Video summary

Everything Perplexity In 34 Minutes

Main summary

Key takeaways

Technology

Summary of the Video (Perplexity: Features, Workflows, and Review-Style Tips)

Tina Huang delivers a fast ~34-minute walkthrough of Perplexity, positioning it as an AI “information discovery / curiosity” tool. The video frames Perplexity as more than chat + search, emphasizing grounded, source-cited answers and its usefulness for research, content creation, automation, and analysis.

Core Perplexity Features Covered

1) Interface + Modes (How to Start Using It)

  • Go to perplexity.ai and start with a chat-style prompt.
  • Key mode choices:
    • Everyday questions
    • Deep research
    • Labs (project-style workflows)
  • The video also highlights:
    • Model switching (including Perplexity’s “Sonar” and other models like Claude/Gemini/GPT, etc.)
    • Search scope options such as web, academics, social, and finances
    • File upload / connecting files
    • Voice / dictation mode

Hidden/Secondary Settings Mentioned

  • Personalization: interests (e.g., stocks/topics), location, and optional self-description
  • Task: scheduled/recurring automations
  • Connectors: integrations like Google Drive/Dropbox/WhatsApp
  • API access: for building workflows/apps

2) Intelligent Search (Grounded Answers with Citations)

The video stresses that Perplexity:

  • Runs multi-source research, including reviews/opinions from sites like Reddit
  • Displays sources and can optionally show an explanation/thought process for how it arrived at answers
  • Offers follow-up query suggestions and supports iterative questioning

Example Use Cases

  • Trip planning: events/festivals + visitor tips for a specific date range, with official links
  • Uses discussion-style inputs (e.g., Reddit advice) for practical guidance like safety/transit tips

3) Spaces (Saving Threads + Reusable “Specialized Bots”)

  • Add to space saves a conversation into a named workspace (e.g., “September trip planning”).
  • Spaces function like specialized assistants/templates (similar in spirit to custom GPTs or Claude projects), including:
    • Stock screeners
    • Topic-specific research templates
  • Supports shared threads/inspiration from other users.
  • Encourages customizing behavior using instructions/system prompts.

4) Generate “Pages” / Report Formatting

A major workflow demo shows that in deep research, results can be converted into a “page” for nicer presentation.

Pages can include:

  • Multiple sections (headers and editable text)
  • Media/images and generated visuals/diagrams
  • Preview and publish
  • Export options such as PDF, Markdown, or document formats

This is positioned as especially helpful for bloggers and marketing content.

5) Deep Research Mode + Finance/Scoped Sources

For more advanced questions:

  • Switch to Research mode
  • Select scoped sources, including finance (example: searching SEC filings)
  • Deep research is described as:
    • Using many sources (example mention: 49 sources)
    • Producing outputs that feel more grounded, with less hallucination
    • Providing CSV/table outputs and an assets view that resembles step/code-like artifacts

Example

  • “Breakout hardware companies to invest in 2025”
    • The demo cites finance sources (e.g., Yahoo Finance)
    • It’s explicitly framed as a demo, not financial advice

6) Labs: Analysis, Dashboards, and “Execution”

The speaker contrasts Labs with normal deep research:

  • Labs can go beyond collecting sources—it performs analysis and interpretation
  • It can support execution-style outputs (e.g., turning insights into structured deliverables)

Example

  • “Analyze my YouTube channel” using Labs:
    • Identifies patterns/trends (e.g., time-bounded learning/video titles)
    • Produces a dashboard (after requesting: “convert this information into an interactive dashboard”)
    • Recommends improvements to content formats and strategy

Positioning vs Other Tools

A stated advantage: Perplexity can locate and use the right sources more autonomously, without requiring the user to provide datasets upfront.

7) Discover Tab (Personalized News)

  • Discover provides a personalized “for you” news feed.
  • Personalization includes:
    • Topic preferences (e.g., tech/AI/space)
    • Adding companies/stocks or other interests
    • Optional location setup
  • The speaker particularly likes it on mobile, where each item includes summaries plus cited sources, helping reduce bias by pulling from multiple outlets.

8) Task Automation (Scheduled Workflows)

The video highlights Task for scheduled automations, including:

  • Example schedule: daily at 7:00 a.m.
  • The task:
    • Analyzes weekly tech developments focused on AI/ML, strategy moves, and emerging trends
    • Sends results via email and mobile notifications

Presented as a “hidden but powerful” feature.

9) Connectors (Integration + WhatsApp Workflow)

Connectors include:

  • Google Drive / Dropbox: search across stored documents
  • WhatsApp: chat with Perplexity and schedule delivery

WhatsApp Demo

  • Sets up a task to send daily tech news summaries via WhatsApp (instead of email).
  • Privacy note mentioned:
    • The speaker claims WhatsApp chats are not recorded in the Perplexity app and are kept between the involved parties.

Pricing Overview (As Described)

  • Free tier

    • Unlimited general search queries
    • Limited counts for pro/deep/labs-related actions
    • Labs access is stated as not included (speaker corrects herself during the segment)

  • Pro tier ($20/month)

    • More citations/answers (described as “10x”)
    • Access to Comet and Perplexity Labs
    • Unlimited files/uploads, image generation, etc.

  • Max tier ($200/month)

    • “Early access,” unlimited features, advanced models, enhanced video generation

  • API

    • “Pay as you go”
    • Pro includes $5 credits for API usage
    • Mentions an enterprise pro option for collaboration/support/security compliance

Comet (Perplexity’s Agentic Browser): Use Cases + Security Caveat

After skepticism, the speaker explains Comet as:

  • A browser paired with an assistant (a companion assistant is referenced, likely Perplexity’s Comet assistant).
  • Capabilities described:
    • Summarize pages/video content
    • Search and refresh pages
    • Answer emails
    • Interact with shopping flows (implied)

Security Warning (Personal Trust Boundary)

She cautions about prompt injection risks and mentions prior security leakage concerns (e.g., exposing sensitive credentials). As a result, she avoids giving Comet access to sensitive info such as login details or credit cards.

Two Preferred Comet Use Cases

  1. Video summarization and Q&A with timestamps (YouTube example)
  2. Testing discount codes for shopping
    • But completing purchases in Chrome, not inside Comet

Perplexity API + Automation Example Workflow

The video concludes by emphasizing that an API key allows:

  • Building complex workflows and applications
  • Using Perplexity’s intelligent search as an upstream step for automation (example described as “NAT/Make-style”)

Example Workflow

  • Take a search query → ask Perplexity for a news/research summary
  • Convert the result into a podcast
  • Email the podcast to the speaker
  • Listen via audio playback at faster speed

Main Speakers / Sources Mentioned

  • Main speaker: Tina Huang
  • Perplexity (product)
  • Third-party sources referenced as part of demos/targets:
    • Reddit
    • Yahoo Finance
    • SEC filings (as a finance deep research source target)

Original video