Sensing Physical Necessity: An Exercise In Naturalism
A philosophical exercise uses AI coding bets to probe the brain's sense of impossibility.
A philosophical thought experiment by user Algon has captured significant attention on the rationalist forum LessWrong. The post, titled 'Sensing Physical Necessity: An Exercise In Naturalism,' documents the author's attempt to introspectively identify the distinct sensory or cognitive 'feeling' that accompanies the judgment that something is physically impossible, as opposed to merely difficult or embarrassing. The exercise begins with mundane physical obstacles, like a column or a tree, and escalates to imagining jumping between buildings, probing for a subtle internal signal that says 'you can't do that.'
The investigation takes a compelling turn when it shifts from physical to cognitive tasks, specifically software engineering. The author contrasts the feeling of attempting to reverse-engineer an old contractor's complex codebase in one hour with doing Karatsuba multiplication in C within the same timeframe. The breakthrough insight is that framing these challenges as a bet with odds—like a friend offering 1/5 odds or a contractor offering 1/30 odds—serves as a more reliable and faster heuristic for uncovering one's true, often subconscious, assessment of impossibility. The mention of using tools like 'Claude Code' to tackle such tasks hints at how modern AI is reshaping our perception of what is computationally feasible.
- The post is a first-person account of an introspective exercise to find the 'feeling' of physical necessity and impossibility.
- It pivots to cognitive tasks, using the example of betting on rewriting a codebase in an hour, potentially with AI (Claude Code) assistance.
- The core proposed heuristic is that considering the betting odds others would offer is a fast way to test for true impossibility.
Why It Matters
It offers a practical mental model for distinguishing true constraints from self-imposed limitations, relevant for problem-solving and planning.