| ▲ | gpt5 2 hours ago | |
It’s a 1,000 pages dense document. I’m not even sure the PDF conversions are accurate enough to feed it ti an LLM. Also - if that’s the future of math (larger and larger proofs and algorithms, to the point they are far beyond a human’s capability to comprehend, then I wonder how mathematicians work would look like. | ||
| ▲ | schleck8 10 minutes ago | parent | next [-] | |
You can cut off the appendix so it's roughly 500 pages long and Gemini 4 can digest it in one go (context window is around a million, this takes up around 300K tokens so it's inside the effective window). | ||
| ▲ | nojito 9 minutes ago | parent | prev [-] | |
>I’m not even sure the PDF conversions are accurate enough to feed it ti an LLM. It's best practice to feed in images to the LLM. It works really really well. | ||