You are sick of reading about AI. That's fair. I am, too. However, I am not sick of writing about AI because I am still figuring out my feelings. Here we go!
LLMs Are Not Useless
I have a problem with people who don't know that LLMs are actually very useful. I myself was a massive hater until I finally paid to use a frontier model. The quality of the output was inarguable. Yes, hallucinations, but not really that often. Yes, code quality issues, but not really that major. And in exchange, I've gained a lot of utility from it. I like learning things from it (which I fact-check, but it's almost always right). I like making plans with it, and I like asking it for advice as a neutral third-party.
Grief
I had a grieving process when I realized that coding, something I had spent a significant amount of my life doing, something that is a strong part of my identity, is kind of over. I still can't fully "vibe-code": I treat Claude as a paired programmer, but truthfully where I wish it was a dialogue between the two of us to work out ideas, it's really just Claude driving all the decisions with very little guidance needed from me. I just look at the outputs, they're almost always correct, then copy and paste. I start skimming more and more, with no consequence. I feel like Claude is doing the work and I'm slowing it down.
Serious Issues
The problem is a failure of our American culture and leadership. We have disgusting, abhorrent people in power, both elected and not. They are what is wrong, not fancy autocomplete. We need laws that protect people, protect labor, that encourage freedom and flourishing and to reject and undo the consolidation of power.
I used to think the problem to everything was capitalism but now I have a more nuanced view: some of these people make irrational decisions that are certainly enabled by the free market but not justified by it. Instead, they're just evil. They want power. This system is failing to hold them accountable, but other systems would fail in different ways. We need to address power itself, we need to decide on when power is just and what to do when it is unjust; pointing to the stock market or a bank account isn't enough.
When I think about the comically evil possible future, I imagine Grok has, by law, been required to refuse to acknowledge more than two genders when generating code. When I think about the real threats today, I see LLMs used by children. We absolutely need to ban these tools for anyone under the age of eighteen.
This Post is 100% Organic
You can use LLMs for reviewing blog posts. "Help me organize my thoughts", "help me improve my thesis" etc. But I refuse, because it all converges towards a specific point-of-view, a specific vocabulary, a specific voice and that must be avoided. Worse to use it for generating any creative stories. Ironically, this kind of convergence towards predictable and bland is what makes it so great for code: I want code to be as readable and predictable as possible.
In Conclusion
I'm sorry, this reads like a high school book report, but I either ramble and publish or I don't publish, and I'm not going to get any better if I don't publish!
I'm pessimistic about the future. Things are going to keep getting worse. But, eventually, things will get better again. I hope. In the meantime, we're living in a bizarre world that is getting eaten by LLMs and it's up to us and our communities to do the best we can.