A post about AI wiping out all knowledge work just hit 60 million views on X.

The core claim: AI can now handle tasks that take a human expert five hours.

That's technically true. At a 50% success rate.

The author left that part out.

He also left out that the benchmark behind the claim measures mostly software engineering tasks, not "all cognitive work" as he implies.

The METR research is real.

The trend is real.

AI capability IS doubling roughly every seven months.

Here's what matters for legal:

The gap between "AI can do this in a lab" and "AI is trusted to do this in a regulated industry" is measured in years, not months.

Liability frameworks. Institutional inertia. The need for human accountability when someone's freedom or fortune is on the line.

These aren't bugs.

They're features of a system built on trust.

The guy who wrote the post, Matt Shumer, has a history of sensationalism.

Most notably a model launch in 2024 that independent researchers couldn't replicate.

He later admitted he "got ahead of himself."

The truth about AI in legal sits between "nothing's changing" and "everything's over."

I meet with law firm leaders regularly....like multiple times a week regularly.

Some of these leaders tell me that they taking a "measured approach" (a.k.a. slow) to exploring tools and alternative service offerings.

"Measured approach" sounds sensible but I think it's a mistake.

If I'm being honest, I'm not sure if the firms moving slow are going to be here in 5 years.

Measured is too slow and reckless is too fast.

Finding the right pace is hard but necessary.

Don't panic but move with purpose.

There is no time for complacency.

The world is changing fast and you must too.