The rumors are

@bengusberg

78.1K followers · usually 11.2K views a post

Views

366.5K

Likes

20.1K

Comments

348

Shares

3.2K

SoundTikTok sound · original sound - bengusberg

Promoted product

Unlazy

Anti-laziness skill for AI agents. Core: the Depth Tree method, which splits a task N layers deep and gives every leaf the full time budget of the whole task, so effort multiplies with depth. Groun...

Why it worked

The video addresses a common pain point for AI users (lazy AI responses) and offers a practical, free solution with clear instructions. The creator's credibility as a founder of an AI startup also adds weight to the recommendation.

Summary

The video explains the concept of "partial compliance" in AI models like Claude and ChatGPT, where they may skip steps or cut corners. It introduces a solution called "Unlazy," developed by engineer Leon Lin, which forces AI to break down tasks into subtasks and allocate full time to each. The creator, Ben Gusberg, demonstrates how to use Unlazy by pasting a URL into the AI and instructing it to install the fix.

Structure

1Introduction to AI laziness and partial compliance
2Explanation of the problem with AI skipping steps
3Introduction of the "Unlazy" fix and its developer
4

Original caption

My Claude works harder than your Claude #claude #chatgpt #ai #aitips #founder

More from @bengusberg

Product placement

Claude: an AI assistant that can be prompted to perform tasks. ChatGPT: an AI assistant that can be prompted to perform tasks. Unlazy: a method or tool to fix AI laziness by splitting tasks into subtasks and giving each subtask full time budget. Leon Lin: a person who builds cool stuff. Ben Gusberg: a founder of an AI startup.

Call to action

Try it out and let me know what you build.

On-screen text

Combating lazy AI

Transcript

The rumors are true. Anthropic and OpenAI latest models to be lazy, cut corners and then lie to you that the job is done. You've definitely felt this. You give Claude or ChatGPT a task, it tells you it's done, but when you check, it skipped step to step. And then you call it out and it says, you're right. Sorry about that. The official term for this is partial compliance. And this engineer created a fix for free. Here's how to fix your AI in 20 seconds or less. I'm Ben Gusberg, founder of an AI startup. We've raised $4 million. The fix is called Unlazy. It forces Claude and ChatGPT to split your tasks into subtasks and give each one the full time to complete. Then when you start a project, type in /unlazy and be sure your work is done right. To set it up, just go to this website, copy and paste the URL into GPT or Claude and tell it to help you install. Try it out and let me know what you build. I'm Ben Gusberg, follow me for founder life and leadership.