Synthetic Competence & The Rise of Prototype Intelligence

AI may massively increase the appearance of intelligence before increasing intelligence itself.

We may be entering an era of Synthetic Competence.

I see people demoing agent swarms, Claude Code, autonomous coding loops. Very impressive. Often on codebases roughly the size of a respectable sandwich.

Then you ask:

• How do you prevent code drift?
• How do you keep agents sane in non-toy systems?
• How does this behave in legacy architecture with 12 years of regret?

Silence. Or a diagram. Usually a very confident diagram.

Some agent demos are just prompt engineering wearing a tuxedo. I call this Prototype Intelligence — things that look profound in demos and become philosophical in production.

There may be serious competence inflation underway:

1. More signals of expertise.
2. Fewer encounters with complexity.

We may be automating confidence faster than judgment.

Quotable thought: AI may not expose weak coders first. It may expose shallow thinkers.

Because real engineering begins exactly where demos stop. And when everyone sounds smart, bullshit detection becomes a strategic capability. Possibly a premium one.