitely does not sound like philosophy. Giving a "why" you want to implement a feature and make particular changes will help the AI stay on track much better than if it is driving blind. It can't make choices without understanding what the desired outcome is.
I would say it is definitely a form of context, but when people think of LLM context windows in terms of coding is more technical context related: "what has been done before, what's the coding task at hand." etc.
However, I think that there is a philosophical portion to that context as well: "What problem is this feature supposed to help with? How would you verify that passing unit tests means that the code is working as intended? Does this feature need to exist at all?" LLMs usually need these to be provided to them explicitly since they are not good at inferring the correct intent compared to humans, otherwise they just make something that looks right but doesn't work right.
