I have seen people just generate large docs with Claude cowork and they themselves have not scrutinized it or know why/how it's useful. It's just kind of impressive in its volume and well formatedness. And then they dump it in your lap as being helpful
This. The amount of long winded unedited docs people think it’s ok to dump on me now is unbelievable.
Organizational self-imposed DOS attacks.
It's ok, you can use AI to summarize the key points. No need to read anymore! /s
I wish. I have to fix all the bs before things get sent out. It’s so exhausting.
> And then they dump it in your lap as being helpful
I've been guilty of this and gotten pushback from my manager: "this feels like homework, cut these options down to 100 words each, max".
Curation and refinement are even more important when you can have genAI generate reams of text.
Seeking outside signals is even more important, like talking to customers, looking at real usage data, and more. It's too easy to trust believe what Claude tells you, even if you say "please argue against this idea", which you always should.
We often see this bizarre workflow where notes, like engineering notes, are converted to large prose using AI. And then, the large prose is converted back to short bullet points on the other end for summarization!
It's all fun and games until some high level executive realizes everyone is using it and still demanding the same paycheck.
So what? They produce more for the same money. What's not to like?
> They produce more for the same money.
The word "more" there is doing a lot of work.
What is the "more"? Is it:
- more documents and text or more understanding
- more code or more valuable features
- more things to throw against the wall or more considered experiments
It's way easier to do the first things instead of the second.
I know, I know. It was more a tounge in cheek comment to something I assume was irony as well. To be fair to GP, management doesn't get any of this...
Adding is always easier than reducing to the essence.
I find it interesting how herding agents has so much in common with being a team lead. Constant struggle between too detailed and too loose instructions. One difference though is that the team learns from you, but with agents it's only you who adapts. Saying that, because I don't count instructions or anything in the context window as adapting.
Yep, I've received a few powerpoints like that.
I'm using Claude to write large files too, but it's a very iterative process and involves a lot of reading and correcting.
I'm beginning to see this in my industry (consulting). I was at a client site last week and in a room with some heavy hitters both from my side and client side but in a casual setting (lunch). Everyone was discussing how they sometimes "cheat" using genAI to put together decks when one of the out-of-the-blue 1 sentence questions that takes 4 hours to answer come down from the c-suite. They all said they heavily edit the output but at least it gives them a place to start. I have my doubts though, i wonder how many times they just take it as gospel and forward the deck on.
to be fair, i've been guilty of this with code. Ask claude to generate a python script that takes X as input and produces Y as output, run it, pipe to more, output looks ok but i don't check everything, write it to a file, send it on.
We've really reached the point where one person uses AI to create an impressive report based on a few prompts with some keywords, and the receiver uses another AI to summarize the report to a short TL;DR that's almost identical to the input prompts.
This, creating order from chaos (reducing entropy) is difficult and requires real intelligence. Inflating some small prompt into a wall of text and creating a bunch of entropy in the process is not as useful as it appears.
"Simplicity is the ultimate sophistication". I'm more impressed by a pithy sentence than 100 pages of statistical fluff.
This reminds me of that game of telephone. Eventually, the message gets morphed and transformed into something different from what was originally said. Is this really what we want?
I'm a victim of this. Very bad taste of AI generated gibberish that was obviously not read through before being sent.
[flagged]