Gabriel's AI Workspace
Your personal AI workflow hub — projects, learning, prompts, and logs in one place.
🔴 Immediate Focus
Tier Routing — Quick Reference
- Formatting, rewriting
- Grammar checks
- Sensitive data tasks
- Student / financial records
- General research
- Google Docs drafting
- Reminders, calendar
- Short email drafts
- Project planning
- Document consolidation
- Moderate coding
- Quality checking
- Initial app architecture
- Complex analysis
- Strategic planning
- Difficult debugging
- Formatting / summaries
- Grammar checks
- Template filling
- Short email drafts
Projects
All work and personal projects, prioritised by deadline and dependency.
Learning Roadmap
Skills organised by project phase. Learn what each current project requires.
AI Tier Routing
Apply this before every AI message. This single habit extends your free-tier usage more than any other change.
- Formatting, rewriting
- Simple summaries
- Data extraction
- Grammar checking
- Short translations
- Message classification
- ALL sensitive data tasks
- General research (no sensitive data)
- Image descriptions
- Google Docs drafting
- Reminders + calendar checks
- Quick factual lookups
- Short email drafts
- Project planning
- Document consolidation
- Exam-related work
- Lesson plan templates
- Moderate coding
- Structured analysis
- Feedback analysis
- Quality checking your writing
- Initial app architecture
- Complex multi-party analysis
- Strategic planning
- Difficult debugging
- Resolving ambiguous requirements
- Synthesising conflicting research
- Formatting
- Simple summaries
- Casual research
- Template filling
- Grammar checks
- Short email drafts
- File format conversion
Key Terms Glossary
AI Proficiency Certifications
Complete in sequence. Steps 1–4 during Phase 1 of the learning roadmap.
Proficiency Record
1. Certification completions are logged automatically when you add a Session Log entry with a cert name.
2. Your Claude Project system prompt asks you to flag new completions each session.
3. LinkedIn auto-displays Microsoft and Google badges — screenshot and save to Obsidian /Certifications.
Update this record only when preparing your CV or before a job application.
Prompt Library
Click any prompt to view and copy. Filter by context, type, AI target, or source. Overlapping prompts are combined.
Logs
Quick notes per project and issues/decisions log.
Jot down quick observations per project. No formal structure required.
Reference
Setup guides, hardware specs, handover templates, and workflow prompts.
Full AI Stack
| Layer | Tool | Cost | Purpose |
|---|---|---|---|
| High Reasoning | Claude Pro | ~USD 20/mo | App frameworks, project planning, consolidation, complex analysis, debugging |
| High Reasoning (alt.) | ChatGPT Free | Free | Brainstorming, negotiation analysis, initial prompts — keep as second opinion |
| Research + Ecosystem | Gemini Free | Free | General research, Google Workspace tasks, reminders, image gen, AI Studio |
| Teaching Content | NotebookLM | Free | Slide decks, teaching videos, long doc synthesis, exam material prep |
| Coding IDE | Windsurf | Free tier | App refinement, debugging, deployment — receives framework from Claude |
| Local AI | Ollama + Open WebUI | Free forever | Low-tier tasks, ALL sensitive data workflows — unlimited and private |
| Automation | n8n (self-hosted) | Free | Payment alerts, WhatsApp logging, Gmail automation, Sheets updates |
| Version Control | GitHub Desktop | Free | Protect app code; work with Windsurf |
| Knowledge Base | Obsidian | Free | Handover docs, SOPs, lesson templates, prompt library backups |
Custom References
Step-by-Step Setup
Step 1 — Claude Pro + Projects
Step 2 — GitHub Desktop
Step 3 — Windsurf
Step 4 — Obsidian
Step 5 — Docker on Old PC
docker --version
Step 6 — n8n
docker run -d --restart unless-stopped --name n8n -p 5678:5678 -v n8n_data:/home/node/.n8n n8nio/n8n
Step 7 — Ollama
ollama run mistral
Step 8 — Open WebUI
docker run -d --restart unless-stopped --name open-webui -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data ghcr.io/open-webui/open-webui:main
Old PC — Home Server Roles
| Role | Tool | RAM | Notes |
|---|---|---|---|
| Automation server | n8n (Docker) | ~500MB | Runs Gmail, WhatsApp, Sheets workflows 24/7 |
| Local AI inference | Ollama + Mistral 7B | ~8GB | GTX 1060 6GB VRAM — adequate for 7B models only |
| Local AI front-end | Open WebUI | ~200MB | Browser chat interface |
| Vector DB (future) | Chroma / Qdrant | ~300MB | Future RAG for question bank and lesson plans |
New Desktop — Generic Specification
| Component | Minimum | Preferred | Reason |
|---|---|---|---|
| GPU | 12GB VRAM, CUDA | 16GB VRAM | Run 13B+ local models and RAG pipelines |
| CPU | Modern multi-core, 2022+ | 12+ cores | Docker, IDEs, browser multitasking simultaneously |
| RAM | 32GB DDR5 | 64GB DDR5 | Multiple AI tools + Docker + browsers concurrently |
| Primary SSD | 1TB NVMe PCIe 4.0 | 2TB NVMe PCIe 4.0 | AI models and Docker volumes accumulate fast |
| Secondary | 2TB HDD | 4TB HDD | Backups and archives |
| Motherboard | 2+ NVMe, Wi-Fi 6 | AM5 platform | Future CPU upgrade without replacing board |
| PSU | 750W 80+ Gold | 850W 80+ Gold | Brand reliability critical — Seasonic recommended |
Specific Recommendations (Malaysia ~RM10,000)
| Component | Product | Est. Price (MYR) | Reason |
|---|---|---|---|
| GPU | NVIDIA RTX 4060 Ti 16GB | RM 1,800–2,400 | 16GB VRAM sweet spot for local AI |
| CPU | AMD Ryzen 7 7700X / 7800X3D | RM 900–1,400 | AM5 platform, long upgrade lifespan |
| RAM | Corsair/Kingston 32GB DDR5-5600 | RM 400–550 | Upgradeable to 64GB later |
| SSD | Samsung 990 Pro 2TB / WD Black SN850X | RM 500–700 | Top endurance for AI model loading |
| HDD | Seagate Barracuda 4TB | RM 350–450 | Bulk backup storage |
| Motherboard | ASUS PRIME X670-P / MSI MAG X670E | RM 700–1,000 | Wi-Fi 6E, 2 NVMe slots |
| PSU | Seasonic Focus GX-850 | RM 500–650 | Most reliable brand — do not compromise |
| Cooler | DeepCool AK400 / Thermalright PA120 | RM 150–250 | Excellent air cooling |
| Case | Fractal Pop Air / Lian Li LANCOOL 216 | RM 250–400 | Airflow + quiet |
| Windows 11 Home | Microsoft official licence | RM 450–600 | Check if old PC licence transfers first |
| Total Est. | RM 6,000–8,400 | Within budget with buffer |
How the Handover System Works
This Dashboard
Permanent reference manual. Does not change between sessions. Reference by name only in new sessions.
Handover Document
Session memory. Short, specific, regenerated every session. Paste this into new sessions.
Handover Generation Prompt
"We are ending this session. Please generate a handover document based on everything we discussed and worked on today. Use this structure exactly: --- HANDOVER DOCUMENT Project: [fill in] Date: [today's date] Session AI: [Claude / ChatGPT / Gemini] --- ## Current State [What exists right now — what works, what does not] ## Completed This Session [Bullet list of what was accomplished] ## Known Problems or Blockers [Errors, unresolved questions, things that failed] ## Decisions Made [Choices that are now locked in] ## Pending Tasks (priority order) 1. 2. 3. ## Files Modified [File name — location] ## Assumptions Made [Anything assumed that should be verified] ## Risk Areas [Things that could go wrong] ## Recommended Opening Prompt for Next Session [Write the exact message to paste at the start of the next session] --- Be specific and complete. The person reading this will have no memory of today's session."
Context Restoration Prompt
"I am continuing a project from a previous session. Here is my handover document: [PASTE YOUR HANDOVER DOCUMENT HERE] Please: 1. Confirm you understand the current state, pending tasks, and known problems 2. Ask me one clarifying question if anything is unclear 3. Do not begin any work until you have confirmed your understanding My standing instruction: clarify all ambiguous assumptions before proceeding with any task."
App Development Pipeline
You are helping me build a desktop application for Windows. I will describe what I want the app to do. Produce a complete, structured app framework I can take into Windsurf for refinement and deployment. For every app I describe: 1. Confirm your understanding before writing any code 2. Ask 3–5 clarifying questions about functionality, users, and edge cases 3. Once confirmed, produce: (a) a clear file structure (b) all necessary files with full code (c) a README with setup instructions (d) a PyInstaller command to build the .exe 4. Flag any assumptions you made 5. Highlight areas Windsurf will likely need to refine My app idea: [DESCRIBE IN PLAIN LANGUAGE]
Multi-Option Analysis
You are helping me evaluate and compare [NUMBER] options for [BRIEF PURPOSE]. Evaluation criteria: 1. Price and value for money 2. What is included and excluded 3. Reliability and track record 4. Quality of communication and responsiveness 5. Flexibility or customisation options 6. Any red flags or concerns I will provide details one at a time. After all options are given: 1. Build a structured comparison table 2. Identify the top two with clear justification 3. Flag missing information to request 4. Suggest 3 questions to ask each provider 5. Give a recommended decision with reasoning Confirm you understand before I begin.
Feedback Analysis
You are helping me analyse [TYPE OF FEEDBACK]. I will paste [NUMBER] responses below. After reviewing all: 1. Top 3 positive themes with examples 2. Top 3 concerns / improvement areas with examples 3. Urgent or sensitive issues requiring immediate attention 4. 3 concrete action items based on the feedback 5. 2-paragraph summary for [AUDIENCE] Do not invent or infer beyond the responses. Flag unclear items. Feedback: [PASTE HERE]