After shipping 100+ AI agents to production, we learned that making them reliable isn't about better LLMs—it's about better architecture.

The solution: task-driven design. Give agents mandatory todo lists with validation gates. Each task needs clear completion criteria and evidence. Agents cannot finish until every todo is verified complete.

This simple pattern prevents 80% of production failures. Structure beats intelligence.

see it in action at justcopy.ai

Based on the use of em dashes and general tone, this post and this comment were both written with AI assistance.

That’s a lot of confidence with a vanishingly small amount of evidence. I am constantly accused of being an AI for articles I blogged more than 20 years ago.

I use—still—em dashes whenever I feel like they are better than commas, parentheses or semicolons.

Because that’s what they are for, and this is English after all. Some people don’t approve, and I approve of their disapproval—it’s a free country and writing is a creative process.

What doesn’t contribute are comments that discredit and disparage without adding anything to the conversation.

Do you think that this post, by an AI company, talking about workflows for AI tools, which is structured into nested headings and bullet points, which uses the red X and green check emojis, as well as the em dash, was not written with the assistance of AI tools?

I think it contributes when people generally disparage AI and the people who abuse it. This is one way that we build up a social awareness that faking your work with AI is not okay.

accussing people that their content is AI generated without merit will do the exact opposite of this