AI summaryⓘ
The authors challenge the usual way neural networks are used by proposing a model that never freezes—its structure and parameters can keep changing as it learns. They create a system for lifelong learning where the model can grow and adapt continuously, with changes carefully managed by checks rather than being fixed early on. Their approach ensures stability while allowing the model's capacity to grow based on need, rather than being limited by an initial design. They tested their method on common lifelong learning tasks, observing that their system can keep learning without hitting a fixed capacity limit. The authors openly share both successes and failures from their experiments to give a clear picture of their approach's strengths and weaknesses.
lifelong learningneural networkscontinual learningmodel capacitystructural plasticitygrowth governancenon-stationary datastabilityauditingparametric and structural change
Abstract
Today, a neural system is almost always used in two phases -- trained, then deployed -- and in that regime it freezes twice: training ends, and the topology itself was never a degree of freedom. We take the opposite premise as an axiom -- total plasticity: no part of a model, including its structure, is ever frozen -- and derive the governance a lifelong learner then requires. The design's target regime is continual, in-service learning: a long-lived model on a non-stationary stream, whose stability comes from governance rather than immobility and whose capacity follows demand. The result is a growable soft model: an algebra of structural operators (width, hierarchy, composition, input interface, grown cycles, attention heads), each exact at application, budgeted, and audited, with adoption decided solely by a held-out reality gate that treats parametric and structural change uniformly. A complete from-scratch system realizes the whole account; its factory surface is operated end-to-end by a production LLM. Two conclusions follow from the axiom by construction: stability under lifelong change becomes an audit property of the lifecycle, and structure that follows demand removes the silent cap a fixed topology places on later capability where the capacity floor binds. A third is measured: in the worlds where this was measured, the marginal value of new capacity was unobservable before adoption, so workable growth governance took its ex-post form. The same governance extends to evaluative signals, and the core method is evaluated on standard continual-learning benchmarks, where governed growth preserves the ability to keep learning along long task sequences. A pre-registered experimental program adjudicates the mechanism and value claims on the tested problems and reports its failures at full prominence; the map -- positive and negative -- is the contribution.