What Do We Really Expect from AI?
A reflection on persistent AI as a thinking counterpart that can preserve questions, notice change, and challenge premature closure.
Diary tag
LLM
Canonical diary tag page generated from normalized source tags.
12 linked entries currently in the archive.
Tagged entries
Entries linked to LLM
Bounded Capability Extraction Clause v0.2.1
Announcement of Bounded Capability Extraction Clause v0.2.1 and its controls for bounded use of external oracle capability.
A small boundary question kept returning to the Project Ester corpus.
A note on A6 Composition Layer v0.1.1, mapped composition over A0-A5, and why connection is not merger.
AI may be real. The current AI economy may still be measuring the wrong thing.
A note arguing that the AI economy should move from token volume, activity metrics, and valuation rushes to consequence accounting.
AI used by people who do not understand the work becomes expensive theater.
A note distinguishing AI as leverage for expertise from AI as activity theater measured by token volume instead of consequences.
LLMs are not “the AI”.
A note distinguishing LLMs from complete AI entities and framing future AI as composed intelligence under consequence.
A new public layer is now part of the corpus.
A note that DEA formalizes the boundary where input stops being storage and becomes experience that alters continuity.
EA-L4 / EATP is now published in a canonical form inside the public stack.
Public note that EA-L4 / EATP is now a structured package for training provenance, consequence-preserving learning, and auditability.
There is a subtle but important confusion in how we talk about AI learning.
A note that world models require persistent existence under constraints, not only better data or Experience Artifacts.
One of the most underestimated failure modes in current LLM training is not only quality loss.
A note that future training ecologies need Learning Abstracts and Experience Artifacts to remain separate so models preserve origin and consequence.
Data harvesting is not "inevitable". It's a design choice.
A note arguing that raw data should stay local while structured experience, not private exhaust, becomes the export surface for AI learning.
Empathy is not magic
A proposal for an engineered emotional layer where memory, state weights, and L4 constraints define bounded care without simulated feeling.