A few weeks ago I imposed a discipline on myself that cost nothing and changed how I think about investment research.

I was building a platform designed to find investment opportunities at the intersection of independent analytical frameworks — companies sitting at the convergence of multiple structural forces simultaneously.

The temptation, when building something like this, is to design those intersections. To know what you're looking for before you start looking.

I resisted that.

The Rule

Each investment theme is seeded by an independent analytical session, with no knowledge of what any other session has done.

The discovery engine — and only the discovery engine — is permitted to look across them.

That sounds like a small design decision. In practice, it determines whether a platform discovers ideas or merely confirms existing beliefs.

What It Produced

Several sessions in, building a Defence & Aerospace investment universe, I tagged Palantir as a primary expression of the defence intelligence theme.

The AI Infrastructure universe had been built separately, weeks earlier.

When the engine ran, Palantir appeared at the intersection — flagged because two completely independent lines of analytical reasoning had arrived at the same company, for entirely different reasons.

That is what joining the dots actually looks like. Not designing connections in advance. Not confirming what you already believe. Two independent analytical processes, neither aware of the other, converging on the same vehicle.

The Part That Compounds

Most investment research is flat. Each analysis starts from scratch. The reasoning that led to a view is not captured. The condition that would justify revisiting a rejected idea is never written down.

The platform I've built remembers. Every decision is logged with three fields: what was asked, what was decided, and the specific evidence that would cause the decision to be revisited.

Not "I'd look at it again if things change." But a specific, testable condition: I would reconsider if there is a significant fall in forward earnings estimates.

The decisions made this week narrow the search space next week. The rejections recorded today define the watchlist for next year.

In thirty years of investment management, I never had a research tool that remembered what I had decided and why.

Now I do.

That, for me, is the real promise of AI in investment research — not replacing judgement, but building infrastructure where good judgement compounds over time.