Researchers have discovered that appending 300 filler tokens (dots) to a multi-hop reasoning question significantly improves a frontier large language model's ability to answer correctly, even though the model initially fails without the filler tokens. The finding suggests that additional tokens provide the model with more processing steps, effectively enabling it to perform the necessary reasoning without explicit chain-of-thought prompting.
If you ask a frontier LLM a multi-hop reasoning question, e.g., "Who won the Nobel Prize for Chemistry in (1900 + Mozart's age when he died)?", it usually can't answer correctly immediately (no thinking) BUT if you ask the same question & append 300 dots, suddenly it can answer?
EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.