About the Project
Who I Am, Why I'm Building Papyrus, and developmental timeline/overall goals
What this project is
My name is Jackson Cohrs, and my academic background is in political science. I graduated with a BA in Political Science from the University of Hawai'i at Mānoa Summa Cum Laude with a 3.9 GPA in May of 2026. With Papyrus, I am exploring how a desktop research environment can keep sources, evidence, citations, and interpretation connected, and integrate GenAI and Machine Learning tools in a conscious, ethical way. It is also a practical way for me to work through questions that sit between research methods, technology, privacy, and environmental responsibility.
I am treating Papyrus as an experimental prototype rather than a finished product. Some described features are present in the current project material, some are partial, and some remain ideas. This website serves as a portfolio and technical write-up: it records the research problem I am interested in, the design choices I am testing, and the limitations that still matter. Currently, it contains no link to download the project, although the eventual goal is to fully open-source it.
Why I started building it
Political-science research often involves moving repeatedly between a source, a note about that source, a broader claim, and the citation that supports it. I wanted to explore an environment in which those relationships remain visible instead of being scattered across unrelated windows and files. I have not been able to find a singular application that allows me to easily switch between reading a source, taking notes on that source, synthesizing and organizing notes across sources, managing citations, and writing based on the notes taken. That interest led to a project centered on reading and organization before automation.
Privacy is one part of that motivation. Research collections can contain unpublished writing, sensitive notes, or material whose context should not be sent to a remote service by default. Papyrus therefore treats local storage and local processing as necessities. No feature of the main app itself uses the internet at all, but plugins can interact with the app and thus should be enabled and selected carefully.
Environmental concerns provide another constraint. Large generative systems consume material resources, but moving a model onto a personal computer does not erase those costs. The project’s more modest goal is to avoid unnecessary computation: use conventional code for suitable tasks, smaller models for narrow jobs, cache reusable work, and reserve generation for cases where it adds something simpler methods do not. Papyrus has not been benchmarked against other research tools for energy, water, or emissions, but it does not require any data centers to run, although local models you can use in the app were trained in data centers.
The class essay that helped shape it
A class essay I wrote titled The Ethical Implications of Using AI for Assignments helped shape these questions. The essay considers the environmental costs of generative AI and the consequences of outsourcing reasoning, especially when a generated assignment is submitted without critical evaluation. It also draws an important distinction between replacing thought and using a tool to support critical engagement.
Papyrus does not fully resolve the ethical problems identified in that essay. Software cannot make a model free of bias, guarantee sound reasoning, or establish that the use of AI is appropriate. Instead, the essay helped define practical boundaries to investigate: fewer unnecessary model calls, visible source connections, reviewable assistance, and workflows in which a researcher remains responsible for interpretation. Privacy became an additional project principle rather than a central claim of the essay itself. These ethical problems are something I've kept in mind while building Papyrus's GenAI features, with local-processing and strict tracking of which information came from GenAI and if it was verified by a human researcher. These are two ways it aims to protect privacy and keep a human-in-the-loop.
Class essay: Read “The Ethical Implications of Using AI for Assignments”
From concerns to design choices
| Concern | Design response | Remaining limitation |
|---|---|---|
| Research privacy | Local processing and explicit network boundaries | Downloads and network-capable extensions still require verification. |
| Unnecessary computation | Selective processing and caching | Resource use has not been measured comparatively. |
| Independent reasoning | Manual workflows and reviewable suggestions | The app cannot ensure critical evaluation by a human, only promote it. |
| Opaque assistance | Recorded inputs, assembled context, model choice, and output, all system prompts viewable and editable | A prompt trace is not access to internal model reasoning, nor can it reveal any inherent biases in training data |
| Unverifiable claims | Source-linked records and explicit prompting to answer only from provided sources | Parsing and association errors remain possible. |
What I am building and learning
The project gives me a place to learn across interface design, document processing, local data storage, retrieval, citations, plugin boundaries, and evaluation of AI-assisted features. I have been working on this project independently, aiming to create a useful application that is thoughtful in its implementation of AI-related features, and can be used for a wide variety of research tasks, regardless of specific field. I have some experience with creating programs in Python, but have used AI coding tools to help with parts of the app while gaining a better understanding of Python. My work focuses on defining research workflows, directing feature development, testing the application, and refining its interface and behavior, rather than strict software engineering.
The technical notes section explains in greater detail how the project works and design principles I've been guided by while building it, the features section further elaborates on the exact capabilities of the application, and the status and limitations section shows the current status of the project and what is and is not working.
What to expect
Development follows my interests and available time. There is no committed launch schedule, release standard, or support promise. I would like to share the source code eventually, but it is not presented here as already public or open source.
For now, Papyrus is best understood as an evolving personal prototype and a record of what I am learning. Continue with Features for the practical view or Status & Limitations for the current state.