CEO's vibe coding' 👏👏

Here's the post I put on LinkedIn on "Chain of Debate" a while back: Chain of thought was just chapter one. Next up: Chain of Debate. When AIs started sharing their chain of thought (CoT), it gave us a window into its thinking – how it reasoned its way from A to B. But at the end of the day, it’s just talking to itself. I think we’re about to see AI become a team sport instead. ❌ a single model serving up an answer ✅ a huge range of models debating, debugging, and improving each other's CoTs. Imagine a kind of AI council meeting to solve problems or find answers for the user. After all, this is how progress gets made in our world! Debate, discussion, bouncing ideas off others and getting feedback. Amazingly it seems true of AI as well. This inter-model collaboration and accountability should help boost trust, accuracy, and quality – pooling models’ strengths while mitigating each other’s limitations or biases. Think of it as the AI version of peer review. But what about the elephant in the room: do we really want an LLM improving other LLMs? Isn’t recursive self-improvement a slippery AI slope? That’s where the transparency of debating “out loud” really matters and makes a difference. We aren’t powerlessly watching a black box. Even as AIs influence and steer each other, the user can see exactly what that influence is – and intervene if needed. “Change their minds” so to speak. There’s that saying “two minds are better than one,” and I feel the same about AIs pushing each other to be better, more accurate, more helpful. “AI” is becoming “AIs.” We’re entering an era where models don’t just think out loud, they think together—and by extension, better.