Not AI psychosis exactly, but I know a lot of people who get access to tools like Claude Code and start to feel the potential, the amazing potential and can’t seem to think about anything other than building their amazing new strongly-typed functional meta-agent Zettelkasten system, and how they’ve finally figured it out, the ultimate organization of all knowledge, and how it will change their life and the world, but there is no externally visible output other than an increase in them talking about said strongly-typed functional meta-agent Zettelkasten system.
This is a very accurate description of my recent behavior, though I’m working on spaced repetition software. It’s something I’m trying to learn to not do.
Claude code can often create the feeling of progress on a project while producing very little useful output. It seems the models have, for some reason, become good at engendering a feeling of clarity, without actually producing very much insight into the problem.
On the other hand, LLM chat in the browser is pretty unbeatable for queries like “what is a CRDT and what human authored content should I read to really understand them” + a few follow up questions.
Additionally…
I’ve had success with various ‘reformatting’ tasks. I.e. turn this paper into a presentation, or these stream of consciousness notes into a more nicely formatted and well organized document. I’ve had mixed success with “review this” kinds of tasks. They often find issues, but also often produce a ton of false positives that waste a ton of time.
Like silentbob I also know engineers with ‘extremely advanced agentic coding workflows’ but I’ve never been able to replicate their success.
I’ve seen some software engineers in tech orgs behave in similar ways, where I’m really unsure whether they actually managed to 10x their productivity using the super elaborate systems they constructed, with hierarchical teams of AI agents, each having different roles, and the humans just being their orchestrator, or whether they just get extremely excited about these possibilities and set up all these layers of “automation” without this actually leading to more useful output. I can certainly imagine it leads to more output, but I’m somewhat skeptical if it’s high quality, sustainable output, or just huge amounts of ~slop.
Shit, that sounds like me. (Minus the Zettelkasten because my new system would obviously be better, but...)
I guess a part of this is a lack of calibration. You already know what you can realistically achieve without an AI. But you are not sure how much you can achieve with the AI. You are still learning how to use them, and the AIs are evolving, so whatever you have achieved until now seems like the lowest bound on what you could achieve...
The failure mode is talking about it but not doing it. In my case, the excuse is that I have a full-time job (where the AI does not help much, because writing code is not the actual bottleneck) and kids, so it feels like the ascension to godhood is right behind the corner, I just need one free fucking weekend… ideally a week.
If my model is correct, the problem will mostly fix itself as people get more experience and find their upper bounds, too.
Not AI psychosis exactly, but I know a lot of people who get access to tools like Claude Code and start to feel the potential, the amazing potential and can’t seem to think about anything other than building their amazing new strongly-typed functional meta-agent Zettelkasten system, and how they’ve finally figured it out, the ultimate organization of all knowledge, and how it will change their life and the world, but there is no externally visible output other than an increase in them talking about said strongly-typed functional meta-agent Zettelkasten system.
This is a very accurate description of my recent behavior, though I’m working on spaced repetition software. It’s something I’m trying to learn to not do.
Claude code can often create the feeling of progress on a project while producing very little useful output. It seems the models have, for some reason, become good at engendering a feeling of clarity, without actually producing very much insight into the problem.
On the other hand, LLM chat in the browser is pretty unbeatable for queries like “what is a CRDT and what human authored content should I read to really understand them” + a few follow up questions.
Additionally…
I’ve had success with various ‘reformatting’ tasks. I.e. turn this paper into a presentation, or these stream of consciousness notes into a more nicely formatted and well organized document. I’ve had mixed success with “review this” kinds of tasks. They often find issues, but also often produce a ton of false positives that waste a ton of time.
Like silentbob I also know engineers with ‘extremely advanced agentic coding workflows’ but I’ve never been able to replicate their success.
I’ve seen some software engineers in tech orgs behave in similar ways, where I’m really unsure whether they actually managed to 10x their productivity using the super elaborate systems they constructed, with hierarchical teams of AI agents, each having different roles, and the humans just being their orchestrator, or whether they just get extremely excited about these possibilities and set up all these layers of “automation” without this actually leading to more useful output. I can certainly imagine it leads to more output, but I’m somewhat skeptical if it’s high quality, sustainable output, or just huge amounts of ~slop.
org-mode mania
Shit, that sounds like me. (Minus the Zettelkasten because my new system would obviously be better, but...)
I guess a part of this is a lack of calibration. You already know what you can realistically achieve without an AI. But you are not sure how much you can achieve with the AI. You are still learning how to use them, and the AIs are evolving, so whatever you have achieved until now seems like the lowest bound on what you could achieve...
The failure mode is talking about it but not doing it. In my case, the excuse is that I have a full-time job (where the AI does not help much, because writing code is not the actual bottleneck) and kids, so it feels like the ascension to godhood is right behind the corner, I just need one free fucking weekend… ideally a week.
If my model is correct, the problem will mostly fix itself as people get more experience and find their upper bounds, too.