The Citrini selloff, the IBM session, and what both mean for your data

A 10% probability scenario moved capital

Citrini Research published a thought experiment in February 2026 and moved billions in capital within 48 hours. They framed it as a macro memo stress test, a speculative scenario from the year 2028 in which AI agents had commoditized entire categories of professional services. They assigned the scenario a 10% probability. Markets didn't care about the other 90%.

Bloomberg called it "the Citrini selloff." ServiceNow dropped more than 3%, DoorDash declined, and American Express fell because the memo described a mechanism portfolio managers could already see forming

Citrini's argument was specific. Machine intelligence reprices the value of human intelligence. When an AI agent can re-shop an insurance policy in four minutes, the information asymmetry that sustains insurance brokerage collapses. When an AI can assemble a tax return from raw documents, professional tax preparation loses its pricing power. When financial advice becomes something a model can generate from your actual data, the value proposition shifts from "access to expertise" to "access to your own information, organized."

Citrini gave the scenario low odds. But financial markets are probability-weighted machines, and even a 10% chance of that much disruption was worth repricing.

Why did IBM fall 13% in one session?

Separate from Citrini, IBM fell 13% after Anthropic announced Claude could work with COBOL code, and what happened in that single trading session is worth sitting with.

IBM spent decades building a services empire around COBOL, millions of lines of legacy code running global banking, insurance, and government infrastructure. The competitive advantage came from scarcity. COBOL developers are aging out of the workforce, and IBM had positioned itself as the indispensable partner for maintaining and modernizing those systems. Relationships built over 30 years, contracts renewed on institutional memory alone.

One product announcement suggested an AI model could read, understand, and work with that same code, and thirty years of institutional relationships repriced in an afternoon because the scarcity justifying the premium had an expiry date nobody priced in.

What did the Citrini memo actually say?

John Lober offered a useful rebuttal built around what he calls the iron rule of dealing with human reality: everything is always more complicated and takes much longer than you think.

He pointed to real estate brokers as evidence. People have predicted their obsolescence for 20 years, Zillow and Redfin and a dozen startups tried to cut them out, and brokers still collect 5 to 6% commissions on most residential transactions. Regulatory capture, emotional complexity, and accumulated customer habits keep incumbents in place long after their informational advantage disappears.

Fair point, and it applies more cleanly to some industries than others. DoorDash, which Citrini included as vulnerable, is a reasonable example because the moat there is a logistics network of restaurants, drivers, and delivery routes that no AI agent can replicate by being smarter. Citrini arguably overreached on that one.

Insurance brokerage, financial advisory, and tax preparation sit in a different category entirely. Their value propositions rest on information asymmetry. You pay a broker because they have read 40 policies and you have read zero. You pay a financial advisor because they can synthesize your holdings, tax situation, and risk tolerance into a recommendation. You pay a tax preparer because they understand which deductions apply to your specific situation. AI agents do not need to be perfect to compress those margins, they need to close the information gap enough that consumers start questioning the fee.

Citrini closed the memo with five words: "The canary is still alive." Maybe so, but canaries don’t leave the mine on their own.

Everyone's analyzing the macro layer and missing you

Citrini, Bloomberg, and the analysts parsing this selloff are all focused on the same question: which companies get disrupted. Nobody is asking the more interesting one, which individuals are prepared.

If AI agents will eventually re-shop your insurance, optimize your subscriptions, assemble your tax documents, and manage your financial records, those agents need access to your personal data, your insurance policies, financial statements, property records, family documents, medical records, and subscription contracts.

Right now that information sits scattered across Gmail inboxes, Google Drive folders, iCloud accounts, filing cabinets, and apps you signed up for three years ago and mostly forgot about. No single agent can reach all of it and no single platform holds a complete picture.

Citrini describes a world where AI agents commoditize professional advice, but commoditized advice still needs raw material. And the raw material is your data, fragmented across a dozen services extracting value from it in directions you never chose and mostly can’t see.

Intelligence infrastructure belongs to platforms, not people

Google reads your email to sell ads. Apple locks your data inside a closed ecosystem while positioning itself on privacy. Dropbox stores your files without understanding them. Every financial app, insurance comparison site, and document manager you use extracts value from your information through advertising, data aggregation, or vendor partnerships.

Nowhere in this arrangement does anyone build intelligence infrastructure that works just for you.

When Citrini imagines AI agents re-shopping your insurance, the unstated assumption is that some platform runs those agents, and that platform brings its own incentives, its own data practices, and its own reasons for steering you toward specific providers.

When the intelligence layer sits on top of someone else's platform, the value extraction continues and gets more sophisticated with every interaction.

What is personal intelligence infrastructure?

We are building MyVault around a premise that sounds simple but turns out to be structurally rare: ownership.

MyVault reads your insurance policies and identifies coverage gaps before you need to claim. It monitors contracts and flags renewal terms 90 days before they expire, tracks every subscription you are paying for and catches the ones you have forgotten, and connects financial records to property records to family documents into a knowledge graph of your actual life.

When you ask what your total insurance coverage is, or how much you are spending on subscriptions you no longer use, MyVault answers from your documents with sources, confidence levels, and reasoning you can follow.

Your passphrase generates the encryption keys and those keys never leave your device, which means we have no ability to access your information. 

MyVault's Insurance Agent reads your policies because they are stored in your vault, encrypted with your keys, with no partner carriers to steer you toward. Your Finance Agent spots the duplicate Hulu charge and the gym membership unused since October because that is what you asked it to do, and the intelligence it builds stays yours.

Preparation looks different than you'd expect

Citrini's memo focuses on corporate disruption, and that analysis is valuable, but the individual implication is just as significant and far less discussed.

If machine intelligence re-prices professional services over the next decade, the starting point is owning your own data infrastructure. Insurance policies, financial records, property documents, and family records organized, encrypted, and queryable by agents working exclusively for you.

Count the platforms holding pieces of your financial life right now. Your email provider, your cloud storage, your bank's app, your insurance company's portal, your investment platform. Each one holds a fragment, none of them talk to each other, and every one has terms of service granting usage rights you probably have not read.

Personal intelligence infrastructure consolidates those fragments into a single system you control. A vault.

MyVault's Private Beta opens in 2026. If you want to build that intelligence layer before the transition arrives, join the Founding Circle.

Markos is the founder and CEO of MyVault, an AI powered personal intelligence platform that reads, connects, and surfaces your scattered digital information, with zero-trust privacy architecture so your data never becomes someone else's asset.

FAQ

What is the Citrini Research report?

• In February 2026, Citrini Research published a speculative macro memo framed as a dispatch from June 2028. It modeled how abundant machine intelligence could trigger cascading disruption across white-collar work, consumer spending, SaaS, private credit, and mortgages. They assigned it a 10% probability, and markets repriced the same day it went viral.

What is personal intelligence infrastructure?

• It's a system that organizes, connects, and surfaces insights from your scattered documents while working exclusively for you. Unlike every other platform, there's no advertising model, no data aggregation, no vendor partnerships steering your decisions. Just intelligence built for your interests.

Who owns your data when AI processes it?

• When you use a general purpose AI tool, the platform typically retains usage rights to your inputs under its terms of service. Your data trains models that benefit millions of other users. MyVault uses zero-trust architecture where your passphrase generates encryption that never leaves your device, meaning we have no ability to read what you store.

What is a personal data vault?

• A personal data vault is an encrypted system where your documents, financial records, and personal information are stored under your own encryption keys. Unlike cloud storage, the vault operator has no ability to access your information, and the intelligence built from your data stays yours.

How does AI change personal data management?

• AI can now read, understand, and surface insights from documents that previously required professionals to interpret. Insurance policies, contracts, tax records, financial statements can be analyzed automatically, which means you don't need to wait for a broker or accountant to tell you what your own documents say.