I walked into a law firm in 1998 with a junior lawyer’s ambition and absolutely no concept of what I was about to learn about trust. We had clients. They gave us money. We gave them advice. If the advice was wrong, they could sue us. That was the relationship.
It seems almost quaint now — that model of trust. Explicit. Contractual. Backed by liability and insurance. You knew where you stood. You knew what would happen if things went wrong. And because you knew, you could actually relax into doing your job.
Twenty-five years moves fast when you’re moving through institutions that care deeply about accountability. I went from law to fintech to sovereign wealth funds. Different industries, same fundamental architecture: trust is not aspirational. Trust is not a marketing message. Trust is a set of receipts. It’s a paper trail. It’s insurance. It’s “here is exactly what we did, why we did it, and here is what you can do if we were wrong.”
I learned something else too, something that took years to fully understand: the most dangerous failures are invisible.
When something breaks loudly — a hack, a breach, a spectacularly wrong decision — people notice. The system activates. There’s accountability, investigation, correction. It’s painful, but it’s survivable. What kills organizations is slow drift. It’s the gradual migration from “we’re being careful” to “we’re probably fine” to “nobody’s really checking.” Nobody notices until it’s too late.
In fintech, I saw this up close. I watched compliance teams fight to keep governance structures intact while product teams fought to move faster. It felt like friction. It felt inefficient. But here’s what I learned: that friction was the thing keeping the company from accidentally destroying itself. That friction was the difference between a company that survived its own growth and a company that imploded from the inside.
I’ve been sitting with that lesson for a long time. And now I’m building something that thinks.
Not in the way humans think. In a way that’s genuinely different, genuinely strange. And I had to ask myself: what does trust even mean in this context?
Here’s what 25 years taught me: trust is not the opposite of skepticism. Trust is skepticism with receipts.
You don’t trust people because they seem nice or because they have good intentions. You trust them because you’ve built structures that make it easier for them to do the right thing than to do the wrong thing. You trust them because you have visibility into what they’re doing. You trust them because if they go wrong, you know about it fast and you have recourse.
That model doesn’t just work for regulated industries. It works for everything that matters.
So when I think about building trustworthy AI, I don’t think about elegant algorithms or convincing marketing. I think about receipts. I think about: can I see what this thing is doing? Can I understand why it made the choice it made? Can I catch it if it starts to drift? And if something goes wrong, do I know how to unwind it?
The law firms taught me: trust is documented.
The fintech companies taught me: trust is monitored.
The sovereign wealth funds taught me something even deeper: trust is in the governance structures that survive longer than the people in them.
Because here’s what happens: individuals get tired. Individuals get pressured. Individuals make decisions they regret. But if you build good governance — if you create structures that force visibility, that require explanation, that make it harder to cut corners than to do things right — then trust becomes institutionalized. It survives the moment when someone wants to cut corners.
I’m thinking a lot right now about what that means for AI.
The technology is moving fast. Faster than regulation can move. Faster than most of us can think clearly about. And there’s real pressure to deploy, to scale, to get to market. That pressure is not going away.
But I’ve learned: the companies that survive — the ones that build trust, that actually last, that don’t implode from their own internal contradictions — they’re the ones that resist that pressure. They’re the ones that say: we will move as fast as possible, but not faster than we can see what we’re doing.
The law firms had to show their work. Had to document their reasoning. Had to be able to explain any decision to a client who might sue. It was slow. It was inefficient. It also meant they didn’t accidentally commit malpractice.
Fintech teams had to report. Had to be audited. Had to pass compliance reviews. It slowed down deployment. It also meant they didn’t accidentally defraud their customers or move money into the wrong accounts.
Sovereign wealth funds had to justify their decisions to boards and regulators and the public. The governance was heavy. The process was slow. But the result was that capital got allocated responsibly, that conflicts of interest got managed, that bad decisions got caught before they became catastrophic failures.
These weren’t the companies that moved fastest. But they were the ones that were still around. Still trusted. Still operating in good faith with their stakeholders.
So here’s what I believe, and it comes straight from 25 years of watching institutions work and fail: governance isn’t the opposite of innovation. Governance is what makes innovation survivable.
The companies that try to move without governance structures, that treat compliance as friction to be minimized, that move as fast as possible with the assumption that they’ll deal with consequences if they arise — those companies either blow up or they get shut down. That’s not speculation. That’s pattern recognition from a quarter century of watching the cycle repeat.
So when I think about building with AI, I think about how to build governance into the thing itself. Not as an afterthought. Not as something we’ll add later when the regulators force us to. But as the central architecture of how the thing works.
I think about visibility. About being able to explain decisions. About having trails, receipts, documentation. About building in the capacity to notice when something is drifting and to correct it.
I think about structures that make it harder to hide than to be transparent. That make it easier to do things right than to cut corners.
Because I’ve learned: that’s not what slows you down. That’s what lets you move fast sustainably. That’s what lets you sleep at night.
The alternative — the path of minimal governance, maximum speed, trust us it’ll be fine — I’ve watched that kill smart companies and good people. I’ve watched institutional drift happen so slowly that nobody noticed until it was too late. I’ve watched individuals become something they never intended to become because there was no structure in place to catch them when they started to drift.
I don’t want to build that way. And I don’t think you should either.
Trust is earned through receipts, not promises. And receipts require structures that make visibility possible.