IBM lost a quarter of its value in a single session this week. The headline is an earnings miss.

The real signal is one level deeper. IBM's own management said it plainly: enterprise clients are moving spend away from software and legacy infrastructure, toward AI infrastructure.

This is not an IBM story. It's a capital reallocation story — and IBM is just where it showed up loud enough to move the stock 25% in a day.

The same week, semiconductor names were swinging double digits. SK Hynix, Micron, Sandisk — all moving on the same underlying question: where does that displaced enterprise budget actually land?

We ran both through Kubera.

The Researcher agent pulled IBM's filing, the earnings call commentary, and the semiconductor movement from the same week.

Then the Analyst agent did something a separate analyst covering each sector would not have done automatically.

It mapped both events onto a single standing thread — enterprise IT budget reallocation — that Kubera had been tracking across prior research runs.

Neither the IBM miss nor the chip volatility triggered a new research project. They extended one that was already open, with the earlier evidence intact and the timeline connected. Two desks. Two events. One thesis. Linked automatically.

Nobody had to remember to check. Nobody had to search across coverage silos.

The system held the context so the analysts didn't have to.

Every investment team rotates analysts. Every coverage universe accumulates thousands of observations across dozens of people over years.

The question isn't whether context gets lost. It is how much — and how often your team rebuilds analysis that already exists in a file no one thought to look for.

Most firms have two analysts covering adjacent sectors, each writing excellent notes, neither aware the other's work exists. Not because of bad process. Because there is no structure that forces related research to stay connected across people, across time, across rotations.

A shared drive stores files. It doesn't connect them. Kubera keeps research in a graph — where every observation connects to the theses, sectors, and entities it actually touches.

When the next data point arrives, the system traverses to what's already known and surfaces it without being asked. Six months from now, if a second legacy software vendor misses on the same budget shift dynamic, your team won't start from scratch.

They'll extend a thesis that already has its reasoning built in.

That's what compounding research looks like.

✓ Research compounds instead of resetting.

✓ Connected knowledge that surfaces patterns before they're obvious.

✓ Institutional knowledge survives analyst rotations.

✓ New events automatically extend existing investment theses.

✓ Your team's reasoning becomes an institutional asset, not an individual's memory.

I'm curious to hear from you.

Read more about Kubera here.

— Santhosh
Founder, Om Labs · Kubera