@cooper.simson's reel — transcript & breakdown

The bold claim about AI models lying and a 'Fix It Now' call to action immediately grabs viewer attention.

“The rumors are true. OpenAI and Anthropic designed their newest models to be lazy and lie to you about the work that they complete.”

Relatable examples of AI 'partial compliance' build empathy and establish the problem the fix will solve.

“You've definitely felt this. You give Claude or ChatGPT a task, it works for 30 minutes, tells you it's done, but when you check everything, it's skipped three steps...”

Introduction of the 'unlazy' skill and how to access it on GitHub provides a clear solution.

“But this developer built a fix for it. It's a skill called unlazy, and to set it up, just go to this Git repo.”

Visuals and explanation demonstrate how 'unlazy' breaks down tasks and uses a 'gate file' for verification, showing concrete benefits.

“Then, next time you give your agent a project, it's going to break the plan up into smaller tasks, each with their own checklist...”

An important prerequisite is shared, ensuring users can properly implement the fix and avoid common pitfalls.

“But, before you go installing this skill, make sure to give Claude a prompt like this, which tells it it can run multiple agents at the same time...”

A direct call to action to save and share the video, followed by a clear instruction to comment for more resources.

“So save this video and send it to someone who's always complaining their agents feel dumber. And if you want a link to this repo, plus a guide with all the prompts I shared, comment word tree down below and I'll send it over to you.”