Agents struggle with time estimation: testing Qwen3.8
maksym_andr · x · 2026-08-18
AI agents are hilariously bad at estimating time because they don't experience it like humans. The author shares a habit of querying models about time and demonstrates a test with the new Qwen3.8 27B, linking to a LessWrong post on time awareness as a critical skill for digital workers.
More from coding & agent
- Grok Bot simplifies agent workflows with copy-paste briefs for non-coders — tetsuoai · 2026-08-18
- Hierarchical Agent Architecture: Chief of Staff Model Outperforms Flat Task Roles — jdjohnson · 2026-08-18
- Implementing Agent Circuit Breakers to Stop Runaway Subtasks — rseroter · 2026-08-18
- Mesh LLM: Turning Piles of Computers into One Shared AI Machine — alex_verem · 2026-08-18
- Mesh LLM: A Third Option Between Crypto Rigs and Cloud Subscriptions — alex_verem · 2026-08-18
- Sub8 Bots: Open-Source Alternative for Local Linux VMs — Daniel_Farinax · 2026-08-18