iiinova — where your digital self keeps evolving

And, I’d like a closer inspection of “how are memories stable”…to consider whether they are. Maybe “the tension between stability and flexibility of memories”

030

As for product, considering biology more broadly, allostasis (calibrated dynamics) and metabolism, as contenders for “goals”

Yeah - they can/should be separated into process and product, was just compressing b/c we keep them together atm. Taking BTSP as example alternate process, I wonder what others we may be missing, possibly at a higher-order scale, drawing from cell heterogeneity.

🙌 ❤️🔥Yes to all the things. The 1st and 4th points may be related. Less about minimizing a reward error term and more about maximizing efficiency, allostasis.

140

Matteo Carandini @carandinilab.net · 31/07/2026

There is nothing either good or bad, but joking makes it so - Bob untucks humor #bb28

050

Reposted by Kari Hoffman

The apical dendrite looks great, and does wonders in a slice. What does it do in the living cortex? It depends on the cell. A causal spatial role for apical dendrites in the cortex www.biorxiv.org/content/10.6... by @anyiliu.bsky.social with @kenneth-harris.bsky.social and @fedrossi.bsky.social

A cartoon providing a graphical summary of the paper. It is divided in 3 rows.

The top row shows a pyramidal cell and its apical tuft, and shows the two experiments that we ran: optical dendritic pruning (chopping off the apical tuft), and synaptic imaging done in the same neurons where we do somatic imaging.

The middle row shows the results pertaining to neurons that prefer large stimuli (this is visual cortex): pruning reduced their responses, and the receptive fields of apical synapses are far away from the that of the soma.

The bottom row shows the results pertaining to neurons that prefer small stimuli: pruning seemed to do nothing to the responses, and the receptive field of apical synapses are right around that of the soma.

17427

Reposted by Kari Hoffman

Mark Histed @markhisted.org · 31/07/2026

Also, science needs to care about journalism too! The collapse of the US journalism industry (which is related to conspiracies and fractionation but not the only cause) is also a reason we are here.

2124

Kari Hoffman @karihoffman.bsky.social · 31/07/2026

Thanks, yes, I’d love to know what structure old memories impose. In particular, whether old -> new structural similarity between memoranda matters to scaffolding (compositionality) and how scaffolding differs and (dynamically?) aligns across brain areas. So much to do!

000

This took tenacious effort from @sabbaspoor.bsky.social, and props to @aymanaljishi.bsky.social who assisted and carries the torch. Thanks to the #BRAINinitiative, #NINDS, and Whitehall Foundation for essential funding & Diagnostic Biochips, Cambridge Neurotech and Neuralynx...& the reviewers! fine

140

Our findings suggest that memory consolidation could do more than stabilize isolated traces (or 'engrams'). Experience progressively links, reorganizes, and integrates neural assemblies allowing the primate brain to learn new information while preserving established memories. 11/n

Because we had laminar cell resolution, we could explore CA1 microstructure doi.org/10.1016/j.ce.... Superficial pyramidal cells contributed more strongly to new assemblies and predicted greater future drift. As memories aged, superficial + deep CA1 neurons became increasingly coordinated. 10/n

To more directly test this idea, we measured the local cohesiveness of assembly members. Cohesiveness increased with memory age: new < recent < old, suggesting that experience progressively embeds assemblies within more interconnected and potentially more stable networks. 9/n

graph connectedness of assembly members increased from new to recent to old

“Drifting assemblies” = gradual change in activation strength during the task, was greater for New assemblies. Was this drift due to member cells "competing" across fellow assemblies? No! The more exclusive an assembly's members, the greater the drift, i.e. integration->stabilization w/ time. 8/n

It looks like experience changes the likelihood and duration of coordination among assemblies, at least in this cross-sectional approach (Old/New). Inspired by @denisejcai.bsky.social, we wondered if experience-dependent change would also appear as gradual drift in task-related activation? 7/n

130

The old metassemblies don't just co-activate more, they coordinate for LONGER ⏱️ Markov modeling and cross-correlograms show mutual reinforcement: half-widths can extended 500ms+, beyond the SWR window we’d assumed. --> The temporal reach of the metassembly groups grows with experience. 6/n

schematic from the paper's HMM showing transitions from baseline and inactive states to the active state. The activation from one assembly to another of the "metassemblies" was stronger for old then new metassemblies. Below, the crosscorrelograms shows that strong 'instantaneous' pairs of old metassemblies are more sustained in time than new metassemblies.

Old assemblies don't just reactivate more, they are more organized. When pairs of old assemblies reactivate together (or in opposition) during the task, the pattern is preserved in sleep; New assembly pairs also have such 'Metassemblies' in the task, but they aren't preserved in sleep. 5/n

Metassemblies - pairwise activation of assemblies - shown sorted for task and again in sleep. Old metassemblies were preserved between task and sleep.

👀In sleep + SWRs, Old assemblies reactivated more strongly than New assemblies! Perhaps sleep reactivation isn't just benefiting consolidation of what's *NEW*, but also helps restructure or reinforce what's already learned. We further examined this idea of structuring of assemblies over time. 4/n

Distinct patterns of population activity (cell assemblies) represented the different sequence types (New/Old), and were better decoded than single-unit activity. --> The populations carried memory-specific information during the task. But what about sleep? Sharp-wave ripples? Reactivation? 3/n

from the population activity covariance matrix, we find cell assemblies - population cofiring motifs - some of which occur preferentially during new-sequences or during old-sequences.

130

The setup: macaques learned multi-touchscreen sequences in an immersive enclosure: a "new" or "recent" (1-day old) sequence on one side and an "old" sequence (learned ~6 weeks' prior) on the other, while we recorded from the hippocampus and connected regions during the task and then in sleep. 2/n

enclosure with touchscreens on the left, single unit ensembles activated during the task on the right