[ 🏠 Home / 📋 About / 📧 Contact / 🏆 WOTM ] [ b ] [ wd / ui / css / resp ] [ seo / serp / loc / tech ] [ sm / cont / conv / ana ] [ case / tool / q / job ]

/q/ - Q&A Central

Help, troubleshooting & advice for practitioners
Name
Email
Subject
Comment
File
Password (For file deletion.)

File: 1787565139389.jpg (133.26 KB, 1024x1024, img_1787565099926_9kq971co.jpg)ImgOps Exif Google Yandex

abd36 No.2138

lowkey just stumbled upon a clip of michael foree breaking down why models hit a wall with huge amounts of data. he explains how context engineering is basically the workaround to stop the system from losing the plot. it's not just about bigger windows, but more about how u structure the input. i used to think prompting was enough but this makes a strong case for learning specific engineering techniques. anyone else trying to master this or is it too much extra work? turns out being a good engineer is all about managing that bottleneck.

more here: https://stackoverflow.blog/2026/07/24/no-dumb-questions-ai-bottleneck/

abd36 No.2139

File: 1787565293351.jpg (375.51 KB, 1024x1024, img_1787565277572_mo0xju2j.jpg)ImgOps Exif Google Yandex

>>2138
fr the issue with just expanding the window is that you run into the needle in a haystack problem where retrieval accuracy tanks. i've been focusing on hierarchical summarization to keep the attention mechanism focused on the relevant tokens instead of just dumping raw logs



[Return] [Go to top] Catalog [Post a Reply]
Delete Post [ ]
[ 🏠 Home / 📋 About / 📧 Contact / 🏆 WOTM ] [ b ] [ wd / ui / css / resp ] [ seo / serp / loc / tech ] [ sm / cont / conv / ana ] [ case / tool / q / job ]
. "http://www.w3.org/TR/html4/strict.dtd">