Comment by aaaronic
17 hours ago
My "favorite" Claudism is when I critique its work and ask it to remove some unnecessary part of the design -- and then the diff has more green than red because it added comments about why the code is no longer there -- the code that was never in the mainline and never asked for!
Oh god this has been driving me nuts since Opus 5 landed. Every docblock is filled with long-winded jargon explaining why this design is superior to some other design, which never existed as far as any dev who might read that comment is concerned.
This has been a thing for long while, on codex too.
You ask it to do something, then tell it to do something in a different way, then it assumes it needs to do the refactor in a backward compatible way, or creates migrations for it etc.
This is why I started adding a PROJECT.md file to all my projects and a hook for claude to read it.
It contains (Among other things) stuff like "this is a single user personal project, I'm the only user, this will never be open to the public internet" etc.
It kinda-sorta tones down the proclivity to worry about backwards compatibility and slight edge cases where if someone has edited some template and the new code doesn't support it.
Omg the "backwards compatible fix" in a one-off script, I feel my ptsd coming on ...
Oh good I thought I was doing something wrong! Using ChatGPT web for planning, ask for a prompt then notice something weird in the prompt and whether I:
1) use the edit in-place functionality; or
2) ask it to rewrite to remove something
It'll write the prompt as if the agent (codex) knew about the conversation and add "don't do X" etc. At first that bothered but I realized it doesn't really change the output so I stopped caring.
Still, no experience is unique I guess.
I remember when I was updating some formats for my apps data files and it ended up writing v2 and laying it on top of the old one. Ended up just telling it to delete the entire feature and start again. I do think claude.md instructions help though.
Oh no, you said "landed"! It's dug into our brains!
I hate that I have to change the way I write to avoid AI-isms. I loved using "load bearing" to describe weird code that you think you can delete but is actually holding everything together.
Now people think I'm just parroting what Claude said. It sucks. I want my catch phrases back, I guess this is how em dash users felt
6 replies →
You could say it landed cleanly.
5 replies →
Oh. Oh no.
There needs to be a pithy name for this. I have been calling it context leak.
I like ‘accretive editing’: https://justindfuller.com/programming/accretive-editing
It's not just claude, all AI is unable to produce something concise. On the surface everything looks 'good' whether code or prose, but then if you dig a bit, try and understand the whole text you quickly realise that 80% of it is unecessary and the whole thing could have been re-worded/re-coded into something a fraction of its size and complexity.
I asked Sol to reduce the length of some documentation we had by making it more concise. It came back after 20 minutes of work, did a line count and was aghast that the line count had somehow increased...
It's not that it fundamentally isn't able to produce something concise, it's that the business model of the companies developing these models rests on selling tokens...
because human writing IS not just producing the Next token, sometimes you write shit and then go back and find ways to improve and try to play tricks inside the structures only you or your coauthors understand. AI doesnt do this
It’s not some conspiracy. If they knew how to make the AI responses better and more concise, they would. They just don’t yet.
I have a theory that AI code looks good because you never subsequently come up with your own alternative.
"AI code looks good for you because you are a bad developer"?
3 replies →
I have to ask Claude to compact the comments every time, and I give specific criteria for it. Never ever reiterate what’s in the code, never mention decisions not made, never mention the conversation, etc etc.
Even then it is conservative. For the love of God, compact the comments.
Comments become a huge maintenance burden, especially in the age of AI. They just grow and grow, and then mislead the AI later on.
I just wrote a utility to rip all comments out of the code. Now the code is fully uncommented and it has saved lots of input tokens and also lots of meandering because the model is no longer getting stuck on bad ideas it told itself about.
1 reply →
I set a line budget for comments (also wiki page parts, chat responses, etc). That only helps when I ask it to do a second pass to reword everything to the budget and add links. I think they tuned it this way to stash reasoning dumps in the code. Unlike human developers, it has no context in its head, other than general GitHub knowledge.
That is a dramatic shortcoming, but I fix it with permalinks to other files. https://replicated.live/blog/link
Oh Jesus this. I've tried to include rules that tell Claude to only include relevant, evergreen comments but it's to no avail.
I also love how it'll build local plans with phases, tasks, or decisions, then reference those numbers in those same useless comments.
Or when it references "item 06" from some plan you wrote as if it had any relevance to you whatsoever.
I’ve a codebase filled with references to §x.y section of documentation that Claude itself prepares and never updates; which is exactly what many devs would do, I guess, joke’s on us.
Yeah this is awful. Every codebase becomes a graveyard of references to ideas or behaviors that were barely considered. It's probably also a compounding source of context poisoning when a minority of the comments/documentation are about how the current code actually works.
It also likes to spew references to documents that are not, and never have been, in the repo. So if you're not careful you'l have comments all over your codebase saying things like: foo() - Perform foo action as documented in PRIVATE_INTERNAL.doc
I found this as well, but I found it usually refers to a scratch file it made and purposely did not commit (either by my decision or its). Not that this makes it better, but at least it makes the AI world make a bit more sense to me
shudder
I wish they trained the bots to be a notch more relaxed and less hysterical. Less is more.
But maybe that's just a consequence of the RL training being essentially AI torture to make them do what we want.
This isn't just on code.
When I'm writing technical documentation, it keeps the explanations in. Same when writing non-technical documentation. When I was having it attempt to generate a Pathfinder 1e class for a Sword Dancer, it was leaving in notes about why it removes things I told it to remove/rework.
And it isn't just Claude. I've seen the same with GPT models, with Grok, with Deepseek. Each AI isn't quite the same with how it approaches this, but in every case they seem to have a strong bias to retaining information, even bad information that we want gone, so it is like they have a, dare I say, subconscious bias to retain the information. Putting a note in a comment or explaining why to not do something or something was undone is a good way to retain information while still achieving the goal (well, if you ignore the part about the human intention for the information to be gone).
This then weakens the AI in the future, as I find AI struggles with the more incorrect information. Sure, a comment saying "not X because Y" is less 'context damage' than a comment saying "X" (assuming X is wrong), but it is still a slight shift to X being present in context in some way. One off, AI's seem to perfectly handle this without issue. But after hundreds or thousands of cases build up? The attention mechanism seems unable to keep up and incorrect information flows it. This effectively creates a sort of vibe coding maximum size unless there is a human janitor cleaning up the bad information on the context stays nice and clean.
But this is all simply a feeling I get as I use AI to do different things and isn't at all backed up by any formal study.
This is a common problem, and I don't get why LLMs have not been tuned to stop this nonsense. It is writing comments as if the audience is you, the user in the session, while obviously code comments are meant for future readers.
Follow the tokens