Discussion
Oh bugger, got my first Claude co-authored PR. Telltale signs were a gigantic wall of text instead of a succinct description, and massive commit messages. Changes look far too big too.
I’m going to see if I can extract the intent of it and re-do it myself in a saner way, and I’m going to have to go through all my projects and put a big “No AI” message on them all. 😩
Right, that's got LLM declarations and contribution guidelines pushed to the 5 most common repos I get contributions to (summary: nah mate). Thanks to folks who already published theirs as examples to draw on! e.g. @dougbinks @floooh @Doomed_Daniel
I reviewed the underlying issue "addressed" in that LLM PR, turns out it was completely unnecessary and down to a misunderstanding. Who knows how many tokens were used to address a problem that doesn't exist if you use the API as intended. They tried to use the wrong tool for the job, when a better one was available, and when it didn't work they asked Claude to write hundreds of lines of code to 'fix' it. If they'd actually asked me (a human), I could have given them the correct answer 🤦♂️
I mentioned earlier that I'm getting fewer support questions these days, and I wonder how much of this is down to people 'fixing' things with their LLMs like this. The trouble is that the LLM only does what you ask, even if you ask the wrong question. I wonder how many ill-conceived patches are being applied to codebases, band-aid upon band-aid, just because someone didn't understand something properly and didn't ask
Install Fedizen
Get the full app experience
I really don’t know why people who use Claude think these massive LLM “brain” dumps are useful. Maybe it makes them feel clever for the sheer number of words spaffed onto the page. But it’s always, always been the case that actually smart people can sum up a complex thing in a small, digestible form. Dumping the text book onto the page has always been the way of the amateur.