The Only AI Tools You Need

Study Guide

The Only AI Tools You Need

Creator: Tina Huang. Published June 29, 2026.

Tina Huang, an AI educator, YouTuber, and founder of the AI education company Lonely Octopus, walks through the AI tools she actually uses "day in and day out." She frames the tour as a "map of my life," organizing her toolkit into four zones: the home base, the builder dungeon, Octopus HQ, and the content studio.

Core Philosophy

Huang's guiding principle for adopting AI tools is to master a small core set of tools really well rather than chase the hype of many different tools. The whole video is structured around this idea: a handful of durable, repeatedly-used tools per area of life.

Zone 1: Home Base (general life)

This is where she "spawns into existence" — waking up, eating, gym, investing, banking, general lifestyle. Three primary tools:

  • Claude Code (Anthropic) — a local AI agent for building custom workflows and automations, much of it no-code. Examples: a custom morning brief compiling emails, Slack activity, calendar, investment portfolio, and daily focus; plus a custom investment workflow with a dashboard and stock research.
  • Claude chat — her general chatbot for everyday questions (how to plan things, how to connect a Logitech mouse, song lookups, flight durations and layover ideas).
  • Perplexity — her general search tool. She prefers it over a chatbot when she wants objective answers across multiple sources and multiple models (not just Claude models). Practical bonus: she lives in Hong Kong, where Claude chat, Gemini, and OpenAI are blocked and require a VPN, but Perplexity works without one.

Zone 2: Builder Dungeon (building software and products)

The tools she uses to build other tools — internal products (slide generator, news tracker, learning app, accounting software, marketing tools) and external client tools (AI databases, CRMs, automations). She splits these into two groups.

AI coding tools

  • Claude Code — best for deep, autonomous work inside their own internal codebase.
  • Warp — best for collaboration and orchestrating many coding agents at once.
  • Cursor — nice when you want a familiar IDE.
  • Codex — most useful for specialized tasks or when another tool requires a coding-agent subscription.
  • Lovable — her general option when she just wants a functional, beautiful prototype without going deep into code.
  • Retool — for building something very business-specific.

Agentic tools

  • Claude Code and Hermes (an open-source local AI agent) for workflows and automations.
  • n8n — her favorite no-code option for building her own agents.
  • OpenAI Agents SDK — her favorite code option for building her own agents.

To support this work she runs a Mac Studio (64 GB RAM, M4 chip) at the office 24/7 to host big open-source local models — Qwen, Llama, GLM, MiniMax — when she wants things private, local, and free. She also notes supporting infrastructure that matters for real building: data stores, MCP servers, VPS, and hosting platforms.

Sponsor Note: SerpApi

One underlying tool she uses when building software workflows is SerpApi (the sponsor of this portion of the video). It programmatically returns clean, structured search results from Google, YouTube, Google Scholar, shopping, news, and more as JSON, avoiding captchas, rotating proxies, and breaking layouts. She uses it for her Hermes news-tracking automations and slide-generator research. Cited stats: 99.9% uptime, ~1.2s average response time, 250 free search credits on signup.

Zone 3: Octopus HQ (running the company)

Her company, Lonely Octopus, has 10+ people (engineers, ops, product/growth, editors, sales) — mostly remote, with a Hong Kong office. Her tools here center on management, communication, and "meta business."

Management and communication

  • Claude Code — custom workflows tracking notifications/messages across Discord, Notion, and Slack, plus drafting emails.
  • Notion + Notion AI — primary database for meeting notes and documentation; Notion AI helps sift through it.
  • Google business ecosystem (Drive, Docs, Sheets, Meet) with Gemini embedded for summarizing, transcribing meetings, and writing.
  • Wispr Flow — a voice/audio dictation input tool she likes.

Meta business (accounting, banking, taxes)

Mostly custom internal software built with the builder-dungeon tools. She builds it herself to keep it private (local AI models) and fully customizable to coordinate accounting, taxes, and banks. She admits this got harder as the company scaled and that she "absolutely hates" meta-business work, but values AI help with it.

Zone 4: Content Studio (content creation)

A surprising admission: despite running an AI education company, she is "pretty low AI" when it comes to content because she enjoys making it and believes the human touch matters. She says her content is "100% human."

  • Hermes — her primary content tool. A workflow tracks AI topics across Reddit, X, Hacker News, Instagram, YouTube and other sources, flags rising topics, and generates a research outline she then develops herself.
  • Claude Code — generates video assets like graphics.
  • Internal AI tools (from the builder dungeon) — produce the free video companions and written resources.
  • "AI Tina" — a chatbot trained on all her released social content. The team can ask it information-based questions instead of waiting on her. She notes it doesn't replace her actual decision-making. (It surfaced her two prompt-engineering mnemonics: "tiny crabs ride enormous iguanas" and "ramen saves tragic idiots.")

She expects her content tools to change the most of any category, since they are still experimenting.

Key Takeaways

  • Depth over hype: a small, well-mastered toolset beats constantly chasing new tools.
  • Organize tools by life function, not by vendor — the same tool (Claude Code, Hermes) recurs across zones because it fits multiple jobs.
  • Local + open-source models (run on a Mac Studio) enable privacy, customization, and cost savings for sensitive work like accounting.
  • Build vs. buy: for business-specific and private needs, she builds custom internal tools rather than adopting off-the-shelf SaaS.
  • AI augments, humans decide: she deliberately keeps content human and treats tools like "AI Tina" as information helpers, not decision-makers.
YouTube