Memory Palace · intake room

Capture first. Judge deliberately.

Potentially valuable outputs of human-machine search should not disappear merely because nobody has yet decided whether they are ideas, knowledge, validated results or worth pursuing. Intake preserves them cheaply without silently promoting them.

Why this room exists

Preservation should be cheaper than attention.

Capture latency should approach zero. Judgment can take longer.

An intake record may explicitly say that warrant is unknown, understanding is incomplete and novelty is unassessed. Those are legitimate epistemic states, not embarrassing blanks to be papered over.

Intake is not knowledge. Intake is not commitment. Intake is the refusal to lose a potentially important object before we know what it is.

Current intake

3 preserved artefacts.

The public projection contains only public-safe artefacts. Sensitive or potential-IP material belongs in an appropriate private store.

conceptual-model · option

Epistemic artefact to frontier state model

Provisional chain separating possibility, directed search, epistemic artefact, warrant, assimilation, capability and reachable frontier.

Warrant
exploratory
Understanding
partially-assimilated
Novelty
established terminology plus synthesis/application; novelty unassessed
Next route
lineage-review-and-multi-episode-test

mechanism-conjecture · option

Constitutional attention as adversarial evolutionary search

A provisional architecture in which strategy-generating and adversarial populations co-evolve under a human-governed constitution of plural worthwhile-future objectives, hard constraints, escalation rules and retained historical failures.

Warrant
exploratory
Understanding
conceptually-understood-not-formalized
Novelty
competitive coevolution, constitutional oversight and selective-human-review antecedents; synthesis novelty unassessed
Next route
formalize-and-adversarially-compare

programme-concept-and-empirical-research-candidate · option-high-value-candidate

The Critical Scholar: AI as Academic Method

MPhil programme that uses LLMs as examiner, adversary, conceptual microscope and metacognitive instrument to improve scholarly reasoning while testing durable human capability through unaided transfer after AI scaffolding is removed.

Warrant
conceptually-developed-untested
Understanding
programme-architecture-articulated
Novelty
AI-literacy, metacognition, tutoring and learning-science antecedents expected; programme-level novelty unassessed
Next route
dedicated-lineage-review-then-minimal-transfer-pilot

Routing rule

Nothing gets promoted merely because it was captured.

Artefacts can route to lineage review, validation, assimilation, formalization, experiment, handoff, the Possibility Portfolio, publication review, realization, parking or retirement.

That lets the research system preserve abundance while keeping scarce attention pointed at the few transitions that could actually change what becomes possible.