Published on February 8, 2026 8:07 PM GMTA worked example of an idea from physics that I think is underappreciated as a general thinking tool: no measurement is meaningful unless it’s stable under perturbations you can’t observe. The fix is to replace binary questions (“is this a degree-3 polynomial?”, “is this a minimum?”) with quantitative ones at a stated scale. Applications to loss landscapes and modularity at the end.
Discuss Read More
Every Measurement Has a Scale
Published on February 8, 2026 8:07 PM GMTA worked example of an idea from physics that I think is underappreciated as a general thinking tool: no measurement is meaningful unless it’s stable under perturbations you can’t observe. The fix is to replace binary questions (“is this a degree-3 polynomial?”, “is this a minimum?”) with quantitative ones at a stated scale. Applications to loss landscapes and modularity at the end.
Discuss Read More
Related Posts
[Paper] Output Supervision Can Obfuscate the CoT
Published on November 20, 2025 10:41 PM GMTWe show that training against a monitor that…
PrincInt (PIBBSS) Opportunities: Summer Fellowship, Postdoc, and Ops Role (Deadlines in January)
Published on December 29, 2025 12:12 PM GMTThree opportunities from Principles of Intelligence (PrincInt, formerly…
Manifold spin off MNX, a real money decentralized market for AI-related assets. Includes levered prediction markets, perpetual futures
Published on February 18, 2026 10:36 PM GMTTrillions of dollars are flowing into AI, and…