# The World with A(G)I Recently, I was fleshing out my thoughts on how the massive resources being dedicated to AI data centers would pay for themselves in the form of immense "optimizations" returned to humanity. The environment, energy, anything to do with allocation, could be solved with a superintelligent model. I was reading the model card for GPT 5.6 at work. I pray, for my job, that scaling laws don't hold. I saw a tweet a couple days ago: > if we could coordinate a global capabilities slowdown today i would likely press that magic button I would also press that button. Would I press it if it was a shutdown, instead of a slowdown? Three scenarios: 1. Scaling laws hold, but AI is not monetized, and scaling stops before we reach superintelligence. 2. Scaling laws don't hold. 3. Scaling laws do hold, and we continue to allocate capital until it reaches a superintelligent state, where it can provide near infinite resources to humanity. _First Two Scenarios_ If the first holds, it becomes a question of how good it gets before centers are allocated towards inference instead of training. These are both likely "bubble" scenarios. I believe the former to be less likely with the existence of Google and Meta, who have revenue to buy all the compute from AI companies that get wiped out post-popping, and continue scaling _(would they actually do that?)_. If the second occurs, a lot of jobs will be gone. I don't really know what the tech world would look like after a pop. I attended a company all hands meeting where Zuckerberg was talking. He believed that AI would be a tool to SWEs, helping us build more. I agree with this, somewhat. I think some companies have much less of a use from AI then others. Companies in "maintenance mode," where the central product is already built, and moving fast/adding significantly more lines of the codebase has diminishing returns, would probably use AI to replace employees. Companies where there is a near-linear relationship between new features built and actual value provided would probably use AI in the way Zuckerberg is saying. I don't know what the split would be between 1. employees "replaced" by AI and 2. employees displaced due to a decline in capital allocated towards AI would be. _The Last Scenario_ My thoughts on this change a lot. I think the central idea is that I would like the delta between AI can do all software engineering <> AI provides infinite resources to be as short as possible. The latter seems like a really big goal - but I believe it to be possible (is it likely?). I don't know what it would look like, or how it would happen, but if we reached superintelligence, it would find a way. I'm not sure how much compute this would require. Looking at what we have now, we would need to continue immense spending on data centers. Meta Q2 2026 earnings showed that investors are already growing skittish, and want profits soon. Perhaps I am too bullish on AI. Telling people that I think there is a good chance it might provide infinite resources to humanity feels stupid sometimes. I guess I'll see.