Morgan Stanley predicted a "shock-level AI breakthrough" between April and June 2026.

They published that on March 13th.

Thirteen days later, Anthropic accidentally leaked Mythos.

Anthropic confirmed it's real. Called it "a step change" and "the most capable we've built to date."

The model is so advanced that Anthropic has been privately briefing top government officials about its cybersecurity risks.

Goldman Sachs' CIO said 2025 saw the biggest changes in his 40 years in technology.

And predicted 2026 would be even bigger.

Neither institution pointed to a specific model.

But both were tracking compute buildout at the major AI labs.

They knew something was coming.

Now we might know what it is.

At Morgan Stanley's TMT conference, a panelist said one recently introduced model "broke the trend in a big way to the upside" on AI task complexity benchmarks.

If Mythos is that model, we're not on the curve anymore.

We're above it.

I talk to law firm leaders multiple times a week.

Many are still taking a "measured approach" to AI.

Measured looks very different when the curve bends upward.