#ai #capitalism

Inspired by this blog post, i made a quippy post on the fediverse but i wanted to actually talk about something a bit deeper. Capitalism fundamentally warps the mind. We (Americans) are deeply indoctrinated into a society built upon extreme violence and theft of land, labor, generational wealth, and the lives of many. We work jobs we don't necessarily agree with morally in order to survive in the economy, which depending on random chance, you will either be born into wealth or poverty or somewhere in between, giving you a large advantage or disadvantage depending on where the dice land. This has a large determining factor in the access you have to further capital. "The rich get richer" as they say.

I open with this because it is important to me to understand how much those who boost agentic coding tools, like Cursor, spend on them. If I was spending almost a grand a month on anything, you bet i would be singing its praises and hyping it up.

How do you know someone has a Tonal? They'll tell you about it. They're expensive (less than $900/mo tho) and backed by a startup so of course anyone who has one wants the company to succeed and not fail as to brick the best part of the machine (imo), the instructor led classes (only available through an additional subscription. The machine works for weights without any internet connectivity), is going to hype it up it. It's the same deal with these agentic ai tools, but there is a lot more at stake than one's personal fitness and health in the slop business. The boosters are gambling with the future of the economy and jobs that currently feed and house us in the software industry. They argue that those who do not submit to the will of the machine gods will be left behind by those who do.

The current rebuttal of most criticism of these tools is one flavor of "have you used it?" and dear reader I have gotten use out of LLMs! But i self host, or use the tools at work my employer paid for (Gemini, GitHub copilot) and so far i have only found them useful on one thing i couldn't solve on my own that with a bit more research i could have figured out. but yes it did save me some time in that one case. And of course this is not to mention the valid concerns of ethical training etc, which is why i rarely use them. I can also say the amount of small bugs i have written by lazily accepting GitHub copilot's suggestions has wasted me more time than i have saved with the tool, but that subscription is on my employer's dime. Should i have been more careful before hitting tab and accepting the bug? Yes! But the point of the tool is to dull your senses and your mind. It was working as intended.

So i do applaud Mr. Steinberger for giving us a glimpse into his LLM spend and i implore the rest of you ai boosters to do the same. For the record, i have not spent any money on LLMs save for the extra electricity it takes for my gaming PC to run open-webui and ollama on WSL2. I do not have a number for that but given the PC isn't even on all the time, it's negligible to my overall power bill. But I wonder how much people will get squeezed for once the VC money dries up and they start demanding their returns?