Rohit Krishnan Critiques AI Responses as Overly Garrulous

Rohit Krishnan posted a reply detailing his heavy use of an AI tool with resets applied between sessions. His main observation was that the responses stayed overly wordy despite these efforts. He mentioned constantly needing to ask for summaries as a result. Krishnan suggested that OpenAI might reduce its output tokens by 80 percent easily through better handling of verbosity. The post comes from the creator of the Strange Loop Canon Substack, who focuses on AI and complex systems. Visible in the conversation around the post, this stands as his primary feedback on the model's output style after extensive testing.