Vector Safety Net
Subscribe
Sign in
Home
Notes
Archive
About
The Game Behind the Agents
The greater risk is from the people building and deploying AI. Not from AI going rogue on its own.
Sep 25
•
Makho Mashoba
The Herd Shares a Genome: Why Rare Situations Matter for AI Safety
When every session runs the same model, a rare failure is a shared vulnerability
Sep 20
•
Makho Mashoba
Teaching AI to Work Well With People
A case for building a business around reliable measures of human–AI cooperation.
Sep 20
•
Makho Mashoba
AI Cooperation: Who Benefits
What the evidence shows, why reputation needs a self, and why the reward has to count outsiders
Sep 19
•
Makho Mashoba
2
LLMs Aren’t “The Best” at Game Theory: Rent the Model, Own the Decision
Building a business that helps AI agents negotiate reliably—and creates useful evidence for training more cooperative models.
Sep 19
•
Makho Mashoba
Why AI Safety Can Swing Between Caution and Speed
What “win-stay, lose-shift” suggests about AI development, safety commitments, and the incentives shaping both.
Sep 19
•
Makho Mashoba
1
How Measuring Human–AI Cooperation Could Change What Labs Optimize
A proposal for making human–AI cooperation a measured, marketable property of model quality.
Sep 19
•
Makho Mashoba
Training AI Agents Beyond the Answer Key
Real interactions, feedback from affected parties, and reputation as training signals.
Sep 19
•
Makho Mashoba
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts
