The great AI race has focused on which company can build the smartest, fastest, most capable model. That's understandable. But while much of the industry was asking how to build a more powerful model, I got interested in a different question: who challenges it?
As these systems get more sophisticated, their answers get more persuasive. That's the problem. Confidence is not certainty. Eloquence is not proof. And no matter how advanced a model becomes, its conclusions should stay open to challenge.
That question led to Bridget; an AI platform built on a different premise: competing models arguing against each other rather than one system delivering an answer. A single model's answer is one opinion presented as an answer. Bridget makes models argue and shows you where they disagree.
Most AI tools are built to sound confident. We built ours to show its seams. When our models can't reach agreement, the output says so – that's a finding, not a failure. Every result carries its full audit trail: which models, which prompt, how many rounds, every rebuttal and appeal.
You can check our work.
None of this means competing models produce truth. They don't. AI can still be wrong. Multiple systems can share blind spots. Human judgment remains indispensable.
That's the point. The future shouldn't be humans surrendering judgment to increasingly powerful machines. It should be humans using competing intelligences to challenge assumptions, expose weaknesses, and decide better.
Which is why Mark Zuckerberg's recent essay is worth reading closely.
In "The Future is for Everyone," he argues that AI's future shouldn't be controlled by any single company, government, institution, or intelligence – that power should be distributed, that competing interests should check one another, that individuals should stay in control. He writes about a balance of power in AI, and about multiple frontier labs building systems with different values that check each other.
Checks and balances only work if the parties are genuinely independent and the record is open. Structure matters more than intent. Zuckerberg is making that argument at the scale of the industry. We're making it inside a single query.
The next phase of AI shouldn't only be about building more powerful intelligence. It should be about building structures capable of challenging it.
