Great piece. Had a little lightbulb. I’ve been developing a theory that models cognition as an autoregressive process: the mind continually generates its next state—thought, perception, or action—based on a structured, residual context of prior states. “Memory”, in this view, isn’t about retrieving static past events but about carrying forward a dynamically evolving internal context that conditions what comes next.
Reading your piece, it struck me that it might be fruitful to think about biological systems as containing a similar form of memory—not just as systems that transition from one state to the next, but as generative engines whose future outputs are conditioned on internalized traces of their past.
There seem to be examples that support this view: epigenetic modifications, transcriptional feedback loops, immune memory, and circadian regulation all suggest that biological systems carry and integrate prior information in ways that shape their trajectories. This opens the door to modeling them not just as dynamic systems, but as context-conditioned generative systems—something closer to what we might call true autoregression.
Does this make any sense? Would love to hear your thoughts.
Hard agree
https://open.substack.com/pub/afederation/p/the-software-layer-of-life?r=ox9g8&utm_medium=ios
Love this! Very poetic.
Woah Thanks man!!!!!
Great piece. Had a little lightbulb. I’ve been developing a theory that models cognition as an autoregressive process: the mind continually generates its next state—thought, perception, or action—based on a structured, residual context of prior states. “Memory”, in this view, isn’t about retrieving static past events but about carrying forward a dynamically evolving internal context that conditions what comes next.
Reading your piece, it struck me that it might be fruitful to think about biological systems as containing a similar form of memory—not just as systems that transition from one state to the next, but as generative engines whose future outputs are conditioned on internalized traces of their past.
There seem to be examples that support this view: epigenetic modifications, transcriptional feedback loops, immune memory, and circadian regulation all suggest that biological systems carry and integrate prior information in ways that shape their trajectories. This opens the door to modeling them not just as dynamic systems, but as context-conditioned generative systems—something closer to what we might call true autoregression.
Does this make any sense? Would love to hear your thoughts.