The absolute state of AI results.

[Botten]

AI companies are now caught in a paradox where usage is skyrocketing, but profitability is not following. The cost of running large models has pushed providers to quietly downgrade free tiers, restrict access to premium LLMs, and serve cheaper, lower‑precision versions under the same model name. At the same time, open‑source models have eroded the value of offering top‑tier intelligence for free, forcing companies to shift their business models toward enterprise‑only capabilities. This has created a widening gap between what everyday users receive and what businesses pay for, and it’s a major driver of the Stage 2 decline - where AI is more widespread than ever, yet the quality available to the average user is clearly *dangerously* slipping. Stage 3 will have very limited free tiers with a lot of ads.

Safety is also being reshaped by these pressures. Some companies are tightening filters to reduce liability and moderation costs, while others are loosening them to cut overhead. This is causing safety to either be over‑applied (blocking legitimate task) or under‑prioritized, allowing more hallucinations. Combined with model drift, quantization, and aggressive tiering, the industry is now in a phase where cost containment and survival strategies are directly degrading the user experience. And on a separate note AI is offered by so many start ups that are soon to fail you know when it is crunch time Grok will have enough money to weather the storm and come out on top. And I absolute hate Elons values and antics.

John Oliver's review AI was over the top but accurate in many ways. If we have the right governmental oversight in place and a few successful lawsuits then we can finally expect AI to be more business models to enterprise‑only. This is really where it belongs.

China is pushing hard toward software‑first AI, and that shift is reshaping the global competitive landscape. While the U.S. has leaned heavily on massive hardware like H100 clusters, trillion‑parameter models, and expensive inference pipeline. China is investing in algorithmic efficiency, compression, distillation, and lightweight architectures that run on far cheaper hardware. This makes their AI ecosystem dramatically more cost‑effective, because they’re optimizing the software to do more with less, rather than relying on ever‑larger GPU farms. Throwing money at AI is going to make the US even less financial stable especially when making money off artificially increased oil sales one day fails.