Now we’re going to use it to build something real, from scratch, in one sitting. You don’t need any coding experience. You’ll describe what you want in plain English and Claude Code will build it for you.

Here’s what we’re building: a personal AI-powered morning newsletter. Every morning, before you even wake up, a polished email lands in your inbox with the best stories from across finance, geopolitics, tech, and the creator economy, curated and written up just for you. You don’t have to do anything. It just shows up, and it’s completely free.

Under the hood, your morning digest pulls hundreds of articles from 15 news sources, sends them to Google Gemini, and Gemini reads through everything, picks the best stories, and writes them up into a polished email for you.

I built this for myself, and I’m going to walk you through exactly how to build your own using Claude Code.

Words you’ll see in this project

A few terms will come up as we go. Here’s what they actually mean.

Script: A script is a plain text file that contains a set of instructions your computer can follow, like a recipe. Ours will say “go fetch raw articles from these 15 news websites, hand them to Google Gemini so it can read and summarize them, get the finished digest back, and email it to me.” It lives in a folder on your computer, just like any other file. You don’t write it. Claude Code does. You just tell Claude Code what you want the recipe to do.

API: A way for two apps to talk to each other behind the scenes. You use APIs all the time without knowing it. When you order an Uber, the Uber app talks to Google Maps to get your route, and talks to Stripe to charge your card. Those conversations between apps happen through APIs. In our case, the API is the bridge between your script and Google Gemini. Your script passes raw articles across that bridge, Gemini processes them into a polished digest, and passes the result back.

GitHub: A website where people store code online. Think of it like Google Drive, but for software projects. We’re using it here because GitHub can run your script every morning on their servers, so your computer doesn’t even need to be on.

Repository: A project folder on GitHub. Same concept as a folder on your computer, just hosted online.

GitHub Actions: A free feature on GitHub that can run your script on a schedule. It’s what makes the “every morning at 7 AM” part work without your computer being involved.

RSS feed: A behind-the-scenes data stream that most news websites have. It’s how your script reads articles from 15 sources without opening 15 browser tabs. You don’t need to set these up. The script handles it automatically.

.env file: A file that stores your private information (like API keys and passwords) separately from the rest of your project. Think of it like a sticky note in your desk drawer. Only you can see it. This matters later when your project goes on GitHub, where other people could technically see your files. The .env file stays private.

Before We Start

You’ll need three things beyond Claude Code that we setup last time:

1. A Google Gemini API key. This is free. It’s what lets your script talk to Google’s AI behind the scenes to write your digest. Go to aistudio.google.com, sign in with your Google account, and grab an API key. No credit card needed.

2. A Gmail account. The digest gets emailed to you. You’ll need to create something called a Gmail App Password (we’ll cover this in the steps below).

3. A GitHub account. This is where your project will live online so it can run automatically every morning without your computer being on. Free to sign up at github.com. (we’ll cover this in the steps below).

That’s all you need. Let’s get into it.

Step 1: Create Your Project Folder

Open VS Code or Cursor (whichever you installed last time). I’ll be using VS Code, so that’s what I’ll reference throughout.

Create a new folder called morning-digest somewhere on your computer. Open that folder in VS Code (File > Open Folder).

Now open the terminal inside VS Code (View > Terminal) and type claude to start Claude Code.

You should see a blinking cursor. That’s Claude Code, ready for you to tell it what to build.

Step 2: Tell Claude Code What You Want

Now you tell Claude Code what you want it to build. Here’s what I told it. Steal this and paste it in Claude Code.

Build me a Python script that does the following:

1. Pulls recent articles from RSS feeds across four categories: finance (CNBC, MarketWatch, Yahoo Finance, BBC Business, Guardian Business, Axios), geopolitics (NPR World, Foreign Policy, Deutsche Welle), tech (Hacker News, TechCrunch, The Verge, Wired), and creator economy (Digiday, Social Media Today)

2. Sends all the articles to Google Gemini using the Gemini API with a prompt that says “act as a sharp, witty friend who actually reads the news” and asks it to write a digest with these sections: Money Talk, World Lore, Tech Tea, Creator Szn, and a Speed Round of quick one-liners

3. Formats the digest as a polished HTML email with a dark header, color-coded sections, and clean design

4. Sends it to me via Gmail using SMTP

Use a .env file for all secrets (Gemini API key, Gmail address, Gmail app password, recipient email). Put everything in a single file called digest.py.

Hit Enter and let Claude Code do its thing.

It’ll ask you for permission along the way, things like “can I create this file?” or “can I run this command?” Hit yes every time. That’s just Claude Code checking in before it does something on your computer.

You’ll see it start creating files, installing things, and writing code in real time in VS Code. It’s pretty cool to watch the first time.

When it’s done, you’ll have a working Python script and a few supporting files.

Step 3: Set Up Your Secrets

Your script needs four pieces of private information to work. Claude Code created a file called .env in your project folder to hold all of them. This is the locked settings file we mentioned earlier. Your passwords and keys go here so they stay separate from everything else.

Open the .env file and you’ll see four lines. Here’s what each one is, why your script needs it, and how to get it.

GEMINI_API_KEY=your-api-key-here
GMAIL_ADDRESS=your-email@gmail.com
GMAIL_APP_PASSWORD=your-app-password-here
RECIPIENT_EMAIL=where-you-want-it-sent@gmail.com

GEMINI_API_KEY: This is what lets your script talk to Google Gemini behind the scenes. Without it, your script can fetch all the raw articles it wants, but it has no way to hand them off to Gemini for processing. This key unlocks that connection so your script can pass articles to Gemini and get a finished digest back.

To get it: Go to aistudio.google.com and sign in with your Google account. Click “Get API key” and create one. It’s completely free. No credit card, no credits to buy. Google’s free tier is more than enough for one digest a day.

GMAIL_ADDRESS: The Gmail address your digest will be sent from. Your script needs a Gmail account to act as the sender. You can use your main Gmail or create a new one just for this. Either works.

To set it: Just type your full Gmail address (e.g., yourname@gmail.com). Nothing else to configure.

GMAIL_APP_PASSWORD: A special one-time password that lets your script send emails through your Gmail account. Google doesn’t let outside software use your regular Gmail password for security reasons. So instead, you create a separate “app password” that’s only used by your script.

To get it: Go to myaccount.google.com. Click Security. Turn on 2-Step Verification if it’s not already on. Then search for “App Passwords” in your Google Account settings. It’ll ask you to name it - type “Mail.” Google will give you a 16-character password. It might show up with spaces (like “abcd efgh ijkl mnop”). Remove the spaces before pasting it here.

RECIPIENT_EMAIL: The email address where you want to receive your morning digest. This can be the same as your GMAIL_ADDRESS above, or a completely different email address. Whatever inbox you check first thing in the morning.

Step 4: Test It

Go back to Claude Code and tell it:

Run the script so I can test it.

If everything is set up right, you’ll get an email within a minute or two. That’s your first AI-generated morning digest.

Step 5: Tweak the Tone

That first version might not be exactly what you want. Mine wasn’t. The first draft was way too heavy on Gen Z slang. “Bestie” and “it’s giving” in a finance briefing. Funny, but not what I wanted to read every morning.

So I went back to Claude Code and said:

The tone is too much. Dial back the Gen Z energy. I want casual but not chaotic. Witty but trustworthy. Keep it conversational but cut the slang.

Claude Code updated the prompt inside the script. I ran it again and it was much better.

That’s basically what vibe coding is. You don’t need to understand what changed in the code. You just tell it what you want different and it adjusts. It’s like giving feedback to a designer. “Make the logo bigger” doesn’t require you to know Photoshop.

You can keep iterating from here. If you want different news sources, tell it. If you want a different email design, tell it. Each tweak takes a couple of minutes.

Step 6: Set Up GitHub

If you already have a GitHub account and Git installed on your computer, skip to Step 7.

Right now, your script lives on your computer. It works, but you have to manually run it every time. To make it run automatically every morning, we need to put it on GitHub so GitHub’s servers can run it for you, even while your laptop is closed.

Two things to set up:

1. Create a GitHub account

Go to github.com and sign up. It’s free. This is where your project will live online.

2. Install Git

Git is a tool that lets you upload your project files from your computer to GitHub. You don’t need to learn it. Claude Code handles all the commands. You just need it installed.

Go to Claude Code and tell it:

Install Git on my computer.

It’ll figure out whether you’re on Mac or Windows and install it for you. Hit yes when it asks for permission. You don’t need GitHub Desktop or any other app.

3. Connect your computer to GitHub

The first time you push your project to GitHub, a browser window will pop up asking you to sign in to your GitHub account. Just log in and approve the connection. This only happens once. After that, your computer remembers.

Step 7: Make It Run Every Morning (Automatically)

Now that GitHub is set up, we can make your digest run on autopilot.

Tell Claude Code:

Set this up to run automatically every morning at 7 AM Eastern time using GitHub Actions. Create a GitHub repository for the project and give me the steps to push it up and configure it.

Claude Code will create the automation file and walk you through uploading your project to GitHub. Here’s the gist of what happens:

1. Claude Code creates a repository on GitHub for you (a folder in the cloud for your project)

2. It uploads your project files to it

3. You add your secrets (Gemini API key, Gmail password) to GitHub’s secure settings so they’re not visible to anyone

4. GitHub runs your script every morning on their servers. Your computer doesn’t need to be on

Claude Code will give you the exact commands to type. Just follow them one by one.

What You Just Built

So here’s what you just built. A fully automated product that:

• Reads 100+ articles across 15 news sources every morning

• Uses AI to curate and write a personalized digest

• Emails it to you in a polished format

• Runs on its own without you touching anything

• Costs nothing

You did it without writing code, without hiring a developer, and without paying for a SaaS tool. All in under 2 hours. This is your first vibe coding project, and it probably won’t be your last.

If You Get Stuck

Things might not work perfectly on the first try, and that’s normal. Here’s what to do:

If something breaks, just tell Claude Code what happened. Copy the error message and paste it in. Say “I got this error, fix it.” Nine times out of ten, it’ll know exactly what went wrong and fix it.

If the email doesn’t arrive, check your spam folder first. Then check that your Gmail App Password is correct and that 2-Step Verification is turned on.

If Claude Code seems confused, you can always say “let’s start over” and re-describe what you want. No progress is lost. Your files are still there.

The whole point of vibe coding is that you don’t need to debug anything yourself. You just describe the problem and let Claude Code figure it out.

We’ll keep sharing projects like this. If you want to keep building, subscribe and follow along.

Make AI do the work. Get Superlinear.

Written by Abhilash and Alisha

Three months ago we cancelled it and moved to Claude. Same $20 a month.

It’s not that ChatGPT is bad. It’s that OpenAI has been so busy chasing consumer features and flashy demos that they’ve fallen behind on the stuff that actually matters for people who use AI to do real work. Claude’s products for knowledge workers and coding are just better right now. Sharper writing, deeper reasoning, and an ecosystem of tools that actually fits how professionals work.

That last part is the one nobody talks about. Claude isn’t just a chatbot. It’s actually three different tools - Claude Chat, Claude Cowork, and Claude Code - each designed for a different kind of work. All included in the same $20 subscription. And figuring out which one to use when completely changed how both of us work.

This post breaks down what each tool does, when to use which one, and how I use all three across my professional and personal projects. If you’ve been using AI but feel like you’re not getting the most out of it, this is for you.

A bit of context on what I do: I lead AI strategy work where I help PE investors, hedge funds, and tech companies figure out where the AI market is heading and how to position for it. That means a lot of research, client prep, financial models, and increasingly, building custom tools and automations for myself and my team.

What AI tools are you using right now - Claude, ChatGPT, Gemini, something else? I’d love to know more. Drop it in the comments.

It started with Claude Chat

Claude Chat is the one everyone knows. Go to claude.ai, type a message, get a response. Same concept as ChatGPT. I still use it every day for quick things:

Refining important emails before sending. Quickly understanding key technical terms of an industry. Brainstorming investment angles. Stress-testing a thesis.

For anything quick and self-contained, Claude Chat is perfect. Fast, frictionless, no setup.

But if you use AI seriously, you’ve hit the wall I hit. I was trying to prep for client calls by uploading company documents into Claude Chat. Investor decks, annual reports, analyst notes, industry research. And it kept falling apart.

The context window (how much information Claude can hold in its head at once) was too small. It would forget things I told it three messages ago. It would mix in random web search results when I explicitly asked it to only use my documents. The output had that unmistakable AI slop quality - polished paragraphs that sounded smart but contained nothing useful.

And I couldn’t even upload everything I needed. Claude Chat has file size restrictions, especially on PDFs. So I’d be sitting there, feeding it documents one at a time, re-prompting, copying output into Word, reformatting for half an hour. Three hours of work to get something mediocre.

I thought I was bad at prompting. Turns out I was using the wrong tool entirely.

Then I found Claude Cowork

Claude Cowork is a desktop app. You download it, point it at a folder on your computer, and Claude can now see and work with everything in that folder. PDFs, spreadsheets, slide decks, Word docs - all at once, with full context. No uploading one file at a time through a chat window.

This solved every single problem I just described.

Client research before first calls. I take a lot of introductory calls where I need to understand a company’s business cold. Now I just drop everything into a folder - investor decks, annual reports, press releases, analyst notes - and point Claude Cowork at it. One prompt: “Give me a structured research brief on this company.” In a few minutes I have competitive positioning, key financial metrics, recent strategic moves, and talking points for the call. All sourced from my materials. No hallucinations, no random web data mixed in.

This used to take me 2-3 hours. Now it takes 15 minutes. And honestly, the output is more thorough than what I was producing by hand.

Proposals. I keep a folder with our company’s branding materials - templates, past proposals, tone guidelines. When an opportunity comes in, I describe the engagement scope and Claude Cowork produces a formatted proposal that reads like something our firm would actually send.

Financial models. Market sizing, capacity planning, unit economics. I drop the raw data into a folder, describe the model I want, and Claude Cowork builds it as an actual Excel file. Gets 80% of the way there on the first pass. The 80% that used to take a full day.

Presentations. Pre-loaded brand materials plus a clear brief equals a deck that needs light editing instead of a full rebuild from blank slides.

The thing that made me shake my head at how much time I’d wasted: Claude Cowork saves output directly as files on your computer. A Word doc, a spreadsheet, a PowerPoint. No copying from a chat window, no pasting into another app, no reformatting. Just the finished document, ready to go.

Going from Claude Chat to Claude Cowork for research-heavy work saved me 3+ hours per task and the quality jumped by 10x. When Claude reads 15 documents at once versus one document pasted into a chat, it’s a completely different experience.

But then I wanted to build things. And that’s when I hit a different kind of wall.

And then Claude Code changed everything

I had ideas for tools that would make my work and life better. A dashboard that tracks our engagement pipeline. An automated news briefing. A way to analyze massive datasets without spending days in Excel. A tool that helps me find edges in prediction markets.

I tried Claude Cowork for some of this. It was fine for one-off analysis but it felt rigid. When I wanted to iterate, change things, connect different systems together, automate something to run on its own every day - Claude Cowork wasn’t built for that. It’s a research analyst, not an engineer.

Claude Code is the engineer.

It runs in your terminal (the command-line interface on your computer - that black screen with text that developers use). You describe what you want in plain English and Claude Code writes real software, runs it, fixes errors, and iterates until it works. It can connect to APIs (basically how different software systems talk to each other), build databases, create automations that run on their own.

I don’t have a CS degree. I can’t write Python from scratch. Here’s what I’ve built anyway:

An engagement dashboard. Auto-populates every morning with how many first calls I’ve taken this month, how many projects we’ve sold, what team capacity looks like, how many new engagements we can take on. Before this, all of it lived in scattered spreadsheets that nobody bothered to update. I described what I wanted to Claude Code over a couple of hours and now it just runs on its own.

A morning news scanner. Monitors key sources across AI infrastructure - chip announcements, data center deals, power procurement contracts, policy shifts - and delivers a personalized briefing before my first coffee. I used to spend 45 minutes every morning scanning manually. Now the important stuff finds me.

A prediction market analyzer. This one’s my favorite. It pulls in news sentiment data, company earnings documents, and macro indicators (things like interest rates, employment data, energy prices). It calculates its own probability estimates for specific events. Then it compares those estimates against live odds on Kalshi (a prediction market platform where you can trade on real-world outcomes) and flags when there’s a meaningful gap. “The market prices this at 35% but the data suggests 55%, here’s why.” I built this over a normal weekend. A weekend where I also went out to dinner and slept eight hours.

The best way I can describe the difference between Claude Cowork and Claude Code: Claude Cowork gives you a move-in-ready house. Claude Code gives you a move-in-ready house where you can redesign every room.

The Three-Second Decision Tree

After months of using all three daily, this is my mental model:

Quick question or short-form writing / text editing? Claude Chat.

Research, documents, or working with files I already have? Claude Cowork.

Building something, automating something, or heavy-duty analysis? Claude Code.

Pick the wrong one, you’ll know within a minute. Switch and move on. They’re all on the same $20 subscription.

What does your setup look like? Let me know in the comments.

Getting started

All three come with Claude Pro at $20/month:

• Claude Chat: Go to claude.ai. Sign up. Start typing.

• Claude Cowork: Download the Claude desktop app. Make a folder with the files you want Claude to work with. Open Claude Cowork, point it at the folder, and tell it what you need.

• Claude Code: Takes slightly more setup (you install it through your terminal), but once it’s running, you interact with it in plain English just like the other two.

What’s coming next

If Claude Chat and Claude Cowork clicked but the Claude Code section felt like a leap - that’s exactly what our next post is about.

We’re going to walk through setting up Claude Code and building your first vibe code project from scratch. Step by step, no coding background needed. If you’ve ever had an idea for a tool, a dashboard, or an automation that would make your work meaningfully better but couldn’t build it because you “don’t code” - that excuse is about to expire.

Subscribe so you don’t miss it.

Make AI do the work. Get Superlinear.

Written by Alisha and Abhilash

This is our first ever post. We’re SO excited you’re here!!!

And with that, moving on to the real thing.

Last month, a senior associate at a PE fund I’d been meeting with called me. Sharp guy, five years in, knows his stuff. He’d just found out that a first-year analyst on his team had used Claude Code to build a fully functioning 20-tab Excel model with data tables, sensitivity analysis, IRR and NPV calculations, bull and bear cases, key risk areas and mitigating factors, all ready as a draft for the senior MD. In a single afternoon. The kind of work that usually takes 3-4 days and significant hand-holding from the senior associate.

He wasn’t mad at the kid. He was terrified he was falling behind.

He’s not an edge case. He’s the norm. And here’s why.

Career success used to have a formula. Get the right degree. Build expertise over years. Climb the ladder. The people who won were the ones who knew more than everyone else in the room.

That formula is breaking. The expertise that took years to build? AI can now do it in seconds. The research, the analysis, the structured thinking. All of it is becoming instant, cheap, and available to anyone with a laptop.

When everyone has access to expertise, expertise stops being the edge.

So now what matters is agency. Who tackles the hardest problems. Who actually ships. And who does it the fastest.

The consultant who synthesizes 200 pages of market research and delivers a client-ready deck in two hours while others are still reading through it. The marketer who launches, tests, and iterates on 20 campaign angles in a single morning.

The professionals who figure this out will have a massive advantage. And that gap will only widen.

What “AI Native” Actually Means

Being AI native has nothing to do with coding or CS degrees or understanding how neural networks work.

It means AI is woven into how you think, work, and make decisions. Most people think “using AI” means typing a question into ChatGPT. That’s like saying you “use the internet” because you Google things.

Being AI native means using AI agents. Tools that actually do work for you.

I have a Claude workflow that monitors my industry overnight and gives me a personalized briefing before coffee. I use Obsidian (a free tool) for my notes so that all files are saved in .md format and it’s easier for my agent to parse through and generate insights. There are several other examples.

AI stops being a search bar. It becomes a teammate.

We’ve Seen This Before

When electricity first arrived in factories in the 1880s, most owners just swapped out their steam engines for electric motors and kept everything else the same. The factory layout stayed identical. One giant power source in the center, machines arranged around it, connected by belts and shafts. But the output barely changed.

It took a full generation before manufacturers realized the real unlock: electricity meant every single machine could have its own small motor. You didn’t need the central engine anymore. That meant you could rearrange the entire factory floor around the flow of work instead of the flow of power.

Henry Ford was one of the first to get this. His Highland Park plant in 1913 didn’t just use electric motors. It redesigned the entire production process around them. Individual motors powered conveyors, tools, and stations arranged by the sequence of assembly, not by proximity to a power shaft.

The result was the first moving assembly line. The time to build a car dropped from 12 hours to 93 minutes. Meanwhile, competitors who simply bolted electric motors onto their old steam-era layouts saw barely any improvement. By the time they caught up, Ford had already captured the market.

Most people are still in the “swap the motor” phase with AI. They ask ChatGPT the same questions they’d have Googled. They use it to do the same work slightly faster. The real advantage belongs to those who redesign how they work entirely. Who give AI its own role in every part of their workflow, not just bolt it onto the old process.

Every major technology shift follows the same arc: a scarce resource becomes abundant, old advantages fade, and new ones emerge. The people who move first capture the most value.

AI is on the same trajectory. The early mover window is open. It won’t be for long.

The Slow Cost of Doing Nothing

The fear is that AI will take your job. The reality is more subtle.

You won’t get fired for not being AI native. Not tomorrow, anyway.

It’s slower than that. You become the person who takes three days to deliver what your colleague delivers in three hours. You stop getting the interesting projects. The gap compounds every month.

The real threat was never AI taking your job. It was always going to be the colleague who figured it out before you did.

So Why Isn’t Everyone Doing This?

It’s not for lack of interest. Most ambitious professionals have tried ChatGPT, maybe even paid for it. They know this stuff matters. But they also have a job to do, clients to manage, deadlines to hit. They don’t have 40 hours a week to tinker with prompts and figure out what works through trial and error.

And here’s the thing nobody talks about: it’s not as easy as the internet makes it look. Anyone who’s actually tried to build something with AI knows the experience is still messy, full of dead ends and weird failures. The polished demos on Twitter don’t show you the six attempts that didn’t work before the one that did (trust us we’ve tried haha!)

What’s missing isn’t more content. It’s a clear, no-BS playbook that tells you exactly what to do, step by step, so you don’t have to figure it all out yourself.

This Is What Superlinear Is For

Remember that senior associate? Five years of reps, pattern recognition, deals under his belt. And a first-year analyst with the right tools just lapped him in an afternoon.

That’s the new reality. And it’s coming for every industry, every role, every level.

We are building Superlinear to make sure you’re on the right side of it. We break down the most powerful AI tools and workflows into stuff you can actually use, this week, at your job, without needing to understand a single line of code.

Your playbook for the AI era. Built for people who’d rather get results than sit through another jargon-filled tutorial.

Here’s what we’ll be sharing in coming days and weeks: practical AI workflows you can actually set up yourself, tool recommendations broken down by what you do (finance, marketing, consulting, legal, ops, etc.), how to think about your career as all of this shifts, and our honest take on the news when something big happens.

A year from now, there will be two kinds of professionals: those who became AI native, and those who wish they had.

Which one will you be?

Signing off for today,

Alisha & Abhilash