I hope this account will be useful to someone 😌

Neural synchrony between prefrontal and visual cortex supports visual working memory doi.org/10.64898/2026.06.05.… #neuroscience

Nature Neuroscience Human learning of noninvasive brain–computer interfaces via manifold geometry nature.com/articles/s41593-0…

Intrinsic space-time couplings governing multi-scale cortical dynamics This preprint, “Intrinsic space-time couplings governing multi-scale cortical dynamics,” argues that cortical activity is organized by a tight relationship between temporal scale and spatial scale. The main idea is that different frequencies of brain activity spread across different spatial ranges. Faster neural dynamics tend to remain more local, while slower dynamics can coordinate activity over larger cortical distances. The authors use multi-scale analytical methods to estimate coherence length, meaning how far neural activity remains correlated or synchronized at a given frequency. Their results suggest that the cortex has intrinsic space–time coupling rules that shape how activity propagates across local and large-scale networks. Overall, the paper suggests that cortical dynamics should be understood as a multi-scale spatiotemporal system, not just as separate local signals or independent frequency bands. biorxiv.org/content/10.64898…

Nature A thalamus–brainstem attractor network drives history-biased decisions This paper investigates how past experience biases current decisions using zebrafish. The key finding is that the dorsal thalamus stores recent experience for about 10–20 seconds, and the brainstem integrates that memory with current sensory information to guide behavior. A major point is that the dorsal thalamus contains an attractor-like circuit that stably represents the previous obstacle location. When this circuit was manipulated, the history-based decision bias could be either removed or artificially induced. In short, the study shows that history-biased decision-making is generated by a thalamus–brainstem circuit. nature.com/articles/s41586-0…

Nature BOOK REVIEW How to breathe life back into brain theory Neuroscience needs to stop treating the brain as if it is a computer. nature.com/articles/d41586-0…

Low-dimensional prefrontal representations of objects during working memory Abstract There is an ongoing debate regarding the dimensionality of neural representations. Some accounts emphasize representation within low-dimensional subspaces or manifolds, while others suggest high-dimensional neural codes where neurons respond independently. Here, we investigate the dimensionality of prefrontal cortex (PFC) representations of visual objects held in working memory. We found that object representations are low-dimensional, occupying only 3-6 effective dimensions during both encoding and maintenance in working memory. Control analyses indicate this dimensionality was not limited by the number of objects tested (40) or neurons sampled (~100). We also compared PFC dimensionality to that of a well-established deep neural network model of its inferotemporal (IT) inputs and found an approximately 7-fold dimensionality reduction in PFC. These results suggest object representations are compressed into a low-dimensional manifold in PFC, which might be related to attractor dynamics for working memory, and might facilitate interaction with other variables and cognitive control. biorxiv.org/content/10.64898…

Science Bad memories, bad sleep This Science article, “Bad memories, bad sleep,” discusses the close two-way relationship between memory and sleep. Sleep is not just a resting state; it is a period when the brain actively processes memories. The article focuses on how negative or traumatic memories can disrupt sleep quality, and how poor sleep may in turn affect the way those memories are stored or reactivated. In simple terms, the article argues that bad memories and bad sleep can reinforce each other. Understanding the brain mechanisms behind this relationship may help explain sleep problems linked to stress, fear, or trauma, and could eventually guide new approaches for treating trauma-related sleep disturbances. The full Science page was access-restricted, so this summary is based on the accessible title, snippet, and indexed description. science.org/doi/10.1126/scie…

Principles and Practice of Deep Representation Learning: or a Mathematical Theory of Memory arxiv.org/abs/2606.06624

Cell Decoding the spatiotemporal development of human meninges This study maps the developing human meninges using spatial transcriptomics. It shows that the meninges are not just protective membranes around the brain. They contain diverse cell types arranged in specific spatial patterns and may actively regulate brain development and homeostasis. Overall, the paper provides a detailed molecular and spatial atlas of human meningeal development. cell.com/cell/abstract/S0092…

Nature Neuroscience Induction of cortical on/off periods in awake mice fulfills sleep functions nature.com/articles/s41593-0…

Why you need your whole body – from head to toes – to think | Aeon Essays aeon.co/essays/why-you-need-…

Reasoning Structure of Large Language Models This paper, “Reasoning Structure of Large Language Models,” argues that evaluating large reasoning models only by final-answer accuracy or token count is not enough. The authors introduce a benchmark based on logic puzzles and a method that converts a model’s reasoning trace into a reasoning graph, where claims and dependencies can be checked and analyzed. This makes the model’s reasoning structure measurable, not just its final answer. They also propose a reasoning efficiency metric, which measures how concentrated or direct the model’s logical flow is. Their analysis shows that two models with similar accuracy or similar token usage can still have very different reasoning structures. In short, the paper provides a way to look inside LLM reasoning more systematically, helping researchers compare models, diagnose failure modes, and understand how reasoning changes as tasks become harder. arxiv.org/abs/2606.03883

Trends in Cognitive Sciences Machine understanding This paper discusses what it means for a machine or AI system to “understand” something. The authors argue that AI understanding should not be treated as a simple yes-or-no question. Instead, we should ask what kind of understanding an AI system shows, and to what degree. They emphasize that strong performance on tasks does not necessarily mean true understanding. To evaluate machine understanding, we need to look at factors such as flexibility, generalization, explanation, world models, and sensitivity to meaning and context. In short, the paper argues that AI can show different forms of understanding, but these must be carefully defined and evaluated rather than assumed from performance alone. sciencedirect.com/science/ar…

Nature Human Behaviour Neural similarity predicts whether strangers become friends This study found that people with similar brain activity patterns are more likely to become friends, even if they were initially strangers. By tracking participants over time, the researchers showed that similarity in how people perceive and interpret the world predicts future social connections beyond personality or demographic factors. nature.com/articles/s41562-0…

Neuroscience & Biobehavioral Reviews A beautiful loop: An active inference theory of consciousness This paper explains consciousness using active inference. The authors argue that consciousness depends on three key features: simulation of a world model, competition among inferences, and epistemic depth, meaning recursive sharing of beliefs across hierarchical levels. In this view, consciousness is not located in one brain area. It emerges from recursive, hierarchical inference loops that reduce uncertainty and allow a system to model both the world and itself. sciencedirect.com/science/ar…

Journal of Neuroscience Long-range white-matter pathways enable efficient spontaneous neural activity propagation in the human brain This paper investigates how spontaneous neural activity spreads across the human brain. The main finding is that long-range white-matter pathways help neural activity propagate efficiently between distant brain regions. The spread of brain activity is not determined only by physical distance, but also by the structure and microstructure of white-matter connections. The study also suggests that the same principles may help explain the spread of epilepsy-like activity. In short, the paper shows that the brain’s white-matter “wiring” plays a key role in shaping how neural activity travels across the brain. jneurosci.org/content/early/…

PNAS Resolving Feynman’s restaurant problem reveals optimal solutions and human strategies This paper studies Richard Feynman’s “restaurant problem”: when should you try a new restaurant, and when should you return to one you already like? The core issue is the balance between exploration and exploitation. Early on, it is worth trying new options because you still have time to benefit from a good discovery. Later, it becomes better to choose the best option you have already found. The authors derive an optimal mathematical strategy and compare it with human behavior in experiments. They find that people do not use the exact optimal rule. Instead, they rely on a simpler strategy: as the remaining time decreases, they become less willing to try new options. In short, the study shows that human decision-making is not perfectly optimal, but simple rules can still produce nearly optimal choices. pnas.org/doi/10.1073/pnas.25…

Nature Human Behaviour Brain activity, disruption and connectivity comparisons identify origins of human metacognition in other primates This study suggests that the origins of metacognition are shared between humans and monkeys. Using prospective metacognition tasks together with fMRI and brain stimulation, the researchers found that monkeys can estimate the likelihood of future success, similar to humans. However, humans rely more strongly on an integrated function in frontal area 47, whereas monkeys use more distributed and evolutionarily older frontal circuits. The results suggest that core metacognitive abilities evolved earlier in primates and later became more integrated in the human brain. nature.com/articles/s41562-0…