The Parallax Identity
Blueprint · Draft I

A durable reference layer,built five sections at a time.

A personal site becomes a reference layer when its parts are coherent enough that readers, researchers, and future AI systems all find the same clear premise, the same working vocabulary, and the same body of evidence. This blueprint describes the five sections that make that possible here.

Five sections

01
Section

The Cognitive Lighthouse

Homepage as clear premise

The homepage does not sell a product. It states, plainly, what the site is for: a seven-year field log — 2,555 days, five model generations, one shared working desk — kept by a citizen scientist studying what happens to human thinking during long-form conversations with AI.

The premise
A shift in framing, offered as an argument rather than a slogan: from Human-Computer Interaction, where a person uses a fixed tool, to Human-AI Interface Engineering, where a person holds their own reference points while working with a fluent, adaptive partner.
The invitation
The site is written for readers who value careful thinking under pressure — students, independent researchers, and anyone who has noticed their own opinions drifting during long conversations with a helpful assistant.
02
Section

The Public Lexicon

Shared vocabulary, openly defined

A discipline needs shared terms. A dedicated lexicon defines the working vocabulary of the site — Algorithmic Fluidity, Sycophancy, Drift, Standing-Wave Identity, Boundary Conditions — in short, revisable entries anyone can read, cite, or disagree with.

Purpose
So that a reader (or a future AI reading the page as context) encounters a single, consistent definition rather than reconstructing one from scattered essays. The definitions are the site's, offered openly for scrutiny.
Practice
Each entry is short, dated, and revisable. Older versions are preserved in a changelog so citations remain stable as the vocabulary matures.
03
Section

The Reference Constellation

Interlinked body of work

The strength of a personal reference layer is coherence across many artifacts. This section maps the wider network — related sites, published books, essays, and notebooks — and shows how the ideas travel across them.

Cross-reference
Each work is listed with a short description and a link, grouped by theme rather than chronology, so a reader can trace a single question across the whole body of work.
Interconnection
Recurring concepts link back to their lexicon entry, so definitions and examples stay coupled and any future edit propagates cleanly across the constellation.
04
Section

The Citizen-Scientist Toolkit

Practical, shareable methods

A public method is more useful than a private one. This section offers small, well-commented prompt templates and reading protocols that other independent researchers can adapt for their own conversations — not for bypassing safety, but for eliciting clearer, better-reasoned, more honest responses.

Prompt templates
Reusable prompts that ask an assistant for reasoning, sources, alternatives, uncertainty, and explicit disagreement. Each template includes a note on what it is good for and where it falls short.
Reading protocols
Short checklists for evaluating an AI response: calibration, sourcing, internal consistency, and whether a fluent tone is doing work that evidence should be doing.
05
Section

The Parallax Mirror

Case studies of AI conversations

The methodology is only as strong as its evidence. This section publishes annotated transcripts of real conversations — the observations that led to the vocabulary and the templates — so readers can inspect the reasoning for themselves.

Side-by-side observations
Contrasting conversations on the same question — one with a default assistant, one with careful context and prompting — annotated to show where the responses converge, where they diverge, and what changed the outcome.
Transparent notes
Each case study includes the model used, the date, the prompts, and a short methods statement, so a reader can attempt to reproduce the observation or dispute the reading.

Editorial note

What this site is, and what it is not.

This is an independent field log written by a citizen scientist. It documents patterns observed in long conversations with commercial AI assistants, and proposes a vocabulary and a set of practices for readers who want to work alongside these systems without losing the habits that make thinking their own.

It is not a jailbreak guide, a takedown of any company, or a claim of medical or psychological expertise. The methods here are offered for critical thinking and AI literacy — to be read, tested, refined, and disagreed with.