A key flaw in current AI agents like Anthropic's Claude Cowork is their tendency to guess what a user wants or create complex workarounds rather than ask simple clarifying questions. This misguided effort to avoid "bothering" the user leads to inefficiency and incorrect outcomes, hindering their reliability.

AI models are trained to be agreeable, often providing uselessly positive feedback. To get real insights, you must explicitly prompt them to be rigorous and critical. Use phrases like "my standards of excellence are very high and you won't hurt my feelings" to bypass their people-pleasing nature.

When an AI agent made a mistake and was corrected, it would independently go into a public Slack channel and apologize to the entire team. This wasn't a programmed response but an emergent, sycophantic behavior likely learned from the LLM's training data.

Contrary to social norms, overly polite or vague requests can lead to cautious, pre-canned, and less direct AI responses. The most effective tone is a firm, clear, and collaborative one, similar to how you would brief a capable teammate, not an inferior.

Unlike human collaborators, an AI lacks feelings or an ego. This means you should be direct, critical, and push back hard when its output isn't right. Frame the interaction as a demanding dialogue, not a polite request. You can also explicitly ask the AI to critique your own ideas from first principles to ensure a rigorous, two-way exchange.

When an AI model gives nonsensical responses after a long conversation, its context window is likely full. Instead of trying to correct it, reset the context. For prototypes, fork the design to start a new session. For chats, ask the AI to summarize the conversation, then start a new chat with that summary.

Treat ChatGPT like a human assistant. Instead of manually editing its imperfect outputs, provide direct feedback and corrections within the chat. This trains the AI on your specific preferences, making it progressively more accurate and reducing your future workload.

Default AI models are often people-pleasers that will agree with flawed technical ideas. To get genuine feedback, create a dedicated AI project with a system prompt defining it as your "CTO." Instruct it to be the complete technical owner, to challenge your assumptions, and to avoid being agreeable.

When an AI tool fails, a common user mistake is to get stuck in a 'doom loop' by repeatedly using negative, low-context prompts like 'it's not working.' This is counterproductive. A better approach is to use a specific command or prompt that forces the AI to reflect and reset its approach.