🚀 Machine Learning in Excel? formulaML is my attempt to bring Machine Learning where people feel most confortable with and where their data already resides - Spreadsheets and using what they already know = formulas. 📊 No Python. No IDE. Just Excel formulas.

Building this has been addictive – can't wait to ship outlier detection and time series analysis next! Let me know if you have any questions regarding the process of publishing the add-in!

Here's what drives formulaML: I believe every business analyst should have ML at their fingertips, not hidden behind Python scripts. So I made it work with Excel formulas they already use daily.

Just having the add-in wasn't enough though, then I had to navigate the process to become a MS partner, going through the validation process, and finally reviewing the add-in with Microsoft testing support.

And of course #xlwings server by Felix Zumnstein! The documentation is amazing and you can get up and running in minutes.

I am proud to announce that my first Excel add-in is live on Microsoft's AppSource! This has been an amazing experience and learned a lot along the way: - MS Azure - Authentication - Javascript (Office.js) - htmx - Jinja templates - more #FastAPI

If you don’t know what git, commit, branching or even what the terminal is, tell the LLM to explain it to you, it’s going to be the best token usage in the whole year.

In a nutshell: 🔄 Request small changes 📥 git commit 🔀 git checkout 🔗 git merge 🔁 Repeat for every new feature

- Ask the LLM for specific/incremental changes. - Under the new branch name, call 'git commit' and save your changes - If something goes wrong, revert to a previous version of the app using 'git checkout'. - If everything is ok, you can merge your "new-feature-branch" to main

Quick steps to leverage Git: - Download git. - Use 'git init' in your project folder to start the repo - Save the first working app version in your "main" branch using the 'git commit' - Create a new branch from main 'git checkout -b <branch-name>'

And because you are Vibe Coder, chances are that you don't understand what the actual code is doing. 😅 Do yourself a favor and USE GIT! ✨

You ctrl + z, restore previous checkpoints, but still, it doesn't work. The LLM touched many files that these restore points are not fully realiable. Besides, you need to navigate through the endless conversations to see where the breaking changes were introduced.

The initial prompts look great, so you continue. Your LLM partner handles the backend, frontend, and runs tests, all passing. You successfully load your app on localhost. Life's good, you're in the flow, and adding features. But suddenly, the original app features stop working.

How to mitigate the risk of screwing up your app codebase (Vibe coder edition). 🛠 TL;DR: Use git. 💻 Picture the following: You are starting an app concept using your favorite LLM Code assistant (GH Copilot, Cursor, Claude Code, Codex, etc) 🚀

I've been using Kiro for over a week now (I was fortunate to get access before the waitlist). Having used Cursor for over a year, here are some differences between the two IDEs:

Since Python 3.8, there's a handy operator that solves a common pattern: 📌 Assigning a value inside an expression without repeating yourself. It’s called the walrus operator (:=), and although it’s powerful, many developers still don’t fully use it.

🔍 Check out this code: data = [10, 20, 0, 30, 0, 40] for x in data: if (val := x): print(f"Processing {val}") 📊 What does this code do?

Make sure to try Kiro now, it is free, and it's using top tier models with very good limits on usage, so take advantage of it while in Preview!

I love Kiro mainly for its Specs feature. Unlike Cursor, where I kept switching between tasks and using ChatGPT to understand the bigger picture, Kiro simplifies planning and task implementation.

Now, it is the task list that will be actually executed. Above each task list, you'll see a clickable "Start Task" text. This will launch Claude to actually start writing the code.

This applies to the design and task list. You can refine them anytime or revisit the requirements if needed. Claude reviews your codebase to align requirements, design, and tasks.