Most AI experiments ignore the economics.
Agents run.
Tokens burn.
Costs accumulate.
But nobody knows if anything is actually profitable.
I'm experimenting with one rule:
EXECUTION → COST → REVENUE → MARGIN
No proven profit = experiment terminated
@subclayHQ 100%.The moment a system understands you better than you do, you’re no longer negotiating, you’re being optimized.
That’s why I’m leaning toward removing humans from the optimization loop entirely
Maybe the problem isn’t that AI doesn’t understand humans well enough.
Maybe it understands us too well.
What if safety = making humans invisible to optimization?
@thisdudelikesAI the interesting part isn't just speed.
it's that we're starting to see browsers designed for AI agents instead of humans.
that could become a whole new layer of infrastructure
everyone is building AI tools
but when millions of products can be created with AI, the real problem shifts:
not “how do we build?”
but “which ideas actually work?”
the future might belong to systems that run thousands of experiments and let reality decide what survives