Module 03Lesson 1

Lesson 1. AI Companies and Their Ecosystems

Theory

Lesson 1. AI Companies and Their Ecosystems#

Why This Matters#

Before choosing tools, it's important to understand the landscape: who builds AI, how they differ, where they're strong and weak. This helps you:

  • avoid overpaying for tools
  • pick solutions for specific tasks
  • understand who you're trusting with your data
  • stay informed about where the market is heading

The Big Six AI Companies#

In 2026, the AI chat and model market is shared by several major players. Each has its own ecosystem: chat, models, API, and additional products. Let's look at each.


OpenAI — Market Leader

Company overview: Founded in 2015, CEO Sam Altman. In a few years it grew from a nonprofit lab to the market leader with a valuation over $150 billion. ChatGPT is the world's most popular AI chat (300+ million users).

Main products:

  • ChatGPT — universal AI chat (text, images, voice, code, files)
  • GPT-5.2 / GPT-5.3 — model lineup (GPT-5.2 — flagship, GPT-5.3-Codex-Spark — for real-time coding)
  • DALL-E 3 — image generation from descriptions (built into ChatGPT)
  • Sora — video generation from text descriptions
  • Whisper — audio-to-text transcription (open-source)
  • Operator — AI agent that controls the browser for you (bookings, form filling)
  • Codex CLI — AI agent for programming in the terminal
  • Custom GPTs — create your own AI assistants without code
  • ChatGPT Store — marketplace of ready-made assistants

Strengths:

  • Largest ecosystem: thousands of Custom GPTs, plugins, integrations
  • Best generalist: handles most tasks well
  • Voice mode — talk to AI like a person
  • Operator — performs browser tasks for you
  • Huge community and learning resources in all languages

Weaknesses:

  • Pro subscription is expensive ($200/month)
  • Occasional "hallucinations" (made-up facts), though GPT-5.2 has significantly reduced this
  • Privacy concerns: data may be used for training (can be disabled)
  • Slowdowns during peak load

For business: ChatGPT Enterprise and Team — for teams. Custom GPTs — quick way to create an AI assistant for clients without a single line of code. API — for integrating AI into your products and agents.

For life: ChatGPT free — for texts, ideas, translations, document analysis. Voice mode — smart companion in your pocket. DALL-E — images on demand right in the chat.

Pricing:

  • Free: GPT-4o mini, basic features
  • Plus ($20/month): GPT-5.2, DALL-E, Sora, Advanced Voice
  • Pro ($200/month): unlimited access to GPT-5.2 Pro and GPT-5.3, Operator, Codex CLI, priority

Link: chatgpt.com


Google (DeepMind) — AI Inside the Ecosystem

Company overview: Google DeepMind is the world's largest AI lab, merging Google Brain and DeepMind. Unique advantage: AI is built into Google Workspace, Android, Chrome, YouTube — what billions already use.

Main products:

  • Gemini — AI chat (ChatGPT alternative), built into Google services, 650+ million users
  • Gemini 3 Pro / Flash / Deep Think — model lineup (Pro — flagship, Flash — fast, Deep Think — for complex reasoning)
  • NotebookLM — AI researcher for working with documents (more below!)
  • Google Workspace AI — AI in Docs, Sheets, Gmail, Slides, Meet
  • Veo — video generation
  • Imagen — image generation
  • Project Astra — multimodal AI assistant (sees via camera, hears, responds)
  • Jules — AI agent for writing code
  • Gemini CLI — AI agent for working with code and tasks in the terminal

Strengths:

  • Deepest integration with Google Workspace (Docs, Sheets, Gmail, Calendar, Meet)
  • NotebookLM — unique product with no direct equivalents
  • Best multimodal understanding (text + images + video + audio at once)
  • Huge context window (up to 2M tokens in Gemini 3 Pro)
  • Generous free tiers

Weaknesses:

  • Less "creative" in text generation compared to ChatGPT
  • Fewer third-party integrations and plugins
  • Google Workspace AI — paid add-on
  • Sometimes abruptly changes or discontinues products

For business: Gemini for Workspace — AI right in documents, spreadsheets, email. NotebookLM — for analyzing large volumes of information (contracts, research, tender documentation).

For life: Gemini free on phone and in browser. NotebookLM free — upload a book or lecture and get a summary, FAQ, or podcast.

Pricing:

  • Free: Gemini 3 Flash, NotebookLM
  • Gemini Advanced ($20/month): Gemini 3 Pro, Deep Think, advanced features, 2M context
  • Google Workspace AI ($20/month): AI in Docs, Sheets, Gmail

Link: gemini.google.com, notebooklm.google.com


Anthropic — Safety and Depth

Company overview: Founded in 2021 by Dario and Daniela Amodei (former senior OpenAI staff). They left to build "safer AI." Their model Claude is one of the best AI for working with long documents, analysis, and code.

Main products:

  • Claude — AI chat with huge context (200K tokens standard, up to 1M in beta ≈ 2,500 pages of text at once)
  • Claude Opus 4.6 — most powerful model (released Feb 5, 2026), leader in coding and reasoning
  • Claude Sonnet 4.5 — fast model for everyday tasks
  • Claude Projects — organize work: upload files, give instructions, manage projects
  • Claude Artifacts — creates documents, spreadsheets, code, visualizations right in chat
  • Claude Code — AI agent for programming in the terminal (more in Lesson 2)

Strengths:

  • Best for long documents: upload a 500-page PDF — Claude reads and answers (up to 1M tokens in beta)
  • Excellent quality of analysis and structuring
  • Most "honest" AI: hallucinates less, admits ignorance more often
  • Opus 4.6 — leader in agentic coding and reasoning (beats GPT-5.2 on benchmarks)
  • Projects — convenient context organization (files + instructions in one place)
  • Artifacts — creates interactive artifacts (documents, visualizations) right in chat

Weaknesses:

  • No image generation (only text, code, visualizations)
  • Fewer integrations than OpenAI
  • More "cautious": sometimes refuses to answer acceptable requests
  • Free tier has fairly strict message limits

For business: Ideal for analyzing contracts, reports, strategic documents. Claude Projects — for managing client projects with persistent context. Claude Code — for building prototypes.

For life: Best assistant for study and document work. Upload a textbook — get a summary and answers to questions. Great for writing and editing texts.

Pricing:

  • Free: Claude Sonnet 4.5, limited requests
  • Pro ($20/month): Claude Opus 4.6, increased limits, Projects
  • Max ($100/month): maximum limits, priority access to Opus 4.6
  • Team ($25/month per person): team features, Admin panel

Link: claude.ai


xAI — AI from Elon Musk

Company overview: Founded in 2023 by Elon Musk. Goal — "understand the true nature of the universe." Quickly became a leader thanks to Musk's resources and integration with social network X (formerly Twitter).

Main products:

  • Grok — AI chat built into X/Twitter
  • Grok 3 — one of the most powerful models in 2025–2026
  • Grok API — for developers and integrations

Strengths:

  • Real-time access to X/Twitter data — knows what's being discussed right now
  • Less "censored": answers questions where other models refuse
  • Good sense of humor and informal communication style
  • Great for trend analysis and current events
  • Powerful Grok 3 model competing with GPT-5

Weaknesses:

  • Tied to X ecosystem (X Premium subscription required)
  • Significantly fewer integrations and additional tools
  • Reputational risks (connection with Musk polarizes audience)
  • Less developed ecosystem compared to OpenAI/Google

For business: Social media monitoring, trend analysis, brand mention tracking on X, real-time news analysis.

For life: Up-to-date answers about events, unconventional perspective on questions, humor, discussion of controversial topics.

Pricing:

  • Included in X Premium ($8/month) and X Premium+ ($16/month)
  • Grok API — separate pricing

Link: x.ai, available in X app


Chinese AI Companies — Budget Alternatives

Summary: China is the world's second AI pole after the US. Chinese models cost 5–20x less than Western ones at comparable quality. Great option for experiments, prototypes, and budget-limited projects.

DeepSeek (DeepSeek AI)

  • What: Powerful model with open weights and transparent reasoning
  • Best model: DeepSeek-R1 (685B parameters)
  • Pros: 3–15x cheaper than GPT-5, visible reasoning chain, excellent code and logic
  • Cons: Weaker on creative tasks, limited documentation in Russian
  • For what: Programming, logic tasks, data analysis, budget projects
  • Price: ~$0.003 per 1K input tokens
  • Link: deepseek.com

Qwen (Alibaba Cloud)

  • What: Model family from Alibaba — one of China's leading developments
  • Best model: Qwen 3 (multimodal, strong reasoning)
  • Pros: Excellent multimodality, good API, active development
  • Cons: More focused on Chinese and English markets
  • For what: General tasks, multimodal analysis, budget alternative to GPT
  • Price: from $0.002 per 1K input tokens
  • Link: qwen.ai

Kimi (Moonshot AI)

  • What: Model focused on ultra-long context (up to 500K tokens without loss)
  • Best model: Kimi K2 (500B parameters)
  • Pros: Huge context window, personalization, good agent intelligence
  • Cons: More specialized for Chinese market
  • For what: Long documents, agent tasks, personalized assistants
  • Price: ~$0.005 per 1K input tokens
  • Link: kimi.ai

GLM (Zhipu AI)

  • What: Model family from Zhipu AI — one of China's leading AI startups
  • Best model: GLM-4.5 (355B parameters)
  • Pros: 3rd place in global agent task rankings, good reasoning and coding
  • Cons: Less documentation in English and Russian
  • For what: Agent tasks, reasoning, coding
  • Price: ~$0.004 per 1K input tokens
  • Link: zhipuai.cn

MiniMax

  • What: Models with giant context — up to 4 million tokens!
  • Best models: MiniMax-Text-01 (456B), MiniMax-VL-01 (visual tasks)
  • Pros: World's largest context window, multimodality
  • Cons: Less known in the West, smaller community
  • For what: Working with huge documents, visual analysis
  • Price: ~$0.004–0.006 per 1K input tokens
  • Link: minimaxi.com

When to use Chinese models:

  • Budget projects and prototypes (up to 90% savings)
  • High-volume request tasks (where price per token is critical)
  • Experiments and learning
  • Programming and data analysis (where they match GPT-5)

When NOT to use:

  • Creative marketing and copywriting (Western models still better)
  • Tasks requiring deep understanding of Russian/Western culture
  • Projects where brand reputation is critical (clients may be wary)

Perplexity — Next-Gen AI Search

Company overview: Founded in 2022 by Aravind Srinivas. AI search engine that doesn't just find links — it gives ready answers with sources. Quickly gaining popularity as a Google replacement for complex questions.

Main products:

  • Perplexity AI — AI search: ask a question → get an answer with sources
  • Perplexity Browser — browser with built-in AI (more below!)
  • Perplexity Spaces — collaborative research spaces
  • Perplexity API — for developers and integrations

Strengths:

  • Answers with sources — you always see where information comes from
  • Up-to-date data from the internet (unlike ChatGPT, limited by training date)
  • No need to phrase queries "correctly" — understands natural language
  • Great for quick research on any topic

Weaknesses:

  • Full features only with subscription ($20/month)
  • Sometimes inaccurate or outdated sources
  • Doesn't create content (not a ChatGPT replacement for text generation)
  • Less developed ecosystem

For business: Market research, competitive analysis, finding statistics and facts, fact-checking, preparing for negotiations.

For life: Google replacement for complex questions. "What laptop to buy under $1,000 in 2026?" — get a detailed answer with reviews and links.

Pricing:

  • Free: basic search, limited queries
  • Pro ($20/month): unlimited, advanced models, file uploads

Link: perplexity.ai


Special Focus: Three Breakthrough Products#

NotebookLM — Your AI Researcher

What it is: Free tool from Google that turns any documents into an interactive knowledge base. You upload PDFs, articles, YouTube videos — and get an AI assistant that knows their content and answers only based on them.

How it works:

  1. Upload sources (PDFs, Google Docs, website links, YouTube videos) — up to 50
  2. NotebookLM "reads" everything and builds a knowledge base
  3. Ask questions — answers are only based on your documents (no hallucinations!)
  4. Generates ready artifacts: summaries, FAQ, timelines, study materials
  5. Key feature: creates an AI podcast — two voices engagingly discuss your documents' content

Practical scenarios (for business):

  • Upload contracts → NotebookLM finds key terms and risks
  • Upload 10 reports → ask questions across all: "what trends repeat?"
  • Upload meeting notes → get a briefing before negotiations
  • Upload competitor documentation → get structured analysis

Practical scenarios (for life):

  • Upload a textbook → get a summary, FAQ, and podcast for review
  • Upload articles on a topic → get a structured overview
  • Upload recipes → ask "what to cook with chicken and mushrooms?"

Why it's revolutionary: NotebookLM doesn't make things up — it answers ONLY based on your documents. Every answer includes a link to the specific source passage. This radically solves the hallucination problem.

Price: Completely free.

Link: notebooklm.google.com

Tutorial videos:


Perplexity Browser (Comet) — AI Inside the Browser

What it is: The first browser where AI is built into every action. No need to switch between browser and AI chat — the assistant works right while you're viewing pages.

How it works:

  1. Open any page — AI summarizes it with one click
  2. Select text — AI explains, translates, expands
  3. Type a question in the address bar — get an AI answer with sources (not just links)
  4. AI "sees" the current page content and answers questions about it
  5. Built-in AI assistant available on every tab

Practical scenarios:

  • Learning a topic: Opened an article → AI creates a summary → ask follow-up questions
  • Shopping: Comparing products → AI summarizes all reviews on the page
  • Business analytics: Reading competitor report → AI highlights key figures
  • Translation and context: Opened a foreign page → AI translates and explains context
  • Reading news: Opened an article → "what's the gist?" → get a brief summary

How it differs from a regular browser: In Chrome you search → click links → read yourself → form an opinion. In Perplexity Browser you ask → get a ready answer → dig deeper only if needed.

Price: Browser is free, full AI features — with Perplexity Pro subscription ($20/month).

Link: perplexity.ai/browser

Tutorial videos:


Manus AI — Autonomous AI Agent

What it is: Next-gen AI agent that performs complex tasks on its own. Unlike regular chats (where you guide the dialogue step by step), Manus receives a task — and does it entirely: plans, searches, creates documents, writes code.

How it works:

  1. You describe the task in plain language
  2. Manus breaks it into steps and makes a plan
  3. Independently searches the internet, opens sites
  4. Creates artifacts: documents, spreadsheets, presentations, code, sites
  5. Delivers a ready result — not text in chat, but files and links

Practical scenarios:

  • "Research the food delivery market in Moscow and create a report with charts" → Manus searches data, analyzes, and creates a PDF report
  • "Make a landing page for my beauty salon" → Manus creates a working site
  • "Find 50 potential clients in Moscow for a B2B service" → Manus searches, filters, and creates a spreadsheet
  • "Analyze reviews of our product and create an improvement plan" → Manus collects reviews, analyzes, and makes a plan

Strengths:

  • Autonomy: no need to guide the dialogue step by step
  • Creates ready artifacts (files, sites, spreadsheets) — not just text
  • Can work with the browser (searches, opens pages, downloads)
  • Performs complex multi-step tasks

Weaknesses:

  • Sometimes unpredictable (may go the wrong way and waste time)
  • More expensive than regular chats
  • Complex tasks take a long time (minutes, not seconds)
  • Periodically limited availability (sometimes a queue)

Price: Free tier (limited requests), Pro from $39/month.

Link: manus.im

Tutorial videos:


Comparison: NotebookLM vs Perplexity vs Manus AI#

These three products solve different problems, though they may seem similar at first glance:

CriterionNotebookLMPerplexity (+ Browser)Manus AI
Main taskAnalyzing your documentsSearching information on the internetAutonomous task execution
Data sourceOnly files you uploadThe whole internet in real timeInternet + creating new content
What it createsSummaries, FAQ, podcastsAnswers with source linksDocuments, sites, spreadsheets, code
AutonomyLow (you ask questions)Medium (searches itself, you guide)High (completes the whole task)
HallucinationsMinimal (relies on your data)Low (cites sources)Medium (long action chains)
PriceFreeFree / $20/monthFree (limit) / from $39/month
Best forStudy, document analysisQuick fact-findingEnd-to-end complex tasks

Rule of thumb:

  • Need to understand your documents → NotebookLM
  • Need to quickly find an answer on the internet → Perplexity
  • Need AI to do the whole task for you → Manus AI

AI Companies Summary Table#

CompanyMain chatStrengthFor businessFor lifeFree?
OpenAIChatGPTVersatility, ecosystemCustom GPTs, EnterpriseTexts, voice, imagesYes
GoogleGeminiGoogle integration, NotebookLMWorkspace AIGemini, NotebookLMYes
AnthropicClaudeDocuments, analysis, honestyProjects, TeamStudy, documentsYes
xAIGrokRecency, X dataSocial monitoringTrends, humorVia X
ChinaDeepSeek, Qwen, KimiPrice (5–20x cheaper)Budget projectsSavingsYes
PerplexityPerplexity AISearch with sourcesAnalytics, researchGoogle replacementYes