You know that meme with a guy gradually painting clown makeup on their face? That might be an allegory of my posts on AI over the last months, as well as an evolution of my thoughts on it. And I know I’m not alone here.
I went all the way from “AI is great” through “I’ll stop using AI cause it fries my brain” to “I’m using AI for work daily”. I did not change my somewhat distanced approach to AI coding, simply because I still feel you need very solid fundamentals to use it correctly. I also have a growing feeling that we will soon need to explore different avenues of using AI in our workflows, cause AI companies delivering those frontier models are mostly running at a loss. We are actually already seeing price hikes, like Github Copilot moving from premium request-based to token-based pricing. In my opinion, using AI coding tools will become significantly more expensive as time goes by. Agentic coding might be the most effective, but is also the most token hungry way of using generative AI for coding after all.
With large language models becoming beefier, they will also require more compute power, which translates directly to even more expenses. And with all significant players in the AI race being for-profit companies, this will translate to further price hikes. One way of mitigating it might be setting up locally running models, which comes with numerous benefits, but ultimately will never be as powerful as frontier models running in massive data centers. They might just be enough for coding work, but I didn’t play around with it enough to be sure yet.
To summarize, I will continue using generative AI for work, since that’s what my company expects from me. I’m in the privileged position of knowing how to code without AI, but learning to use it for my daily work is broadening my skillset. I will also continue coding my passion projects by hand, to keep my brain engaged, while using AI to plan out features. Utilize it to its fullest it while it’s still cheap.