I dread having to find answers on random pages now. Its this weird discomfort of not knowing if I’ve gotten dumber and don’t understand stuff I should be able to, or if it just doesn't make sense. Its so much work to figure out, and im just not sure which at any given time anymore.

The antidote is in the poison. We're going to have to start self-hosting smaller, local LLMs specifically tailored to searching the web to find what you're looking for (and providing links instead of summaries), while ignoring all the garbage.

If it can check thousands or millions of pages in a few seconds, it can find stuff buried so deep under all the slop that we would never find it manually.

Even if you're not ready to accept this yet, at least download a copy of a model or two from huggingface before nvidia locks it down. 7-14B parameters is probably okay if you don't have much space/memory.

Eventually we can use codeberg or ipfs to redistribute what we salvage as a collective. If everyone mirrors a few different models we'll be in much better shape later down the road when big tech companies start approaching their endgame (full user-lock and ecosystem control expanding into the internet itself)