Shubham's AI Playbook
A personal reference for navigating the modern AI ecosystem - tools, workflows, models, and principles.
Choose Your Path
Pick where you want to go. Each path is organized for a different purpose.
Start Here
New to AI? Begin with orientation and quick start guides.
Learn
Follow a guided learning path based on your role and goals.
Decide
Compare tools, models, and frameworks. Make informed choices.
Reference
Glossary, cheatsheets, principles, and quick lookups.
Research
Latest trends, model releases, and open-source developments.
Deep Dives
Technical deep dives from fundamentals to production systems.
Resources
Curated papers, tools, communities, templates, and case studies.
Claude
The Anthropic ecosystem: models, API, Claude Code, MCP, skills, and enterprise deployment.
OpenAI
The GPT ecosystem: models, Responses API, Codex, Realtime API, MCP, and enterprise deployment.
DeepMind
Google's AI research lab: Gemini models, Antigravity, Veo, Gemma, AlphaFold, and science breakthroughs.
DeepSeek
The cost leader: open-weight models, dual API, 15+ agent integrations, up to 95%% cost savings.
Community
Contribute, report outdated info, and join the conversation.
New to the AI ecosystem?
For a structured map of the whole landscape — how each major lab (Anthropic, OpenAI, Google DeepMind, DeepSeek) is organized, and where tools like Cursor, Perplexity, and LangChain fit — see The AI Landscape.
Otherwise, pick a path above, scan the comparison below, or jump into the Models Decision Guide.
Quick Comparison: Top Conversational AI
| Feature | ChatGPT (GPT-5.5) | Claude (Claude Opus 4.8) | Gemini (Gemini 3.1 Pro) | DeepSeek (DeepSeek V4 Pro) | Perplexity |
|---|---|---|---|---|---|
| Context window | 1M | 1M | 1M | 1M | Varies by model |
| Knowledge cutoff | Dec 2025 | Jan 2026 | Jan 2025 | — | — |
| API pricing (in/out per 1M) | $5/$30 per 1M | $5/$25 per 1M | $2/$12 per 1M | $0.435/$0.87 per 1M | Uses partner models |
| Reasoning model | o3 ($2/$8) | Extended thinking (built-in) | Deep Think (built-in) | DeepSeek R1 ($0.435/$0.87) | Uses partner models |
| Image input | ✅ | ✅ | ✅ | ❌ | ✅ |
| Web browsing | ✅ | ✅ (claude.ai) | ✅ | ❌ | ✅ Core feature |
| Code execution | ✅ (Code Interpreter) | ✅ (Artifacts) | ✅ | ✅ | ❌ |
| Subscription | Plus $20/mo | Pro $20/mo | AI Pro $20/mo | Free + API | Pro $20/mo |
| Free tier | GPT-5.5 Instant generous | ~20–40 msgs/day | Generous Flash tier | Unlimited (chat) | 5 Pro searches/day |
| Agentic tools | Operator, integrations | Claude Code, Cursor, Windsurf | Workspace integrations | Open weights | Research synthesis |
| Best for | General + reasoning | Writing, long docs, complex reasoning | Very long context research | Cost-conscious teams | Real-time + cited sources |
Model specs (context, pricing, reasoning, vision) are pulled from the playbook's model dataset. Verified as of 2026-06-05.
Tip: If you hit free limits often, the API is often cheaper for heavy use - you pay per token with no daily cap. For regular personal use, a $20/mo Pro plan on your most-used tool pays for itself quickly.
Complete Beginner? Here’s Your Path.
If you’re new to AI, don’t try to absorb everything at once. Follow this sequence over 4–6 weeks. Each step builds on the last.
1. Days 1–7 - Pick one AI assistant and use it every day
Don’t compare them yet. Use it for real tasks - drafting emails, summarising articles, explaining things you don’t understand. Volume over perfection.
- ChatGPT - most popular, widest ecosystem
- Gemini - best if you use Google Docs / Gmail
- Claude - excellent for writing & long documents
All three are free. Pick one and stick with it for at least a week.
2. Week 2 - Learn what you’re actually working with
Before adding more tools, spend 30 minutes understanding the basics. It changes how you use everything.
- Read the Glossary - token, prompt, context window, hallucination
- Read Common Confusions - debunks myths everyone believes early on
- Watch 3Blue1Brown’s “But what is a GPT?” on YouTube - 20 min, best visual explanation available
3. Weeks 3–4 - Add one specialised tool for your biggest pain point
Don’t sign up for everything. Pick the one that removes real friction from your day:
- Coding / refactoring → Cursor ($20/mo) or Claude Code (pay-per-token)
- Too many meetings → Otter.ai or Fathom (transcription + summaries)
- Research-heavy work → NotebookLM (free, excellent for synthesis)
- Lots of writing → Grammarly or Claude.ai Pro
- Need web answers with sources → Perplexity ($20/mo Pro)
4. Month 2 - Stay current without getting overwhelmed
AI moves fast. You don’t need to read everything - just stay loosely informed.
- One newsletter: The Rundown AI (daily, 5 min) or The Batch by Andrew Ng (weekly, deeper)
- Follow Ethan Mollick on X - best practical takes on using AI at work
5. Month 2+ - Pick your lane
Go deeper based on what you actually care about. Once you have the basics, use this playbook as your reference. Pick the path that fits your interest:
- Agentic AI & Coding - Cursor (78% SWE-bench) → Claude Code → Agent Skills → Windsurf (75%) → Agents page
- Creative work - Midjourney → Runway (video) → Suno (music) → Creative workflows
- Building products - Cursor → Lovable → n8n automation → Tools page
- Business & automation - Microsoft Copilot → Zapier → n8n workflows → Workflows page
- How AI works - fast.ai → Karpathy’s YouTube → Read the Glossary & History pages
- Privacy / self-hosting - Ollama → LM Studio → run Llama locally → Open-Source Models page
- Ethics & Safety - Read the Principles page - agentic AI risks, alignment, responsible deployment
The biggest mistake beginners make: signing up for 10 tools in week one. Master one thing first. Focus on agentic AI tools (Cursor, Claude Code) if you code - they’re transformative in 2026.