Coding with AI
A beginner's tutorial on using AI to code away work (and have fun).
Hey, Talent Code fam! One quick favor before we get going—
The GHC vote is open! If you haven’t already, please take two minutes to vote for our session proposal for the Grace Hopper Celebration this fall.
It genuinely makes a difference, and I am deeply grateful for every click.
Now, let's get into a topic that everyone’s talking about: coding with AI.
Current AI tools have fundamentally changed how we work and how we code, and has dramatically changed who is able to do it. Personally, I’m hugely grateful for it—I’ve been a policy wonk for so long with very little time for hands-on-keyboard time, and this has let me get back in the game, building tools instead of just talking about them.
So now, if that’s something you’re interested in, I’m going to take you through how to get started coding and building tools with AI.
Let’s decode it. 🚀
How to Use AI as Your Coding Tutor
And build your own solutions to your workflow problems.
Let me tell you how I’d teach you to code today if I had an hour with you.
I wouldn’t start with syntax. I wouldn’t start with a textbook. I definitely wouldn’t start with a tool. I don’t run my data and AI literacy classes that way and I wouldn’t run a coding class with you that way today.
I’d start with a problem you already have—something you do every week that takes longer than it should. A report you build by hand. A spreadsheet you update the same way every time. A document you recreate from scratch when a template would do.
And then I’d sit next to you while you described it to an AI, watched it write the code, ran it yourself, and did your critical evaluation and troubleshooting.
That’s it. That’s the class.
This is what’s changed. You no longer need to learn to code before you can build things with code. AI can write the code.
Your job is to know what problem you’re trying to solve, be able to describe it clearly, and be willing to experiment.
If you’ve been reading this newsletter, you already know how to do all three of those things.
What “Coding With AI” Actually Looks Like
Here’s what the process looks like in practice.
You open a conversation with an AI. You describe your problem like you’d describe it to a smart colleague who just happens to know every programming language ever invented. You tell them what you’re starting with, what you want the end result to look like, and any constraints that matter.
The AI writes code. You run it, see what happens, and either use it or tell the AI what’s off.
That’s vibe coding. It sounds like coding when you’re in the zone (anyone remember Hugh Jackman enjoying Chateau Margeaux and banging on his keyboard and six screens to develop a master virus in Swordfish?) but it’s really just the practice of building functional tools through natural-language conversation with AI, without needing to fully understand the underlying code.
You’re describing the outcome, and AI is doing the technical execution.
The key mindset shift: you don’t need to understand every line. You need to understand the problem and the logic, and evaluate the solution!
Three Things You Can Build This Week
Let’s build some stuff.
Here are three real projects—each one practical, each one achievable in a really short amount of time, and each one using nothing more than an AI chat window and tools you already have.
Build 1: The Smart Budget Macro
What it is: A spreadsheet macro that turns a raw dump of bills, expenses, and forecasted costs into a living budget—one that updates itself, flags risk before you’re in trouble, and tracks where your money actually goes over time.
Most people manage budgets reactively. They find out they overspent when the month is over. This macro flips that—it projects available resources, tracks 3-month rolling averages by category, and turns red when you’re within 10% of a limit. You paste in your expenses; it does everything else.
What you tell the AI:
“I have a spreadsheet where I paste in my monthly expenses and bills. Each row has: date, category (like rent, groceries, utilities, subscriptions), amount, and whether it’s actual or forecasted. Build me a macro that: (1) automatically totals actual vs. forecasted spend by category, (2) calculates a 3-month rolling average per category, (3) computes remaining available budget based on my income, which I’ll put in a fixed cell, and (4) highlights any category in red where I’m within 10% of my limit or over it. I’m using [Excel / Google Sheets].”
You’ll walk away with:
Auto-updating totals that refresh every time you paste new data
A rolling 3-month average per spending category — so you can see trends, not just snapshots
A risk flagging system that warns you before you’re in trouble
A clear picture of what’s left — before you commit to spending it
Tip: Build a dedicated “Data” tab for your raw expense dump and a separate “Summary” tab for totals and flags. Ask AI to reference both.
Build 2: The Delinquent Task Email Generator
What it is: A script that reads your task list, identifies everything overdue, groups it by owner, and generates a pre-populated follow-up email—formatted, professional, and ready to send.
If you manage a team, you know the Friday afternoon task of manually combing through a project list to see what fell through the cracks. This script does that in seconds—and writes the email for you.
What you tell the AI:
“I have a task list in [Excel / Google Sheets / a CSV export from my project tool] with columns for: task name, assigned owner, due date, and status. Write a Python script that: (1) reads the file, (2) identifies all rows where the due date has passed and status is not ‘Complete’ or ‘Done’, (3) groups those tasks by owner, (4) sorts them by how many days overdue they are — most overdue first, and (5) generates a formatted email summary that lists each owner, their overdue items, due dates, and days overdue. Output the email text to a file called ‘followup_email.txt’ so I can copy it into Outlook or Gmail.”
You’ll walk away with:
A script that scans your task list and flags issues in seconds
A grouped, owner-by-owner breakdown with urgency built in
A pre-written follow-up email ready to send with no need for formatting or manual list-building
An optional add-on: ask AI to generate one email per owner so you can send individual nudges
Tip: Run the script on a small test export first (10 or 15 rows) before pointing it at your full list. Verify the output against what you see in the file. This is your CLEAR step.
Build 3: The Budget Dashboard App
What it is: A browser-based visual dashboard—built with a free Python tool called Streamlit, which I love—that reads your budget spreadsheet and displays it as interactive charts: category breakdowns, spending trends, and a forecast vs. actual comparison. Upload a new file and the dashboard refreshes automatically.
What you tell the AI:
“I have a budget spreadsheet with columns for: date, category, amount, and type (actual or forecasted). Build me a Streamlit app in Python that reads this file and displays: (1) a bar chart of actual vs. forecasted spend by category, (2) a line chart showing total spending by month over time, (3) a pie chart of spend distribution by category for the current month, and (4) a summary table showing total actual, total forecasted, and the difference for each category. I want to be able to upload a new version of the spreadsheet and have the dashboard refresh automatically. Include a requirements.txt file and a one-line command to launch the app.”
You’ll walk away with:
A visual budget dashboard that runs in your browser from your own computer
Charts you can screenshot and drop into any briefing or meeting
A forecast vs. actual gap analysis that catches budget drift early
A reusable framework you can adapt for any data — not just budgets
Tip: Streamlit is free and runs entirely on your local machine. Nothing goes to a server. Ask AI to include setup instructions so you’re not Googling how to install it later.
The Most Important Thing About Getting Started
Don’t get tripped up by the expectation that things will work perfectly the first time.
Life happens. The script won’t run, or it’ll produce output that’s close but not quite right, or you’ll realize halfway through that you described the problem wrong. That’s process, not failure. It’s like Battleship, in that sometimes you figure out where the ships…aren’t. Every error message is information. Every wrong output tells you something about how you described the problem.
Paste the error back into the AI and say: “I got this error. What went wrong and how do I fix it?” Nine times out of ten, it walks you right through it.
You’re not learning to be a developer. You’re learning to be a precise thinker who can collaborate with a very capable tool. You already are that person.
Five Simple Apps You Can Build This Week
Want to go further? Here are five more tools you can build in an afternoon (or less)—each one useful, each one buildable with nothing but a prompt and Python. All five use Streamlit, which is free and runs locally. Just describe the app, run what AI writes, and you’ve built something real!
App 1: Meeting Cost Calculator
What it does: Enter the number of attendees, their average salary band, and the meeting length—it tells you the real dollar cost of that meeting.
Prompt to use:
“Build a Streamlit app that calculates the cost of a meeting. Inputs: number of attendees, average hourly rate (or annual salary to convert), and meeting duration in minutes. Output: total meeting cost in dollars, cost per attendee, and a simple verdict tied to a threshold I can set. Keep it clean and easy to use.”
App 2: Habit Tracker with Streak Counter
What it does: Check off daily habits, see your current streak for each, and get a weekly summary—all stored in a local file so your data persists between sessions.
Prompt to use:
“Build a Streamlit habit tracker. I want to: add habits to track, check them off daily, see my current streak per habit, and view a weekly summary showing which habits I hit each day. Store data in a local JSON file so it saves between sessions.”
App 3: Team Sentiment Pulse Checker
What it does: A simple weekly pulse survey for your team. They rate their week 1–5 and add an optional note. You see a trend dashboard over time.
Prompt to use:
“Build a two-page Streamlit app. Page 1: a form where team members enter their name (or stay anonymous), rate their week 1–5, and add an optional comment. Page 2: a manager dashboard showing average weekly rating as a line chart over time, individual ratings as a table, and any comments submitted. Store responses in a CSV.”
App 4: Job Application Tracker
What it does: A clean dashboard to track every application — company, role, date, status, next step, notes. Sortable, filterable, and far better than a scattered spreadsheet.
Prompt to use:
“Build a Streamlit job application tracker. I want to: add new applications (company, role, date applied, status, next action, notes), view all in a sortable table, filter by status (Applied, Interview, Offer, Rejected, Withdrawn), and edit or delete entries. Store data in a local CSV.”
App 5: Decision Matrix Tool
What it does: Enter your options and the criteria that matter, weight each criterion, and the app scores every option and tells you which one wins—with a chart to back it up.
Prompt to use:
“Build a Streamlit decision matrix app. Inputs: a list of options, a list of criteria, a weight for each criterion (1–5), and a score for each option on each criterion (1–10). Output: a weighted score for each option, a ranked results table, a bar chart of scores, and a CSV download of the matrix.”
I may or may not be using the last one to weigh post-Army options.
So now I can code?
More or less. You can write a sentence with the help of AI. Where training and classes come in handy is helping you evaluate whether or not it’s a good sentence.
Take the AI slop you see everywhere. LinkedIn. Newsletters. Memos. The “it’s not this, it’s this,” hyperbolic claims of game changing, “in an ever-changing world…”
If you don’t understand the mechanics of coding, your code will be sloppy.
In a lot of cases, that won’t matter. You’re making something simple for your own use that operates locally, so scaling, security, and maximizing compute don’t have to be things you consider. Those are things, though, that you need a software engineer for. And designing UI/UX is still an art where experience and expertise matter.
In the realm of AI, you’re not paying the photographer for your photo when you can make your own edits through AI, but you are paying them for their eye. AI can’t put together the original and creative ideas that artists can, even if they can help you get around technical gaps—and that’s true in software, too. There is an understanding of flow and usability that tech can help you with, but there’s really not much substitute for expertise and experience.
So yeah, you can code. You can create apps and plug-ins and a whole lot of other things. You can do what I’m doing and try to vibe code a new framework for your website (mixed success). But I wouldn’t be calling yourself a software engineer yet.
But that’s not your goal here. Your goal is to learn how to use AI to code and start creating easy apps to automate the things we’ve been talking about automating in your workflow, to create prototypes that you can test out and see how much efficiency you gain, and to just get more confident with using AI beyond generating funny pics.
Your assignment this week is to pick one project from this list, preferably one that solves a problem you actually have right now. Describe it to an AI tool, run what comes out, check it against CLEAR, and adjust.
Something to sit with this week:
What’s one task you do on a regular basis that follows the exact same pattern every time — and what would it mean for your team if that task took 90 seconds instead of 30 minutes?
I’d love to hear about what you build and what worked for you, so please do share!
