Proposing archiveJ. (version 1.2)
Hello. In 2026, we actively applied AI to archivej and used approximately 160 billion tokens over 2 months. ^^ We propose a more advanced version of archivej, a tool for easily creating digital archives. (version 1.2)
1. Expanding from digital archives to current work
Looking at the various AI and chatbots being released today, there is a lot of talk about memory. Memory is both personal and a social product, and in companies, various outputs of memory are produced in the form of documents. I believe we have finally reached a point where software can actively store memories. Previously, putting both conversation records and context records into a database was heavy, but now we can store them lightly so that AI and Agents can understand users.
Over the next 6 months, I expect many “memory” services will emerge beyond records and databases. In other words, the daily memories and recollections that people create, as well as the activities of companies and organizations (including non-profits), will all be created and managed in the form of “memories.”
“archivej goes beyond the basic version and adds a new service called ‘current memory.’”
Simply put, these are tools for handling current work. The users of these tools are offices where humans and agents work together. In August 2026, TCG-OFFICE and FLOW-OFFICE are being actively tested, and we will gradually deploy them to users without token issues. Please contact us anytime with questions and search for Cloudflare OS.
If you have experience owning or operating an archive, and you understand that once it’s organized and enters the information retrieval system, the massive amount of memory you created will influence current and future work that can be expanded, you’ll quickly understand this.
2. Regarding formats / Current work forms
Through archivej, we can respond to digital archives and the changing web.
-
We respond to digital archives.
-
We configure them in the form of websites and web services. (Idea -> Execution archive -> Case sharing)
Formats being updated in archivej
-
book (public book format through continuous editing)
-
article (research paper format through continuous research)
-
application (apps that execute when you request music at a cafe, software maintained by developers)
-
office (employees working together in a work environment that includes ChatGPT)
-
skill (functional skills important in digital archives)
What archivej is trying to solve
-
Geographic information matching / Visual expansion
-
Digital archive description and expansion for AGENTs, which make up over 60% of users / Monetization
-
Natural presentations, PPT, Keynotes based on materials
3. Responding to wrist-aching digital archive material organization and photo organization
The most time-consuming task when editing a white paper is having humans read countless materials, edit them again, and set directions through conversations or organization with senior decision-makers. Similarly, the most concerning part when developing a digital archive is organizing materials and creating various content based on those materials. archiveJ, which we developed directly, accesses the relevant folder and attaches a nearly free AI LLM AGENT to proceed with related work.
Since we’re using LLM, doesn’t it cost a lot due to token usage?
Yes, if you don’t understand the code or related processing methods and don’t optimize while entrusting extensive editing, it will waste tokens doing unnecessary work and tasks you didn’t assign. But the digital archives and white paper materials we manage have a fixed quantity, and these are tasks humans could do too. We’ve prepared a way for individual researchers to move forward by attaching workers (Agents) to areas that only organizations with budgets or higher-level institutions could do.
3-1) Agent work organizing records
(Audio) When you leave oral record files (transcriptions) of interview content, the agent automatically checks the context. (Photos) The agent attaches descriptions to photos and understands the context. (PDF, Hangeul, txt) Converts documents into readable formats and reads existing business and documents. (Video) Understands video content and grasps the audio or content within. (Video without audio) If transcription isn’t done, understands and describes the video content (footage). (Links, sites) Captures the site and reads what it contains to provide descriptions.
3-2) Agent work creating new content based on records
(General public) Provides exploration/search for organizations to understand year-by-year work through digital archives (Researchers) Publishes research guides for researchers to read deeply into related materials (Researchers) Provides basic materials that can be used in papers through cross-comparison of related content (Content) Publishes blog content based on generated content (Content) Combines various media in the direction the organization wants to publish exhibitions with themes and intent (Content) Publishes numerous collections based on published keywords
3-3) Translation work and multilingualization
(Translation) Through this system, about 2-3 hours of work and translation can handle approximately 1000 pages. This isn’t simple sentence translation but includes basic information about the item, collected keywords, related language-specific pages, and all generated content. (Translation and website) If 10,000 items are archived, all 10,000 items can be served in each language. (Translation) Including Google Translate or direct translation, we can more easily publish related values in multiple languages.
4) Ongoing work and limitations
(Improvement) Related instructions, missions, and material context are organized through discussions between the archive owner and development team, and the AI Agent references this. (Limitations) AI Agents don’t do everything perfectly. When continuous time and effort are invested, humans help by setting accurate guides and missions for good performance. (Expansion) As it expands, researchers and customers will want to create more archives. But refinement and guidance of content will require more time. (Customers) We will work slowly and diligently to ensure proper archives and processes are established with customers. (Costs) Token usage costs money. We also propose ways for individuals and organizations to secure or manage this.
Recent News
- We expanded the page that organizes by time and chronology.
- We shared results from the living lab in Jecheon.
- Read what was explained at Nomad Coders, run by a smart couple. It’s important to keep Agents closer to you. It’s surpassed 60% of internet searches.
First written June 13, 2026 (Updated August 14) TCG Han Un-jang