I found myself increasingly suspicious of everything I read online. I see people using LLMs for the most random tasks, things that they would be better off doing themselves, and it would turn out better both in terms of time spent and quality.
There's this propaganda of universal productivity enhancement which stems from case studies showcasing dramatic throughput gains in basically every task thanks to the use of this technology. The narrative is also sound:
The AI automates the boring, routine stuff, while the human focuses on the higher-value work
This sounds immaculate to me. I don't disagree with it and I'm looking forward for such a future with excitement. But here's the problem I think it creates. Once you witness the potential of this technology, you don't stop at the boring parts and you offload more and more, gradually slipping into a frenzied state where you just need to move on.
Something I noticed on myself, and I came to despise, was the nonsensical rise in the number of tasks I worked in parallel, because I was sure that delegating the others to an AI agent would give me a result good enough, that I would revise later. Guess what. That later never came: by the time I might have gone back, I was already buried with ten new priorities. There's no more time to care about the fine details or do things with genuine care, because your social media feeds keep telling you that you are falling behind, that you are not shipping enough.
And who are we (supposedly) falling behind? Twitter/Linkedin gurus (most of the time bots) who show off far more than they actually create, instantly switching to the latest framework/tool of the current wave and going full agentic on everything they do. I believe that those examples have a corrosive effect on our psyche. I stumbled multiple times in the error of comparing my real, messy but careful work against posts and threads of those omnipresent buildfluencers [1]: engineered to keep interactions flowing and pumping the latest token-wasting project powered by an army of AI agents. But that's the pitfall of starting to delegate everything to AI: each offload feels like a relief, so I'm tempted to take on more, and more, until I've lost track of what I'm doing.
When efficiency becomes the ultimate measure of value, human beings are tempted to see themselves as a project to be optimized rather than as persons.[[2]]
That's why I have the ominous feeling that the fear of missing out is dangerously becoming the driver of the important work, where the only goal left is to keep pace with the agent.
As regards the quality of the output, when discussing with friends I really like to quote a majestically written blog post by Caleb gross, which says:
Creation is the distillation of intent into form
To comment on that: when it comes to create something novel, in my opinion, you should at least be able to work alongside the agent, and that requires you to accept that you are the bottleneck in the process. That's a good way to keep you in the loop. You should slow the process down to the rate at which you can still clearly think, judge and own the result. Constraining the system to do so it's not a failure to optimize away; it's the only thing that keeps the work yours.
Buildfluencers are engagement bait accounts on social media that constantly show off their productivity. If you see punchlines like "
This changes everything. 🧵 (1/N)" or "We're so early. If you're not running 12 agents in parallel while you sleep, you're already behind", well, you found them.↩︎Magnifica Humanitas, 112.