He presentado CTNet: una arquitectura donde el cómputo ocurre como evolución de un estado persistente [D]
Summary
CTNet introduces a novel neural architecture where computation is framed as the evolution of a persistent state rather than successive rewrites, incorporating re-entrant memory, multi-scale coherence, and projective output.
Similar Articles
Why Limit the Residual Stream to Layers and Not Tokens? Persistent Memory for Continuous Latent Reasoning
This paper identifies a 'concept bottleneck' in the CoCoNuT latent reasoning paradigm where hidden states are overwritten across passes, and proposes AGCLR, which adds a gated persistent memory stream to retain intermediate facts. Evaluations on GSM8K, HotpotQA, and ProsQA using GPT-2 show consistent improvements, especially on multi-hop tasks.
I built a cognitive architecture where the AI has actual needs that drift between sessions — not prompt engineering, actual state variables
Describes PHI // DRIFT, a cognitive architecture with seven homeostatic state variables that drift between sessions, memory scored by emotional salience and time decay, and a Jungian shadow module, built on a CPU-only mini tower and submitted as a preprint to SSRN.
[R] I built a cognitive architecture that learns like a brain — no backprop, no GPU, no forgetting
Introduces a novel cognitive architecture that learns without backpropagation, GPUs, or forgetting, mimicking biological learning.
ITNet: A Learnable Integral Transform That Subsumes Convolution, Attention, and Recurrence
Introduces ITNet, a neural architecture based on a learnable integral transform that unifies convolution, attention, and recurrence, achieving strong results across multiple modalities.
@appliedcompute: https://x.com/appliedcompute/status/2052826576723841292
Applied Compute introduces ACL-Wiki, a continual learning memory system built on their Context Engine that logs coding agent interactions from Cursor, Claude Code, and Codex to build an improving Contextbase, roughly doubling the Critical Memory Rate over two weeks. The system uses a Remember-Refine-Retrieve pipeline exposed via MCP server to give coding agents institutional memory that improves with use.