@GoSailGlobal: Claude Code 102 by Dr. Mushtaq Bilal for Academic Researchers. His first post had 4M+ views / This one is a hard follow-up. Core argument: Claude Code can fully serve as a workflow engine for 'long-term research projects'…

Claude Code 102 for academic researchers, by Dr. Mushtaq Bilal, is out. His first post garnered 4M+ views / this one is a substantial follow-up. His core argument: Claude Code can be fully utilized as a workflow engine for 'long-term research projects'—applicable to PhD theses, long-form writing, and multi-year research endeavors. Here are 5 core modules, ranked by importance by me: · First: Organize everything with 'Folders + CLAUDE.md' Your main folder is 'My Thesis'. Subfolders include: - Literature (PDFs + reading notes) - Chapters (chapter drafts) - Data (datasets) - Notes (meeting minutes / inspiration) - Correspondence (emails from advisors / communication with co-authors / reviewer feedback) Place a local CLAUDE.md in each subfolder, and a global CLAUDE.md (project constitution) in the main folder. Examples: Chapters/CLAUDE.md: 'When reviewing drafts, structure feedback by Argument / Evidence / Literature / Counterarguments. Use MLA 9 for all citations.' Data/CLAUDE.md: 'Treat all CSVs as raw data. Never overwrite originals. Save cleaned versions with a _clean suffix.' Correspondence/CLAUDE.md: 'Prioritize overlapping points between reviewer comments and co-author feedback.' When Claude Code works in a specific subfolder, it reads both the global CLAUDE.md and the local one for that folder. The result? When you work on chapters, it knows MLA citation rules. When you handle data, it won't overwrite original files. When you process reviewer letters, it first identifies cross-cutting feedback. The essence of this design is turning a general-purpose model into your project's 'dedicated assistant'. · Second: Plan Mode + Custom Slash Commands Plan Mode (Ctrl + Shift + M) makes Claude outline the steps before executing them. You review and approve before it starts. Applicable to: Tasks involving more than 3 steps, cross-folder operations, or long outputs. Examples: 'Synthesize notes from 35 papers', 'Filter systematic reviews', 'Clean dataset + generate codebook'. Custom Slash Commands serve as shortcuts for repetitive tasks. Let Claude create them for you with a single sentence: 'Create a slash command called /firstdraft that turns my notes in the Notes folder into coherent paragraphs.' Claude automatically writes a .md file into .claude/commands. Mushtaq's two iron rules: 1. Don't create slash commands for tasks used less than once every six months. 2. If a command exceeds 15 lines, split it into two. · Third: Subagents to solve 'context clutter' The biggest issue with long projects: You ask Claude to read 20 PDFs and then write an outline. Its responses become slow and vague because all previous content remains in its working memory. Subagent = A mini-Claude with its own independent context window. It reads its own PDFs and materials without polluting your main conversation. Mushtaq provides 4 subagent templates for researchers: - Literature Reviewer: Automatically generates structured summaries based on your arguments whenever a new PDF is added. - Citation Checker: Verifies if each citation matches the literature database. - Methodology Auditor: Checks consistency between the methodology section and the data. - Reviewer 2: Acts as a hostile reviewer to critique your drafts. The most powerful usage: Parallel critique. Command: 'Have Methodology Auditor and Reviewer 2 evaluate chapter_4 in parallel, each generating a report.' Both subagents run in their own contexts and generate separate reports, leaving the main session completely unpolluted. · Fourth: Connectors / Integrating Research Tools In Claude Code, go to 'Customize → Connect your apps' at the top. This lists officially supported MCP connectors: Zoom, Google Drive, Zotero, PubMed, arXiv, etc. Practical prompt example: 'Pull transcripts from my last three Zoom calls with co-authors, extract all comments related to Chapter 4, and save them to a new file in the Correspondence folder with today's date.' Mushtaq's warning: Do not connect apps containing confidential data (e.g., don't connect if Slack contains unpublished data). · Fifth: Hooks and Scheduled Tasks to Automate Everything Hook: Scripts that run automatically when a specific event is triggered. Most practical example: Auto-backup before editing. 'Create a pre-edit safety hook that backs up the current version before editing any chapter.' Later, when you ask it to modify chapter_4, it automatically backs up before making changes. Scheduled Task: Runs on a timer. Most practical example: Every Monday at 9 AM, use the PubMed MCP to pull this week's new papers, pass them to the Literature Reviewer subagent for filtering, and save the table to the Weekly Scans folder. Mushtaq's iron rule: Never create hooks or scheduled tasks that involve 'deleting files'. · Putting these five modules together essentially treats a PhD thesis as a software engineering project. Mushtaq has helped non-technical researchers accomplish something difficult: translating the perception of 'Claude Code = programming tool' into 'Claude Code = project workflow engine'. This framework is equally applicable to: - Book writing (one subfolder per chapter) - Long-term consulting projects (one subfolder per client) - Household budgeting and long-term planning (one subfolder per domain) As long as the task spans months, requires accumulating context, involves repetitive tasks, or needs multi-perspective reviews, this template can be applied. Link: https://x.com/MushtaqBilalPhD/status/2053829787219595725… (The 101 beginner guide had 4M+ views / scroll up on the same account to see the previous post)