The Dark Arts of Skill Engineering — Paul Bakaus, Renaissance Geek (Impeccable)

By ai.engineer

Categories: AI, Tools

Summary

Prompting alone can't overcome AI's inherent biases—Paul Bakaus reveals that 250 lines of crafted prompts fail to change the model's 'gravity,' and the real solution is harness engineering: treating skills as MCP extensions rather than packaged prompts to systematically eliminate AI slop.

Key Takeaways

  1. Bans in prompts cause displacement, not elimination—telling Claude not to use Inter or purple gradients just pushes it to the next cluster in latent space. The solution requires systematic harness changes, not negative instruction.
  2. Slop is a moving target that evolves with design trends. Purple gradients and Tailwind defaults dominated early AI generation; now it's 'Claw beige' aesthetics. Monitoring and adapting constraints is critical.
  3. Harness engineering beats pure prompting: reframe skill-building as MCP extensions to your coding environment, not standalone prompts. This architectural shift enables systematic design system normalization at scale.
  4. Impeccable started as a single 'normalize' skill to map Claude's designs back to design systems, then expanded into a suite of design-focused skills. Document this evolution pattern—constraints breed tooling.
  5. The 'claw beige' aesthetic problem emerged because unconstrained AI generation converges on high-probability outputs in latent space. 55-line prompt-only approaches lack the control needed for production design systems.

Related topics

Transcript Excerpt

Hello everybody. How's it going? Okay, I think we still have some people trickling in, but I'm super excited to be here. Okay, first off, let me reload these slides because my claw code was still building something on it. Um, okay. So, the contrast is a little bit low. Um, so please bear with me. I'm going to try to cover uh what you cannot read as much as possible. Uh, I also have heard that the Wi-Fi is not the strongest. So um while it is a workshop uh hopefully you just do take away a lot of the lessons um and then can apply it whenever you want to. Um but I do have a sample repo if you want to follow along. Um okay first of all hi my name is Paul. Um I'm really glad you found your way into this room. Um I'm the author of Impeccable. Who here has used Impeccable by any chance? Can I se…

More from ai.engineer