↓ Skip to main content
Rosetears
Rosetears

Rosetears

I do not know where I am going, but I am already on the road.

By Rosetears  ·  Founded in 2025  ·  rosetears.cn

❤ This is my rose

This is where I collect technical practice, AI collaboration, tool notes, writing fragments, and the small echoes of everyday life.

May every pause help you pick up a little light, and leave another path forward.

Recent writing

Latest posts

Posts

Which Chart Fits Which Data? Part 1: How to Choose the 6 Most Common Chart Types

To make a data chart communicate its conclusion at a glance, first clarify whether the question is comparison, ranking, trend, composition, relationship, or distribution. This article explains the suitable data, common misuses, decision order, and pitfalls for bar charts, column charts, line charts, pie charts, scatter plots, and histograms, with basic chart-selection logic and examples for data analysis, business reporting, academic writing, and everyday reports.

2026-06-04  ·  2755 words  ·  13 min read

LLM

2026 Update: What to Do When a Paper's AI Rate Is Too High? A Guide to Handling AIGC Suspicion Rates

To reduce an AIGC suspicion rate, the core task is not to turn a paper into casual speech, but to first locate risky sections, then preserve academic facts while strengthening concrete context. This article provides three workflows: segmented Agent rewriting, web-based AI handling, and solidifying rules as a Skill. It is useful for graduation-thesis revision and emphasizes that terminology, data, citations, and section numbers must be manually reviewed, especially for drafts that already have a detection report.

2026-05-26  ·  1802 words  ·  9 min read

office

How to Convert Full-Width and Half-Width Punctuation in Word? A Detailed Tutorial for Two Batch Methods

In Word, there are two practical ways to batch-convert Chinese full-width punctuation into English half-width punctuation: Find and Replace is suitable for one-time cleanup, while a VBA macro is better for repeated reuse. This article lists the replacement range, operation entry points, macro-code idea, and quotation-mark conversion notes step by step, and explains when manual replacement is suitable and when automation is better. After reading it, you can follow the steps directly without checking punctuation one by one.

2026-04-23  ·  1366 words  ·  7 min read

LLM

Still Unsatisfied with Paper Polishing? Use a Roundtable-Simulation Prompt for Top-Journal-Style Rewriting

This SCI paper-polishing prompt uses a roundtable-meeting format and forces the AI to simulate an editorial workflow similar to Nature. It does not rewrite immediately. Instead, a logic architect and reviewers first run three rounds of debate, then a writing specialist outputs two versions, Draft A and Draft B. The process raises information density and reduces the mechanical feel of AI-generated prose, making it suitable for detailed academic-paper revision.

2026-02-11  ·  6035 words  ·  29 min read

LLM

Claude Code Supervising Codex: A Reproducible Acceptance Framework to Keep AI from Drifting

A framework for keeping long-running AI development from drifting, using tasks.md for process, feature_list.json for state, progress.txt for handoff, and reproducible validation bundles with scripts, logs, and screenshots. Worker implements and delivers a bundle; Supervisor runs acceptance and records evidence; after PASS, state and commits are archived.

2026-01-31  ·  847 words  ·  4 min read

zotero

What Changed in Zotero 8: Upgrade Steps and Key Improvements

Zotero 8 was released on January 22, 2026 and is based on Firefox 140 ESR. Key improvements include a unified citation dialog, merging item-list and library modes, PDF reader themes such as Dark, Snow, and Sepia, annotations shown directly in the item list, reading mode for web snapshots, and ARM64 Linux support. It requires macOS 10.15+ or Windows 10+ and no longer supports Windows 7/8 or macOS 10.14. Check plugin compatibility before upgrading.

2026-01-25  ·  1251 words  ·  6 min read

Prompt

Cannot Find Relevant Literature? Two Fast Search Methods (CNKI + WOS)

This article introduces two methods for literature search: when you already have papers, expand the literature pool through CNKI citation networks or Inciteful.xyz; when you only have a topic, build and iteratively refine search queries with CNKI professional search and Web of Science advanced search. It includes operation steps, a query-generation prompt, and an optimization workflow using 5–20 seed papers.

2026-01-24  ·  2130 words  ·  10 min read

LLM

Goodbye Vibe Coding: Spec-Driven Development with SpecKit + OpenSpec

A practical explanation of Spec-Driven Development (SDD): write What/Why and acceptance criteria as a living specification before asking AI to implement How. The article explains why vibe coding easily loses intent, and shows how SpecKit (Constitution → Specify → Plan → Tasks) fits 0→1 development while OpenSpec (proposal → apply → archive) fits 1→n iteration.

2026-01-21  ·  1058 words  ·  5 min read

LLM

Stop Mixing Them Up: What Are the Differences Between Prompt, MCP, and Skills?

This article clarifies six often-confused AI tool concepts: Prompt handles temporary conversation instructions, Skills preserve standard workflows, MCP exposes external tools, Commands create shortcut templates, Agent runs autonomous loops, and Subagent isolates parallel tasks. It compares definitions, use cases, engineering cost, reuse, and decision criteria based on Anthropic documentation and LangChain-style agent concepts.

2026-01-15  ·  1157 words  ·  6 min read

Browse more posts