The trust layer · how these pages are made · July 13, 2026 · in review 📋
The notebook comes first.
These pages begin with Carmen's notebooks, essays, drawings, and recorded conversations. Claude is used to produce proposed drafts, software, and translations from that material. Carmen decides what publishes, and every model-assisted page says whether she has reviewed it. The rules are on the AI governance page.
KISOP's ideas predate today's model-based tools. Carmen developed them in notebooks, journals, drawings, essays, and recorded conversations. She uses Claude to prepare drafts. Carmen reviews the work and makes every decision involving consent, safety, privacy, or publication.
Source material
The name KISOP is roughly sixteen years old. Its source material includes handwritten notebooks and journals, drawings and diagrams, Carmen's essays from 2023 through 2025, a recorded conversation from April 2025, patent drawings from May 2025, and a long essay completed in June 2026.
That material was not machine-generated. Work on a new page or component begins there, and Carmen reviews what comes back.
The loop, step by step
Source material, draft, correction, disclosure, review. Carmen removes the mark only after her review.
1.
Carmen provides the source material. Each working session starts with a journal entry, drawing, essay, voice memo, or other record that holds the idea. One session produces one deliverable.
2.
Claude is used to prepare a draft. The model produces proposed pages, software, and translations from Carmen's materials and the project's dated working notes.
3.
Carmen reviews and corrects the draft. She decides what is overstated, what is missing, and whether the voice is hers. The work can go through several rounds.
4.
People make protected decisions. Carmen decides anything involving consent, privacy, safety, or an agreement. Some pages are reserved for her writing and are never machine-drafted.
5.
The project records what happened. Pages remain marked 📋 until the required human review. Every session is logged, every page is dated and versioned, and AI-assisted pages disclose the assistance in their footers.
6.
Carmen decides what publishes. The project logs each session, dates and versions every page, and records software test counts. Proposed rewrites remain proposals until she accepts them.
What the machine does, and what only people do
Model-assisted work
Page copy — from her notebooks, essays, and working notes
Software, with tests — later checked against independent adversarial test suites
Spanish drafts — shipped with 📋 until a native-speaker pass
Research summaries and straight option lists for her decisions
Only people decide
Every promise a person agrees to — guidelines, terms, privacy
All safety wording — warnings, "we will never ask" lists, crisis lines
What publishes, and when — a named human presses every button
Consent rulings and the vision itself — the founder's, on the record
Information kept outside model-based systems
No artificial-intelligence system runs on this website. The plan, letter, kit, and brief run as ordinary code in your browser. Text entered into them is not sent to a model or to KISOP. Automated release checks cover this behavior, and the page source can be inspected. Model output was used while writing some tools; no model runs inside them.
Carmen's personal archive is out of bounds. The sessions used to make this website are forbidden from reading it. KISOP applies its data-control rule to its founder's archive too.
The same rule applies to future member tools. As "Keeping Your Data Safe with AI Tools" explains, an AI system must request access through the consent gate and is subject to logging and refusals. Member data is not used for model training.
The cost we pay attention to
Large AI models use electricity, water, and hardware. We use the smallest tool that can do the work: larger models for architecture, adversarial review, and final drafting; smaller or local tools for routine tasks. This began as a budget rule and now also guides how we limit environmental cost. The AI governance page explains the policy in more detail.
How you can always tell
·
The footer line. Every AI-assisted page says "Drafted with AI assistance" and states whether Carmen has reviewed it. No exceptions — this page included.
·
The 📋 mark. Wherever you see it, a human pass is still owed — a translation, a page walk, a lawyer's reading. It comes off piece by piece.
·
Dates and change notices. The "New in this place" board on the main site lists visible changes by date. If an AI-assisted page is missing its disclosure, use Write to us → "feedback" on the main site.
Where this stands
Carmen's review is pending. She will also decide whether this page should speak as "we" for the project or "I" in her own name. Until then, the page uses "we" and identifies her role directly.