[firasd]

I've always thought that extensive throat-clearing and prefixing the Treaties of Westphalia-length instructions into the context window was unnecessarily baroque when you can just talk to the agent.

I guess part of it is also that I don't mind doing 'hand-edits' like for example LLMs love to say "// so and so removed" I just go and remove that manually later rather than being like "don't comment about what you removed!11" cause you're really fighting deep grooves in the model's behavior at that point.

But I also have a hands-on human-in-the-loop working style so I guess maybe for people who just want to say "implement all open features in github issues" and walk away maybe there needs to be more of all this CLAUDE.md stuff

However I suspect there was always some gearhead type attraction to setting up detailed harness configs that may be unnecessary and more like hobbyist tinkering.

[zahlman]

> I guess part of it is also that I don't mind doing 'hand-edits' like for example LLMs love to say "// so and so removed" I just go and remove that manually later rather than being like "don't comment about what you removed!11" cause you're really fighting deep grooves in the model's behavior at that point.

I feel like this is the way. There are surely things where it's faster; certainly it's more pleasant to do simply things yourself than repeatedly try to figure out the magic words to communicate the idea while outsourcing it. Whether it's to an LLM or to another person.

[anon22981]

I’m ashamed to admit that I once suddenly realized what I had done: I asked the LLM to rename a variable. It took it some minutes or so. With IDE by hand it would have been seconds.

[mexicocitinluez]

The other day I asked it to update a handful of files using a new pattern I built. It was tedious work, and probably would have taken me 10 minutes, but I figure why not just let Claude do it. Well, it spun up multiple agents, each taking a different approach (one writing a Perl script, the other just doing fancy regex, etc) and 30 minutes later had the result. Granted, it wasn't my 30 minutes it was using, but it felt wasteful nonetheless.

Now, if it's something minor, I'll ask Claude about it, and get the gist of what I need to do, and just do it myself.

[jayd16]

I'm hard pressed to believe it didn't also take 10 minutes of your time checking in and head shaking and reviewing and testing.

[rendaw]

I'm ashamed that I use a programming language where the IDE can't rename things, so I rely on AI to do it...

[zahlman]

Am I the only remaining person who enjoys going through a manual rename as an excuse to review the code and refresh my mental model of it?

[mettamage]

The annoying issue sometimes is that while it is faster to do it yourself, the context of the LLM is not updated. So now you have you to mention that and mentioning it is a bit fuzzier sometimes than the LLM simply doing the rename itself, so the context stays up to date.

[acedTrex]

The llm will figure it out when it relooks at the file.

[fibonachos]

This feels like a good place for tighter integration between editors and agents. The editor could notify the agent of user authored changes.

[acedTrex]

This is very trivial to implement in any serious tool/workflow. I have a version of it in my pi nvim setup.

[duxup]

Agreed I just find my own dance and adjust as I go.

Sometimes like a verbose coworker who just is that way… fine Claude, you be that way now. Things seem to change here and there anyway and sweating the small stuff of having to repeat myself, that’s ok.

Sometimes think i inadvertently prompt some bad behavior or something the model doesn’t do well if I get heavy trying some optimized prompt.

[ComputerPerson]

I was surprised by how abstract the article was.

I fall between your human-in-the-loop and hobbyist tinkering limits, where I want to force Claude to atop and talk to me at only a few specific points. I'm still not sure if my 600-word prompt templates are overbearing or not.

[rendaw]

I think it's action bias https://en.wikipedia.org/wiki/Action_bias

[Y_Y]

> But I also have a hands-on human-in-the-loop working style

Literally HitL-er

[novaleaf]

> I guess part of it is also that I don't mind doing 'hand-edits' like for example LLMs love to say "// so and so removed" I just go and remove that manually later rather than being like "don't comment about what you removed!11" cause you're really fighting deep grooves in the model's behavior at that point.

Maybe I'm misunderstanding you, but that's just about the best example possible for using AGENTS/CLAUDE.md. Just add "don't comment about what you removed!11" and you never have to say it again...

...but you'll get constantly nagged about the `11` of course!

[anon-3988]

> Maybe I'm misunderstanding you, but that's just about the best example possible for using AGENTS/CLAUDE.md. Just add "don't comment about what you removed!11" and you never have to say it again...

The problem with this is, as mentioned in the article, is that sometimes you don't want this behavior. Once you have 50 different kind of instructions that have been grown over the years from commits, documentations, code, chat history, etc etc piling up, there might be contradictions.

The point is to go back to basic. Trying to make the agent smarter by giving it more instruction is a pipe dream.

[flawn]

I feel that. I thought I am the only one who has this issue, with Claude always adding this superfluous BS. I feel totally backwards when realizing I am trying to teach this model how to write good technical docs, while I could have already done it in the same time.

Anyone has experiences with other models? I feel like GPT is much more concise?