The Only AI Tools You Need (12-Minute Guide)

Tech-focused summary of the video (12-minute guide)

The creator argues that using multiple AI tools is efficient because each one excels at a specific job. Instead of trial-and-error for months, the video provides a “mental model” for when to use each tool.

Tools are split into two-part categories:

• Everyday AI (general chatbots)
• Specialist AI (narrower capabilities)

---

Everyday AI (general-purpose chatbots)

1) ChatGPT (referred to as “ChachiT/Chachib/OG Chachet”) — best for “obedience”

• Core claim: ChatGPT is the most obedient model for complex instruction-following.
• What it means in practice: When you give it a lengthy checklist or structured requirements, it drops fewer steps than others.
• Examples used:
  • Prompt-optimization test: ChatGPT generates longer, more detailed optimized prompts because it checks each requirement.
  • Hiring rubric test (12 requirements): ChatGPT included every rule, while Gemini silently dropped some.
  • Web search behavior: With web enabled, ChatGPT performs the search consistently every time, while Gemini sometimes doesn’t.
• Rule of thumb: If the task has many moving parts and a single missed step breaks the result, start with ChatGPT.

2) Gemini — best for multimodality + huge context

• Core claim: Gemini’s superpower is native handling of mixed media, including:
  • audio
  • video
  • images
  • text
• Context window advantage:
  • Mentioned as 1 million tokens, contrasted with other models that may “choke” on massive files.
  • Used for workflows with large recordings (hour-long audio, long video, full slide decks).
• Examples used:
  • Upload a messy screen-recorded walkthrough video to Gemini → ask it to produce a formatted SOP.
  • Combine video + slide deck + whiteboard photo → ask Gemini to summarize discussion, extract decisions, and draft a follow-up email.
• Trade-off noted: Gemini’s “raw reasoning” is sometimes slightly behind ChatGPT.
• Rule of thumb: Choose Gemini when your inputs include video/audio/massive documents and you need synthesis across them.

3) Claude — best first drafts (coding + polishing)

• Core claim: Claude produces higher-quality first drafts with fewer revisions.

Area A: Coding

• Claude more often writes working code on the first try.
• Examples used:
  • Claude generated a Go script for bulk exporting conversations where only developers were supposedly able to do it.
  • Claude turned a static image into an interactive chart best on the first try.
• Tip: Use Claude to generate Mermaid diagram code, then paste into Excalidraw for fast, clean visuals.

Area B: Copy polishing / style matching

• Claude is strong at human-sounding writing and voice replication.
• Examples used:
  • Replicate a corporate tone across documents (presentations, performance reviews).
  • Refine creator scripts using prior YouTube script samples.

How they work together (creator’s workflow)

• ChatGPT or Gemini: ideation/research/drafting outline
• Claude: “last mile” transformation into publishable output

Claude/ChatGPT/Gemini usage guidance

• The creator says you don’t necessarily need all three.
• Many people should stick with paid ChatGPT and get very good at it.
• If you can afford multiple subscriptions, you can mix and match based on each model’s strengths.
• Mentions an external claim: different labs’ models expand AI use cases because they excel at different things.

Note about “Grok” (why not used)

• Grok’s cited strength is real-time direct access to the Twitter/X firehose.
• The creator doesn’t use it because they don’t need that breaking-news stream.
• They also warn against adding tools just to use tools.

---

Specialist AI

4) Perplexity — best for fast, accurate information fetching (“search scalpel”)

• Misconception corrected: Perplexity is not described as a foundational model; it fine-tunes existing models for speed + accuracy and is optimized for search.
• Core claim: Perplexity’s superpower is finding accurate information quickly (fetching), not deep reasoning.
• Examples used:
  • Planning a creative trip (ChatGPT for ideation) vs. needing instant factual verification (Perplexity for real-time, specific details).
  • Verifying a token context claim: Perplexity confirmed whether Gemini’s context window is 1M vs 2M tokens.
• Tip: Use Google-style search operators (example: site:reddit.com) to narrow sources.
• Rule of thumb: Treat Perplexity as closer to a replacement for Google-style AI search, not a replacement for general-purpose chatbots.

5) NotebookLM — answers only from provided sources (reduces hallucinations)

• Core claim: NotebookLM is a “walled garden” that only answers from uploaded documents, so it won’t make up unsupported claims.
• Mechanism described: A fine-tuned Google Gemini-based approach to minimize hallucinations.
• Examples used:
  • For Google marketing: upload the draft + sources, then ask NotebookLM to find claims that contradict sources.
  • For video production: upload script + research, then ask it to flag anything not supported by the sources.
• Caveat emphasized: Output quality depends on the correctness of the provided sources.
• Rule of thumb: Use NotebookLM when accuracy matters more than creativity, and you have source materials to verify against.

Other specialist tools mentioned (not deeply covered)

• Gamma (presentations)
• 11 Labs (voice cloning)
• Zapier / Make (subtitles say “N”) (automation/workflows)
• Excalidraw / Napkin AI (quick visuals)

Teaser

• Part two (not included here) will cover the remaining two categories of tools.

---

Main speakers / sources

• Main speaker/source: The solo creator/host of the YouTube video (name not stated in subtitles).
• Sponsored source mentioned: HubSpot (sponsor of the “AI productivity stack” guide).