Great article. AI seems to be a faster version of a big tech wave ie TV did not replace radio completely. And you are right… practice makes perfect with anything….
Software-as-a-Service is evolving into Employee-as-a-Service, bringing massive disruption in business quarters, not years. By the end of 2025, the workforce will be in an obvious state of collapse—prepare for it.
Interesting. Sounds pretty natural to me. And if anything, it would sound more like Claude not ChatGPT as I used the former.
Considering it has performed really well, the question arises - when does that language become too much for the reader? This piece has been shared a ton and a lot of people find it valuable.
Yes, there are several dead giveaways in content structure.
It has performed well because fo the effort you put into inserting graphs and creating a narrative. And also because not everyone yet detects ai sentence structure.
But most importantly, it is performing well because the message is something people want to hear even if it is not entirely right. There are large parts of knowledge work where there is no way for people to charge what they used to in the past. Your message in its own way is right but only to certain types of work.
On the first two points - isn't the narrative and arguments what matter most? And if everyone doesn't detect AI sentence structure, what's the value in them learning to do so if they are just here for the points being made? Now if you are reading something purely for its writing style and the art of writing, that's a different story.
As for the third point on knowledge work - that is a fact. But I don't think there is any way to go back to the past. These workers will have to augment with AI in these roles or there will be no role IMO. In my piece, the top 2 roles we are discussing are writing jobs and software development. In both of these, the roles will change dramatically due to AI.
I think people spend too much time trying to detect AI in writing in places where it is less important. In this piece in particular, I explicitly state that AI is used in the writing. There is "good" AI writing and "bad" AI writing for the sake of not making this too long. The point is, eventually all of our content out there will be shaped in some way by AI.
I'm not sure it's about spending time detecting it. Once you use AI as often as I do in your workflow, you pretty soon figure out its sentence structure.
That's why I said 'the ONLY problem'. The fact that the rest of the post was well-written made the AI-generated sentence structure all the more jarring.
I think people are going to get trained pretty fast and will start detecting it since all LinkedIN posts now have the same structure.
That said, with a bit of intelligent prompting, any LLM can create very high quality human-style text... it's just that most people haven't yet figured which prompts work closest to their writing style.
LinkedIn is unbearable. I think it will get pushed out with time.
But yeah, I agree with your last point. That's what I try to help people do. With structured prompts, custom instructions, project knowledge bases with a lot of real examples, etc. you can really create a pretty good clone of yourself. The problem is most people don't make it to that point.
This piece was actually written before the release of Claude custom writing styles. I urge readers to check out some of my newer ones and compare. You might be able to see what the evolution looked like.
Great article. AI seems to be a faster version of a big tech wave ie TV did not replace radio completely. And you are right… practice makes perfect with anything….
Software-as-a-Service is evolving into Employee-as-a-Service, bringing massive disruption in business quarters, not years. By the end of 2025, the workforce will be in an obvious state of collapse—prepare for it.
The only problem is that the language of most of this post feels chatgpt generated.
Interesting. Sounds pretty natural to me. And if anything, it would sound more like Claude not ChatGPT as I used the former.
Considering it has performed really well, the question arises - when does that language become too much for the reader? This piece has been shared a ton and a lot of people find it valuable.
Yes, there are several dead giveaways in content structure.
It has performed well because fo the effort you put into inserting graphs and creating a narrative. And also because not everyone yet detects ai sentence structure.
But most importantly, it is performing well because the message is something people want to hear even if it is not entirely right. There are large parts of knowledge work where there is no way for people to charge what they used to in the past. Your message in its own way is right but only to certain types of work.
On the first two points - isn't the narrative and arguments what matter most? And if everyone doesn't detect AI sentence structure, what's the value in them learning to do so if they are just here for the points being made? Now if you are reading something purely for its writing style and the art of writing, that's a different story.
As for the third point on knowledge work - that is a fact. But I don't think there is any way to go back to the past. These workers will have to augment with AI in these roles or there will be no role IMO. In my piece, the top 2 roles we are discussing are writing jobs and software development. In both of these, the roles will change dramatically due to AI.
Appreciate your points.
I think people spend too much time trying to detect AI in writing in places where it is less important. In this piece in particular, I explicitly state that AI is used in the writing. There is "good" AI writing and "bad" AI writing for the sake of not making this too long. The point is, eventually all of our content out there will be shaped in some way by AI.
I'm not sure it's about spending time detecting it. Once you use AI as often as I do in your workflow, you pretty soon figure out its sentence structure.
That's why I said 'the ONLY problem'. The fact that the rest of the post was well-written made the AI-generated sentence structure all the more jarring.
I think people are going to get trained pretty fast and will start detecting it since all LinkedIN posts now have the same structure.
That said, with a bit of intelligent prompting, any LLM can create very high quality human-style text... it's just that most people haven't yet figured which prompts work closest to their writing style.
LinkedIn is unbearable. I think it will get pushed out with time.
But yeah, I agree with your last point. That's what I try to help people do. With structured prompts, custom instructions, project knowledge bases with a lot of real examples, etc. you can really create a pretty good clone of yourself. The problem is most people don't make it to that point.
And I'm sure that will only get better with time.
This piece was actually written before the release of Claude custom writing styles. I urge readers to check out some of my newer ones and compare. You might be able to see what the evolution looked like.