Build real online income with AI

NotebookLM directory — a practical guide

Resource · November 2026 · 4 min read

NotebookLM is Google's free AI research assistant. You upload your own sources — PDFs, web pages, YouTube videos, or pasted text — and it answers questions using only those sources, showing you exactly where each answer came from. That makes it far more trustworthy than a chatbot guessing from memory.

Step 1 — Create a notebook

Go to notebooklm.google.com and sign in with your Google account. Click New notebook. One notebook is one topic — keep a separate notebook for each subject you are learning.

Step 2 — Add your sources

Click Add source and upload PDFs, paste web links, or add YouTube video URLs. You can add up to 50 sources to one notebook. Start with the 3–5 sources that matter most; more is not always better.

Step 3 — Ask questions with citations

Type your question in the chat. Every answer comes with numbered citations — click one to jump to the exact passage it came from. If an answer looks wrong, check the citation before trusting it.

Step 4 — Listen to the audio overview

Click Audio Overview to generate a short podcast-style conversation about your sources. It is a good way to revise while walking or commuting, and it often surfaces connections you missed while reading.

Step 5 — Generate a study guide

From the Studio panel, generate a study guide, a set of FAQs, or a mind map from your sources. Use the study guide to test yourself: cover the answers and try to recall them before checking.

Step 6 — Use it for real learning

Load a course's PDFs into a notebook, then ask it to explain a chapter in simple words, compare two ideas, or quiz you with five questions. The notebook remembers your sources, so you can return to the same topic for weeks.

Tips

• One notebook per topic keeps answers focused.
• Check citations before quoting anything.
• Remove sources you no longer need — stale sources give stale answers.

← Back to resources

AI tools list — the ones worth knowing

Resource · November 2026 · 4 min read

The tools worth knowing, grouped by job. All of them have a free tier unless noted. Pick one tool per job, learn it properly, and ignore the rest until you need them.

Writing and thinking

• ChatGPT — general writing, brainstorming, coding help, and everyday questions.
• Claude — long documents, careful writing, and analysis where accuracy matters.
• Perplexity — research with sources cited, so you can verify what it says.

Images

• Midjourney — stylised, artistic images for covers and creative work.
• ChatGPT image generation — make and edit images by describing changes in words.
• Leonardo — a generous free tier, good for product-style and consistent images.
• Ideogram — images with readable text, useful for posters and thumbnails.

Video

• Google Flow (Veo) — cinematic AI video from text and images.
• Runway — video generation plus editing tools in one place.
• Kling — realistic video at a lower price than most rivals.
• Seedance — realistic UGC-style video, good for ads that look self-recorded.

Voice and audio

• ElevenLabs — realistic voiceovers and voice cloning.
• TTSMaker — free text-to-speech for simple voiceovers.

Automation

• Zapier — connect apps without code; the easiest starting point.
• Make — visual automation with a generous free tier.
• n8n — powerful and free if you host it yourself.

Research and learning

• NotebookLM — learn from your own sources, with citations.
• Gemini — multimodal AI built into Google's tools.

How to choose

Start with the free tier of one tool per job and use it for a real task this week. Pay only when the free tier limits your actual work. Revisit this list every few months — new tools appear constantly, but change your workflow only when a new tool clearly saves you time. Do not collect tools. One tool per job, used well, is worth more than twenty you barely touch.

← Back to resources

GitHub automations — a beginner's guide

Resource · November 2026 · 4 min read

GitHub Actions is a free automation runner. You write a small text file describing what should happen and when, and GitHub runs it for you — on a schedule, or whenever something changes. No server to manage, no code to deploy.

Step 1 — Create a repository

Go to github.com, sign in, and click New repository. Name it my-automations and choose Public or Private. This repository will hold your workflow files.

Step 2 — Create the workflow folder

Inside the repository, add a folder named .github and inside it a folder named workflows. Workflow files live here, and GitHub finds them automatically.

Step 3 — Add a workflow file

Create a file like daily.yml. It has three parts: a trigger (on: schedule with a cron line — 0 9 * * * means every day at 9am UTC), a job (runs-on: ubuntu-latest, which is GitHub's free server), and steps (for example, actions/checkout to get your files, then a run command or a ready-made action). Start from a template and change only the schedule and the command.

Step 4 — Use secrets for keys

Go to Settings, then Secrets and variables, then Actions, and add your keys as repository secrets. Refer to them in the file as secrets. Never type a key directly into a workflow file — anyone who can read the repository could use it.

Step 5 — Test manually

Add workflow_dispatch to the triggers so you can run it by hand. Then go to the Actions tab, select the workflow, and click Run workflow. Watch it run once before trusting the schedule.

Step 6 — Read the logs

After each run, go to the Actions tab and click the run to read its logs. The logs show every step and any error. Fix problems there — the log tells you exactly which step failed and why.

Three automations to start with

• A daily digest: fetch an RSS feed and email the headlines to yourself.
• A scheduled post: send a file's content to a Zapier webhook for publishing.
• A weekly backup: copy a folder to a second repository automatically.

Rules

• Start with one automation and test it by hand first.
• Keep secrets out of files — use the Secrets settings.
• Check the logs weekly; a silent failure is the only real risk.

← Back to resources

How to start earning with AI this week

Blog · November 2026 · 5 min read

Most people wait for the perfect skill, the perfect tool, or the perfect moment. None of those arrive on their own. What works instead is much smaller: one service, three samples, and ten people who might pay for it. This is a day-by-day plan you can follow this week, even if you have never sold anything online before. No course to buy first, no audience needed.

Day 1 — Pick one service, not five

Start with a service you can already deliver in an afternoon. The most reliable first services are the ones businesses already pay for:

• AI-written social media posts and captions for a local business
• AI video ads and product videos for a shop or a brand
• AI product photos and portraits for sellers and creators
• AI chatbots and simple automations for a small business

Pick the one that feels easiest. Your first clients do not pay for your tools. They pay for the result: posts that go out, ads that look professional, hours saved.

Day 2 — Learn the tool, not the theory

Spend two hours doing, not watching. Open ChatGPT or Claude and write five real captions for a real business. Or launch an image or video tool and make one ad. The goal is a finished piece you can show someone, not a certificate. If you get stuck, ask the AI itself how to do the next step — that is part of the skill.

Day 3 — Make three samples

Pick one type of business — a cafe, a gym, a boutique, a coaching page — and make three finished samples for it. Use a made-up business name if you like. Samples do the selling for you: they prove you can do the work before anyone pays you. Save them as images or short videos you can send in a message.

Day 4 — Put a price on it

Keep pricing simple. Charge per piece or per project, not per hour. A reasonable starting point is ₹500–2,000 per post or caption set, and $25–100 for a short AI video, depending on the client and the work involved. Say the price clearly and in writing. A clear price makes you look professional, and it filters out people who were never going to pay.

Day 5 — Tell ten people

Message ten businesses directly — on Instagram, WhatsApp, or email. Keep it short: who you are, what you make, one sample attached, and your price. Also post your samples publicly once, and list a simple gig on Fiverr or Upwork. Ten honest messages beat a hundred silent profile views.

Day 6–7 — Deliver well, then ask

When someone says yes, deliver fast and deliver well. Then ask for two things: a short testimonial you can show the next client, and a referral to one other business. One happy client who talks about you is worth more than any ad you could buy.

What not to do

Do not buy a course before your first client. Do not wait until you feel like an expert. Do not offer five different services at once — it makes you look unsure. And do not work for free beyond your three samples; free work rarely turns into paid work.

The goal of week one is not riches. It is proof: one person paid you for AI work. Everything after that is repetition — better samples, better prices, more clients. Start on day one.

← Back to all posts

5 AI agents you can build without code

Blog · November 2026 · 5 min read

An AI agent is simply an AI that does a repeatable job on its own. You give it a task, a few rules, and a place to send the result. No coding is needed: you connect tools with Zapier or Make and write the instructions in plain English. Here are five agents that solve real business tasks, and the first step for each one.

1. Inbox triage agent

It reads your incoming email, sorts it by urgency, and drafts replies for you to approve. You stop drowning in your inbox and only touch the messages that need a human.

• Tools: Gmail, ChatGPT or Claude, Zapier
• First step: write down your rules — what counts as urgent, what tone replies should use, and what should never be answered automatically.

2. Research digest agent

Every morning it gathers news and updates on your niche — competitors, trends, prices — and sends you a short summary. You stay informed in five minutes instead of scrolling for an hour.

• Tools: RSS feeds or Google Alerts, Claude, email or Telegram
• First step: list five sources worth following and the three questions you want answered every day.

3. Content repurposing agent

You paste in one long post, article, or video transcript, and it turns it into a thread, a caption, and a short video script — each in the right style for its platform.

• Tools: ChatGPT or Claude, a shared document
• First step: write one master prompt that lists the exact formats you want and the tone to use.

4. Lead-list agent

It finds businesses in your niche, checks their websites and social profiles, and builds a shortlist with contact details — so your outreach starts with people who actually need your service.

• Tools: Perplexity or Claude, a spreadsheet
• First step: describe your ideal client in one paragraph, including the signs that they can pay.

5. Social scheduling agent

It takes your ideas, drafts a week of posts, and queues them for your approval. You review once a week instead of writing every day.

• Tools: ChatGPT or Claude, Buffer or Zapier
• First step: approve every post yourself for the first month. Automate publishing only after the drafts are consistently good.

Three rules that keep agents useful

• Start with one agent, not five. One working agent beats five half-built ones.
• Approve everything at first. Let the agent earn trust before it acts alone.
• Give it written rules and review its work every week. An agent is only as good as its instructions.

You do not need five agents. You need one that saves you an hour a day. Build that one first, and the rest will follow.

← Back to all posts

Automating your content with GitHub

Blog · November 2026 · 5 min read

GitHub is not only for programmers. It is a free place to store your content as plain text files and run small automations on a schedule. This tutorial sets up a simple content pipeline in about 20 minutes. You will mostly copy and paste, and you do not need any coding experience.

Step 1 — Create a repository

Go to github.com, sign in with a free account, and click New repository. Name it content-pipeline and choose Private if you do not want it public. A repository is just a folder that lives online and keeps a history of every change.

Step 2 — Add your content

Inside the repository, create a folder called posts. Add one Markdown file per post — a plain text file with a title, the text, and the date you want it published. Markdown is simple: a line starting with # is a heading, and the rest is just text. Writing your content as files means you own it and can reuse it anywhere.

Step 3 — Add the automation

Create a folder called .github/workflows and add a file named publish.yml. This file tells GitHub what to run and when. It has three parts: a trigger (a schedule, written as a cron line — for example, 0 9 * * * runs every day at 9am UTC), a job (it runs on GitHub's free servers), and steps (for example, send the post to a publishing tool like Zapier or Buffer through a webhook). Start from a template and change only the schedule and the webhook address.

Step 4 — Add your secrets

Go to Settings, then Secrets and variables, then Actions, and add your keys — for example, your Zapier webhook address. Refer to them in the workflow file as secrets. Never paste a key directly into the file itself; anyone who can read the repository could then use it.

Step 5 — Test it by hand

Before the schedule takes over, run the workflow once yourself. Add workflow_dispatch to the triggers, then go to the Actions tab, select the workflow, and click Run workflow. Check that the post actually went out. If it did not, the run's log will show the error — fix it there, not by guessing.

Step 6 — Let it run

Once one manual run works, leave the schedule on. Add new posts to the posts folder whenever you write them, and the automation will publish them on time. Check the Actions tab once a week to make sure the runs are green.

What this gives you

One place for all your content, a publishing schedule that runs without you, and a complete history of everything you have posted. If a tool changes, you only change one file — your content stays yours.

Keep it simple

Start with one platform and one scheduled post a week. Add a second platform only after the first one has run smoothly for a month. The goal is a system you forget about because it works, not a system you have to babysit.

← Back to all posts