AI, Jobs, and the Quiet Compression of 2026

UBI feels less urgent to me than it did 24 months ago.

The 2025 data just doesn’t support panic.

The U.S. economy grew strongly: GDP +3.8% in Q2 and +4.4% in Q3. Yet payroll employment grew only 584,000 jobs. A dual-method estimate suggests 200,000 to 300,000 jobs were displaced or never created due to AI in 2025.

That’s not mass unemployment. But it’s not nothing either.

And here’s the key: it wasn’t mass AI layoffs. It was hiring suppression. Companies didn’t fire people in waves. They stopped backfilling. They slowed hiring. They absorbed productivity gains quietly.

That’s why UBI feels less urgent today.

But 2025 was the year of copilots. Drafting. Summarizing. Coding help. Task-level productivity. Copilots compress margins. They don’t remove layers.

2026 may be different.

Anthropic’s Claude Opus 4.6 is being called an “AGI moment” by some observers. That might be exaggerated. But the reliability jump is real.

Across models we’re seeing:

• Output per prompt increasing significantly

• Error rates in structured workflows dropping 50%+

• Multi-step task chaining becoming viable

• Agents operating across tools and systems

Internal enterprise tests suggest gains approaching 5x output per prompt in some structured environments.

That’s not just task efficiency. That’s coordination compression. And that’s where roles start getting squeezed.

The job numbers won’t show this immediately. You’ll see it first in fewer junior openings, wider spans of control, and higher output per mid-level employee.

A labor market can look stable while structurally thinning.

So yes, UBI feels less urgent. But if agent-level automation scales this year, should it?

I discussed this in detail with Dalibor Petrovic, who works on post-labor economic modeling alongside David Shapiro.

Full conversation here:

https://youtu.be/4KWzvRL8n9o

Are agents a 2026 inflection point? Or are we still in the productivity phase?

Let’s discuss.

0 comments