@vgr

In @yak study group recent calls about Claude code, @anurajenp used 2 illuminating metaphors: a) It has a flywheel effect on doing stuff b) It feels like loading up a dragster (with config info) and flooring the accelerator Both suggest a particular interpretation in OODA loop terms: The observe/decide/act (O_ODA) parts have been disproportionately accelerated compared to “orient” This means the **mean free path** of decision loops between reorientations has increased. This is like liquid evaporating to gas. The molecules move faster and longer and collide and change direction relatively less. This is a similar effect to ZIRP funding era in VC. “Free capital” led to overcapitalized startups and overleveraged early risks. Since in OODA learning only happens during reorientation, you end up with more bet on each unit of learning. Capital too cheap to meter = startups that grow faster than they learn. Intelligence too cheap to meter = projects that grow faster than they learn (in OODA sense which is also the “science” sense of falsification etc). Not sure what to go with this thought yet.

5 replies

1 recast

11 reactions