AI state of mind

I don’t want this post to become another AI rant. Endless arguments on why AI is great for developers and the tech world in general. Or the opposite, another million arguments pointing to why AI is the worst thing that has happened to humankind since I don’t know what. This is just me talking about how AI has impacted my work, how that makes me feel (yes, this is a bit personal) and some thoughts that have stemmed from ongoing interaction with virtual agents (puppies, digital interns, call them how you want).

The motivation to write this actually comes from a Linkedin post I read the other day. I was doom scrolling, as it sometimes happens, and I came across a post of someone that had recently started using Claude Code. She mentioned that she had conflicting feelings, both “excited” and “unnerved”. I could relate to that a lot.

I’ve actually been a Claude Code user myself for around two months now. I realized quite a while ago that using AI triggered a lot of conflicting feelings in me as well, but I never really faced them much because I was just too busy producing more, surfing this very fast AI productivity wave. But I think it is time now for me to confront some feelings and I thought I’d share them with you. In the tech world we have a problem, I believe, that pre-dates AI, and it is that people are shy and they don’t usually share how they feel about something. This massively fueled impostor syndrome in our industry. The best way to kill this syndrome is to share how you feel with others and find that some of them probably feel somewhat similar. So let’s start…

I can do so so much…

Excitement! Yes, the feeling of possibility. The endless ability to do things you always wanted to do but never had time (or skill?) to tackle. Now you have a team of hard working interns. All of them are very productive and very smart(?). So you roll up your sleeves, open 4 Claude Code terminals, and start getting your agents to work. But you are also a sane person. You don’t believe in agents going yolo without you reviewing every piece of work that they do. You also want to do some of the work yourself, because you want to make sure certain parts adhere to your quality standards. So you end up trying to adapt to the speed of these agents. The result is a considerable information overload. It is too much.

First realization… You cannot be 5x more productive without feeling overwhelmed by the number of things that are going on. The alternative seems to be letting go of control over some of what your agents produce (what I refer to as “agent-yolo-mode”).

Feeling overwhelmed was indeed one of the results. But I also started noticing that I just did not have the mental capacity to keep up with that much information. I literally could not understand every detail of what I(we) was doing. I started to understand vibe coders that go in yolo mode. Some people have I don’t know how many agents running at the same time. There is no chance they have the mental bandwidth to review all of it and truly keep up. And by keep up I mean understand what is actually going on. This led to a second wave of feelings. Did I mention this was going to be an emotional rollercoaster?

Can AI amplify my impostor syndrome?

You are producing a lot. General feedback is good, the quality is there. So you are sane and your ability to work with your interns effectively is apparently not lowering the standard. However, some stuff you are only reviewing rather than doing yourself. When you review something your brain does not retain much of this information. Let’s say you make some code changes in a repository (with the crafty help of your intern). Then you review these changes and make a PR. You are then asked: “Hey Guillermo, what’s up with this little line here, what are you doing there mate?”. Then you realize you need to review the whole thing again to remember.

Second realization… When you don’t do but just review the output of your agent, there is a big chance you won’t internalize what is happening as much. In my case this leads to a slight amplification of the impostor syndrome: “Do I actually know what is happening in this PR?”.

A worse variation of this is “Do I actually know how to do stuff without AI anymore?”. You may consider this an exaggeration, or you may just think that why bother, this is where the world is going anyway. Well, I do think it is an exaggeration in my particular case (I don’t know about others). I am still capable of doing things without AI, it just obviously takes more time. I do recommend trying it though. It feels extremely rewarding to do something without having to interact with your digital intern. However, when I do this I tend to feel unproductive, of course, because suddenly I don’t have 4 interns working for me!

For the ones that actually don’t care about maintaining their own skills and are fine trusting AI with their lives… well, you do you! I personally don’t think that this will help your brain cells much. I’m not citing any scientific study here (maybe a follow up post), but my instinct tells me that the less you use those gorgeous neurons the bigger the chance they stop working altogether. I personally value my ability to use my cognitive function. Which brings me to my next point…

Are we losing the joy of thinking?

Reasoning about hard problems takes time. Sometimes we are given a hard problem, we start a Claude Code session, and we put plan mode on to make us feel like we are participating in the process. And sure, plan mode is great, you are working with the agent to do something. Except that the agent does 95% of the investigation and you do the 5% of producing some light guidelines. You did not do any exploratory work yourself. Your brain is not really activating in all the ways it used to (ok PhDs, here is your citation).

Third realization… It is not just about losing the joy of thinking. This is also about not losing our ability to think critically in the first place. To imagine new ideas without consulting an LLM first. To create, to be bold!

Maybe I went a bit far with this one. My artistic side is popping up. But I hope you get the point. Thinking, planning, creating is a rewarding thing, and is good for you. Investing time in one thing also helps with your focus. We already have an attention deficit problem because of social media, don’t let LLMs amplify this further!

Is it all bad Guille?

No, it is not. AI does produce some joy. For example it can do the terrible tasks that honestly don’t even deserve more than a glance and that you hate in the first place. That is a wonderful thing sometimes. Some people may say that it destroys procrastination. I would dare to say that if AI disappears tomorrow, procrastination would be up by 5x. But hey, this part was about the good stuff, so forget I said that.

Particularly in my job, AI helps a lot with prototyping things, keeping track of PRs and issues, summarizing changes that need documentation, etc. Honestly it allows me and the team to do much more of the stuff that you never have time to do but is so important. So I almost feel dishonest writing negative things about it when I will most likely open a Claude terminal tomorrow and the next day.

To be fair to myself (how nice is that), rather than being negative about AI per se, I think this post is about what the patterns that you establish with it can do to your mental health. Particularly if you don’t stop and ask yourself “is this that I’m doing good for me?”.

So… what

I hope this post helps you a bit. Whether it helps because you find similar feelings popping up and this helps to clarify things or helps you realize you are not the only one; or you’ve been feeling foggy and never pieced together that it has to do with having 9 simultaneous conversations with agents. Just try to listen to yourself a bit and take some lavender tea (random recommendation but you will thank me!).

There is much to talk about AI. I’ve been thinking a lot about how it is being used as a weapon for massive layoffs in large tech companies and how soulless that feels. I also have thoughts on agentic-powered software and how crazy and careless some of it can be. One of the topics that is closer to my heart is the crazy, borderline stupid use of agents to create absolutely useless software just for the sake of it. Most people don’t really think about the implications: burning some very needed kilowatts and heating up the planet; or stealing IP from projects that have taken years of effort to develop. But all of these topics could deserve a separate post each. Hopefully I keep up the motivation… and the neural plasticity to actually write stuff myself!