How I stopped shopping for different models and started designing different tasks to achieve true functional independence.
Or: How I stopped shopping for different models and started designing different tasks.
Before: Two models answering the same prompt → phi correlation = +0.42 → ensemble behaved like 1.7 independent judges.
After: One answers, the other attacks the verdict → phi correlation = -0.80 → true functional independence.
No third model. No different vendor. Just a different job description.
Model A (Participant): Groq/Llama 3.1 8B – answers the original prompt.
Or: How I learned that "independent validators" are like siblings – they share the same...
Like a lot of developers, I’ve integrated AI heavily into my daily workflow. But recently, I ran into...
How I work with AI, and why the system around the model matters more than the model.
One AI Vendor Is a Single Point of Failure. Treat It Like One. The AI model you built your...
What I actually found when I set out to test heterogeneous AI code review. For the last couple of...
I used to have one AI assistant. It did everything — coded, wrote docs, answered questions, monitored...
