Conceptual Architecture
The framework uses five cooperating elements: intake, memory, proof, parsing, and presentation. A website requests contextual guidance, not raw personal identity. The visitor controls what kind of signal is used and when.
Whitepaper and technical aspects
A consent-based memory layer for context-aware web parsing.
CommonGround separates useful personalization from invasive surveillance by routing adaptation through purpose, protected context, and explicit consent.
From personas to portable memory
The next shift is reader-owned context.
Digital communication has moved through several stages. People once wrote primarily from their own starting point. Over time, better communicators learned to tailor messages to a reader, audience, or use case. More recently, data has informed how material is adapted to users, often through behavioral signals, analytics, segmentation, and prediction.
That data-driven approach is evolving into persona-based experiences: systems group readers into likely types and adapt content around those inferred profiles. Personas are useful, but they are still approximations. They describe a category of reader, not the actual person sitting in front of the material.
Portable Memory changes the center of gravity. Instead of asking a website to guess the reader's context, the reader brings a persistent, permissioned memory that helps the material adapt to them. CommonGround's work represents the first substantial steps toward that model: a persistent portable memory that adapts material for the reader while preserving purpose, consent, and authorial intent.
01 Engagement and purpose intake
02 Persistent user-controlled memory
03 Zero Knowledge Proof (ZKP)-supported sensitive context
04 AI and parsing transformation layer
05 Adaptive website presentation
Engagement and Purpose Intake
Engagement and purpose intake is the starting point for CommonGround. Instead of quietly watching a visitor and guessing what they want, the system begins by asking why they are there. The goal is to understand the visitor's intent before changing the experience. Visitors arrive while exploring, comparing options, completing a task, learning a topic for the first time, returning with prior knowledge, or looking for specific guidance. Those are very different situations, even when the visitor lands on the same page.
This intake is short and respectful. It uses a focused set of questions, a selected goal, a declared role, or a simple statement of need. The point is to give the website a clear way to ask, "What makes this useful to you?" That answer becomes the first signal for adaptation. It influences the level of detail, the order of information, the examples shown, the tone of guidance, and the next action presented.
This element matters because it shifts personalization away from extraction and toward consent. CommonGround does not begin with purchase history or hidden behavioral scoring. It begins with purpose. That makes the experience more legible to the visitor and easier for the website to justify. The visitor knows why the site is adapting, because the adaptation follows from what they asked the experience to help them accomplish.
Persistent User-Controlled Memory
Persistent user-controlled memory is the part of CommonGround that lets context travel with the person over time. It is not a memory of a single conversation. It is a durable profile of useful context that the person has chosen to preserve: their work, experience, location, goals, constraints, learning preferences, accessibility needs, language comfort, and other factors that affect how information is explained. The important point is control. The memory belongs to the visitor, not to the website.
When a person visits a participating experience, CommonGround uses approved parts of that memory to help the page meet them where they are. A healthcare worker, a job seeker, a parent, a veteran, a student, and a small business owner need different explanations from the same material. Persistent memory recognizes those differences without forcing the visitor to rebuild their context every time they arrive somewhere new.
This memory is structured, portable, and permissioned. Some fields are broad and low-risk, such as preferred reading depth or current task. Other fields are sensitive and stay protected from direct exposure to a website. The memory layer therefore acts less like a marketing profile and more like a private context vault. Its purpose is to improve comprehension, application, and relevance while keeping the visitor in charge of what context is active in each experience.
ZKP-Supported Sensitive Context
ZKP-supported sensitive context is the privacy layer for information that is useful but remains protected. ZKP makes it possible to prove something about a visitor without exposing the underlying fact. In CommonGround, a website receives a trusted signal that certain adaptation is appropriate while never learning the sacred or sensitive information behind that adaptation.
For example, a visitor has context such as a health condition, a disability, a financial constraint, a legal status, a military background, a trauma-related preference, or a geographic risk factor that makes certain explanations more useful. A normal personalization system often requires the visitor to disclose that information directly. CommonGround avoids that pattern. The site does not need to know the private fact. It only needs to know what kind of adjustment is allowed: use plain-language instructions, avoid a certain assumption, show a safer pathway, provide eligibility-aware guidance, or change the examples.
In the current proof point, this begins with permissioned signals that establish the privacy boundary and cryptographic direction. Sensitive facts remain protected while still being useful. That is the ethical center of the ZKP layer. It gives adaptive systems a way to serve people better without asking them to surrender private information as the price of being understood.
AI and Parsing Transformation Layer
The AI and parsing transformation layer is where CommonGround turns context into a better experience. This layer reads the structure of the webpage, the visitor's stated purpose, and the approved memory signals. It then decides how the content is adapted. The page receives a shorter explanation, a deeper technical path, a step-by-step version, an example from the visitor's field, a different order of sections, or a prompt that moves the visitor toward the next useful action.
This is also where uploaded material becomes part of the experience. This technology powers the GoHireMe webpage: a visitor uploads a resume, receives adaptive responses within the page, and is directed by AI parsing toward what helps next. AI parsing extracts relevant career signals, identifies gaps, and connects the resume to the visitor's goal. The result is not a generic chatbot sitting beside the website. It is the website becoming more responsive to the person's actual situation.
This layer is bounded and transparent. It identifies what parts of the page are safe to adapt, what information is used, and what remains unchanged. It avoids inventing facts or over-personalizing beyond the visitor's consent. Its job is not to manipulate the visitor. Its job is to parse, translate, simplify, deepen, sequence, or direct material so the visitor understands and applies it more effectively.
Adaptive Website Presentation
Adaptive website presentation is the visible result of the CommonGround system. This is what the visitor actually experiences: the page adjusts in a way that feels clearer, more relevant, and easier to use. The adaptation is often subtle. A paragraph is rewritten in plainer language. A technical term is expanded. A section moves higher on the page because it matches the visitor's goal. A resume upload triggers directed guidance. A complex topic is broken into a practical sequence.
The presentation layer matters because good context is only useful because it changes the experience at the right moment. A visitor does not have to dig through a full site to find the version that fits them. CommonGround makes the website feel like it understands the purpose of the visit, while still preserving the integrity of the original material. The site remains recognizable. The content remains accountable. The path through it becomes more personal and more useful.
This layer also makes adaptation visible enough to be trusted. Visitors see when content has been tailored and why. They change their goal, reduce adaptation, or turn it off. CommonGround does not hide the fact that the page is adapting. It presents adaptation as a consent-based feature: a way to create clarity through connection without turning the visitor into a data product.
Persistent Memory
Stores durable user context such as work domain, lived experience, location relevance, learning needs, accessibility preferences, language comfort, and current goals.
Engagement Model
Begins with explicit questions about why the visitor is present and what outcome would make the experience worthwhile.
Adaptive Parser
Reads page structure and produces alternate paths, summaries, examples, warnings, or application-focused explanations.
Zero-Knowledge Support
Allows sensitive attributes to influence adaptation through proof or eligibility signals without disclosing the underlying information.
Consent Boundary
Keeps the visitor in control of which memory classes are active for a site, a session, or a specific content request.
Website Integration
Current proof point: this technology powers the GoHireMe webpage, where adaptive responses happen within the page and uploaded resumes are AI parsed and directed. Active beta paths include a client-side layer (beta), server-side adaptation service (beta), CMS plugin (beta), and standards-oriented context exchange (beta).
The technical goal is simple: adapt the message without exposing the person.