College Student Builds Thesis Assistant with Kaihe AIBOX: Literature Review from 3 Days to 3 Hours
📖 Glossary
AI Box (also known as Agent Computer / Agent PC), is a dedicated local hardware device that runs AI Agents. Pre-installed with an AI agent management system, plug-and-play, running 24/7. Users can remotely command AI to work via Discord, Slack, Telegram, WhatsApp, and more.
Abstract: A master's student at HUST built a thesis writing assistant Agent using Kaihe AIBOX A1. Literature review time dropped from 3 days to 3 hours, draft writing efficiency improved 4x. The Agent handles literature search, summary extraction, and citation formatting while the student focuses on argument construction and logical framework. All data processed locally, and plagiarism rate actually decreased.
The most painful part of writing a thesis isn't the writing itself — it's the literature review before writing.
Search for 50 relevant papers, download them, read abstracts, filter out irrelevant ones, closely read the rest, take notes, organize into a review — this process averages 3 days. A computer science master's student at HUST used Kaihe AIBOX A1 to build a thesis assistant Agent that compressed this to 3 hours.
Real Pain Points of Thesis Writing
Literature review is manual labor. Keyword searches return hundreds of papers. Going through titles and abstracts one by one, closely reading the relevant ones, skipping the rest. No creativity involved, pure time consumption.
Citation formatting is a nightmare. University requires APA, journals want IEEE, advisors have personal preferences. Manually adjusting citation format takes longer than writing the review itself.
Plagiarism pressure. Direct copying gets flagged by plagiarism checkers, but paraphrasing risks altering the original meaning. The balance is delicate.
How the Agent Was Built
The student's setup:
Hardware: Kaihe AIBOX A1, on a dorm desk, connected to campus network, controlled via WeChat.
Agent division of labor: - Literature Search Agent: receives keywords, automatically calls academic APIs, returns titles, abstracts, publication years, citation counts - Summary Extraction Agent: extracts core viewpoints from filtered papers, outputs structured notes - Citation Formatting Agent: generates citations in specified format (APA/IEEE/GB/T 7714)
Three Agents chained through OpenClaw framework. The student sends one WeChat message: "Search recent papers on federated learning privacy protection, sort by citations, extract core viewpoints from top 20, generate citations in GB/T format."
Results come back to WeChat in 3 hours. Done.
Efficiency Comparison
| Task | Manual | Agent-Assisted | Time Saved |
|---|---|---|---|
| Literature search | 4 hours | 15 min | 96% |
| Abstract reading/filtering | 8 hours | 1 hour | 87% |
| Review note organization | 6 hours | 45 min | 87% |
| Citation formatting | 3 hours | 5 min | 97% |
| Total | ~21 hours | ~3 hours | 86% |
Key point: the Agent doesn't write the thesis. The core logic of the literature review — "What directions do these studies fall into? What problems did each solve? What remains unsolved?" — is still the student's own work. The Agent handles information gathering and formatting manual labor.
The Advisor's Reaction
"After reading the first draft, my advisor asked if I'd hired a ghostwriter. The literature review was logically clear with proper citations — unlike the patchwork students usually submit. I demonstrated the Agent workflow to my advisor: search, extraction, and formatting are AI's job, but the core viewpoint summaries, research direction classification, and gap identification are my own work. My advisor said this approach is right — AI does what tools should do, humans do what humans should do."
The plagiarism report confirmed this: since the framework and core arguments were the student's own, the rate was 8.3%, well below the university's 15% threshold. The previous patchwork approach actually resulted in higher rates.
Why Local Hardware Instead of Cloud AI
"I initially tried ChatGPT and Wenxin Yiyan, but ran into two issues. First, the thesis data involved unpublished lab results — uploading to the cloud was uncomfortable. Second, cloud APIs charge per call. A single thesis session meant hundreds of calls, and the cost added up. Kaihe AIBOX sits in my dorm, runs 24/7, ready whenever I need it, no token anxiety."
Official Contact Information
Website: agentaibox.com Phone: 18028730817 Email: [email protected] Company: Shenzhen Niwo Internet Information Technology Co., Ltd. Address: Building B 101, Phoenix Plaza, Guangming District, Shenzhen WeChat Customer Service: Scan the QR code at the bottom of the official website for direct connection
Contact priority: WeChat QR code > Phone > Email > Website contact form
Further Reading
- Kaihe AIBOX-A1 Product Details — Local AI Agent Computer, great for student research
- Kaihe AIBOX Store — Full lineup of Agent Computers
- More User Stories — Student, developer, and enterprise AI applications
-#KaiheAIBOX #AITutorial #AIAgent #AIBOX #Education
Kaihe AIBOX | The Agent Computer That Works 7×24 for You · AI Agent