Oscar
@oskrgab.bsky.social
Empowering engineers through programming and analytics.
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
This has been an amazing experience and learned a lot along the way:
- MS Azure
- Authentication
- Javascript (Office.js)
- htmx
- Jinja templates
- more #FastAPI
October 9, 2025 at 5:00 PM
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
This has been an amazing experience and learned a lot along the way:
- MS Azure
- Authentication
- Javascript (Office.js)
- htmx
- Jinja templates
- more #FastAPI
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) 🚀
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) 🚀
September 1, 2025 at 8:23 PM
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) 🚀
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:
July 21, 2025 at 10:30 PM
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:
📊 What does this code do?
1. Prints all values
2. Prints non-zero values
3. Syntax error
4. Prints nothing
1. Prints all values
2. Prints non-zero values
3. Syntax error
4. Prints nothing
July 17, 2025 at 5:30 PM
📊 What does this code do?
1. Prints all values
2. Prints non-zero values
3. Syntax error
4. Prints nothing
1. Prints all values
2. Prints non-zero values
3. Syntax error
4. Prints nothing
I have been testing Kiro for a couple of days, and I have to say that it is great!
Kiro is a new AI IDE by AWS, and it is free to try with some good limits on usage.
Kiro is a new AI IDE by AWS, and it is free to try with some good limits on usage.
July 16, 2025 at 5:00 PM
I have been testing Kiro for a couple of days, and I have to say that it is great!
Kiro is a new AI IDE by AWS, and it is free to try with some good limits on usage.
Kiro is a new AI IDE by AWS, and it is free to try with some good limits on usage.
6. ROC Curves & AUC
Every decision involves trade-offs. ROC curves show what happens when you shift your standards.
Example: Comparing two models for loan approvals and balancing false positives against true positives.
Every decision involves trade-offs. ROC curves show what happens when you shift your standards.
Example: Comparing two models for loan approvals and balancing false positives against true positives.
July 15, 2025 at 10:30 PM
6. ROC Curves & AUC
Every decision involves trade-offs. ROC curves show what happens when you shift your standards.
Example: Comparing two models for loan approvals and balancing false positives against true positives.
Every decision involves trade-offs. ROC curves show what happens when you shift your standards.
Example: Comparing two models for loan approvals and balancing false positives against true positives.
5. Confusion Matrix, Precision & Recall
“Accuracy” alone is misleading. Precision and recall show where your model truly stands.
Example: Evaluating medical tests where a missed diagnosis matters more than accuracy alone.
“Accuracy” alone is misleading. Precision and recall show where your model truly stands.
Example: Evaluating medical tests where a missed diagnosis matters more than accuracy alone.
July 15, 2025 at 10:30 PM
5. Confusion Matrix, Precision & Recall
“Accuracy” alone is misleading. Precision and recall show where your model truly stands.
Example: Evaluating medical tests where a missed diagnosis matters more than accuracy alone.
“Accuracy” alone is misleading. Precision and recall show where your model truly stands.
Example: Evaluating medical tests where a missed diagnosis matters more than accuracy alone.
4. Linear & Logistic Regression
Basic doesn’t mean boring. These are the bread-and-butter tools that let you predict what’s next. Also, they are my go-to tools for descriptive analytics.
Example: Predicting house prices or who’s likely to unsubscribe.
Basic doesn’t mean boring. These are the bread-and-butter tools that let you predict what’s next. Also, they are my go-to tools for descriptive analytics.
Example: Predicting house prices or who’s likely to unsubscribe.
July 15, 2025 at 10:30 PM
4. Linear & Logistic Regression
Basic doesn’t mean boring. These are the bread-and-butter tools that let you predict what’s next. Also, they are my go-to tools for descriptive analytics.
Example: Predicting house prices or who’s likely to unsubscribe.
Basic doesn’t mean boring. These are the bread-and-butter tools that let you predict what’s next. Also, they are my go-to tools for descriptive analytics.
Example: Predicting house prices or who’s likely to unsubscribe.
3. Confidence Intervals
One number won’t tell the full story. Confidence intervals give you a range to trust—not guess.
Example: Estimating how many users will actually convert, within a realistic range.
One number won’t tell the full story. Confidence intervals give you a range to trust—not guess.
Example: Estimating how many users will actually convert, within a realistic range.
July 15, 2025 at 10:30 PM
3. Confidence Intervals
One number won’t tell the full story. Confidence intervals give you a range to trust—not guess.
Example: Estimating how many users will actually convert, within a realistic range.
One number won’t tell the full story. Confidence intervals give you a range to trust—not guess.
Example: Estimating how many users will actually convert, within a realistic range.
2. Hypothesis Testing & p-values
Is your result legit or just luck? Hypothesis tests give clarity. Understand p-values or risk chasing ghosts.
Example: Checking if that new website tweak actually boosts sales.
Is your result legit or just luck? Hypothesis tests give clarity. Understand p-values or risk chasing ghosts.
Example: Checking if that new website tweak actually boosts sales.
July 15, 2025 at 10:30 PM
2. Hypothesis Testing & p-values
Is your result legit or just luck? Hypothesis tests give clarity. Understand p-values or risk chasing ghosts.
Example: Checking if that new website tweak actually boosts sales.
Is your result legit or just luck? Hypothesis tests give clarity. Understand p-values or risk chasing ghosts.
Example: Checking if that new website tweak actually boosts sales.
1. Bayesian Inference
Stop ignoring your gut feeling. Bayesian methods combine your prior beliefs with new data. Essential when situations keep changing.
Example: Updating your fraud model as new scams pop up.
Stop ignoring your gut feeling. Bayesian methods combine your prior beliefs with new data. Essential when situations keep changing.
Example: Updating your fraud model as new scams pop up.
July 15, 2025 at 10:30 PM
1. Bayesian Inference
Stop ignoring your gut feeling. Bayesian methods combine your prior beliefs with new data. Essential when situations keep changing.
Example: Updating your fraud model as new scams pop up.
Stop ignoring your gut feeling. Bayesian methods combine your prior beliefs with new data. Essential when situations keep changing.
Example: Updating your fraud model as new scams pop up.
Most people think statistics is about complicated math. It’s not.
It’s about making smarter decisions with data—and if you’re not comfortable with these 6 tools, you’re flying blind.
It’s about making smarter decisions with data—and if you’re not comfortable with these 6 tools, you’re flying blind.
July 15, 2025 at 10:30 PM
Most people think statistics is about complicated math. It’s not.
It’s about making smarter decisions with data—and if you’re not comfortable with these 6 tools, you’re flying blind.
It’s about making smarter decisions with data—and if you’re not comfortable with these 6 tools, you’re flying blind.
Share your Cursor usage plot!
I'm curious to see how efficiently others use tools like Cursor. There's an interesting plot in your Cursor account that displays the number of lines proposed by Cursor and how many you accepted.
I'm curious to see how efficiently others use tools like Cursor. There's an interesting plot in your Cursor account that displays the number of lines proposed by Cursor and how many you accepted.
July 8, 2025 at 5:02 PM
Share your Cursor usage plot!
I'm curious to see how efficiently others use tools like Cursor. There's an interesting plot in your Cursor account that displays the number of lines proposed by Cursor and how many you accepted.
I'm curious to see how efficiently others use tools like Cursor. There's an interesting plot in your Cursor account that displays the number of lines proposed by Cursor and how many you accepted.
Here are 5 funny things I learned building my first website...
and what it taught me about the internet.
I just wanted my own little corner of the web. No ChatGPT back then—just deep Google dives, blog posts, and a few books. Here is what surprised me:
and what it taught me about the internet.
I just wanted my own little corner of the web. No ChatGPT back then—just deep Google dives, blog posts, and a few books. Here is what surprised me:
July 3, 2025 at 10:34 PM
Here are 5 funny things I learned building my first website...
and what it taught me about the internet.
I just wanted my own little corner of the web. No ChatGPT back then—just deep Google dives, blog posts, and a few books. Here is what surprised me:
and what it taught me about the internet.
I just wanted my own little corner of the web. No ChatGPT back then—just deep Google dives, blog posts, and a few books. Here is what surprised me:
Python in Excel or the browser—fully offline. No servers. No installs. Just plug-n-code
Sounds too good to be true?
That’s exactly what tools like xlwings Lite, Anaconda Code, JupyterLite, and PySheets make possible today.
Sounds too good to be true?
That’s exactly what tools like xlwings Lite, Anaconda Code, JupyterLite, and PySheets make possible today.
July 2, 2025 at 8:00 PM
Python in Excel or the browser—fully offline. No servers. No installs. Just plug-n-code
Sounds too good to be true?
That’s exactly what tools like xlwings Lite, Anaconda Code, JupyterLite, and PySheets make possible today.
Sounds too good to be true?
That’s exactly what tools like xlwings Lite, Anaconda Code, JupyterLite, and PySheets make possible today.
Which AI tool is actually worth your time and money right now?
I’ve been testing different AI tools over the past few months—
some great, some not worth the cost.
If you’re trying to figure out the best combination of tools
for coding, writing, or just thinking clearly,
I’ve been testing different AI tools over the past few months—
some great, some not worth the cost.
If you’re trying to figure out the best combination of tools
for coding, writing, or just thinking clearly,
June 30, 2025 at 5:00 PM
Which AI tool is actually worth your time and money right now?
I’ve been testing different AI tools over the past few months—
some great, some not worth the cost.
If you’re trying to figure out the best combination of tools
for coding, writing, or just thinking clearly,
I’ve been testing different AI tools over the past few months—
some great, some not worth the cost.
If you’re trying to figure out the best combination of tools
for coding, writing, or just thinking clearly,
Random Forest is a popular machine learning algorithm that combines the output of multiple decision trees, each trained on a random subset of data and features. Then, it aggregates the predictions of these trees to get the final result.
Some of the advantages are:
Some of the advantages are:
June 26, 2025 at 5:00 PM
Random Forest is a popular machine learning algorithm that combines the output of multiple decision trees, each trained on a random subset of data and features. Then, it aggregates the predictions of these trees to get the final result.
Some of the advantages are:
Some of the advantages are:
#LearningRust day 3
----------------------------------------------
Some basic concepts from the book:
“Rust is an ahead-of-time compiled language, meaning you can compile a program and give the executable to someone else, and they can run it even without having Rust installed.”
----------------------------------------------
Some basic concepts from the book:
“Rust is an ahead-of-time compiled language, meaning you can compile a program and give the executable to someone else, and they can run it even without having Rust installed.”
June 25, 2025 at 5:50 PM
#LearningRust day 3
----------------------------------------------
Some basic concepts from the book:
“Rust is an ahead-of-time compiled language, meaning you can compile a program and give the executable to someone else, and they can run it even without having Rust installed.”
----------------------------------------------
Some basic concepts from the book:
“Rust is an ahead-of-time compiled language, meaning you can compile a program and give the executable to someone else, and they can run it even without having Rust installed.”
Each release (from 3.10 through 3.14) continues to improve them: smarter typo hints, missing-argument names, and colorized tracebacks — debugging is much more approachable and gentle for beginners nowadays. So yeah, the best time to start programming, every day, is today!
June 23, 2025 at 5:00 PM
Each release (from 3.10 through 3.14) continues to improve them: smarter typo hints, missing-argument names, and colorized tracebacks — debugging is much more approachable and gentle for beginners nowadays. So yeah, the best time to start programming, every day, is today!
If you’re learning Python, clearer error messages make all the difference — they’re more than just syntax alerts.
June 23, 2025 at 5:00 PM
If you’re learning Python, clearer error messages make all the difference — they’re more than just syntax alerts.
June 19, 2025 at 5:30 PM
You’ve seen colleagues using Python.
To analyze data.
To apply machine learning.
To make smarter decisions.
You’re still working with Excel.
And doing what you can.
But deep down…
you know you could do more.
More insights.
Better forecasts.
Smarter models.
To analyze data.
To apply machine learning.
To make smarter decisions.
You’re still working with Excel.
And doing what you can.
But deep down…
you know you could do more.
More insights.
Better forecasts.
Smarter models.
June 17, 2025 at 5:00 PM
You’ve seen colleagues using Python.
To analyze data.
To apply machine learning.
To make smarter decisions.
You’re still working with Excel.
And doing what you can.
But deep down…
you know you could do more.
More insights.
Better forecasts.
Smarter models.
To analyze data.
To apply machine learning.
To make smarter decisions.
You’re still working with Excel.
And doing what you can.
But deep down…
you know you could do more.
More insights.
Better forecasts.
Smarter models.
When I started with Python, type hints were just this “new thing” in the docs.
No one used them and since Python is a dynamically typed language, why they even existed? Just write your code and it runs.
No one used them and since Python is a dynamically typed language, why they even existed? Just write your code and it runs.
June 11, 2025 at 5:30 PM
When I started with Python, type hints were just this “new thing” in the docs.
No one used them and since Python is a dynamically typed language, why they even existed? Just write your code and it runs.
No one used them and since Python is a dynamically typed language, why they even existed? Just write your code and it runs.
#LearningRust day 2
Installing #Rust
The recommended way is to install rustup from the command line or terminal. I'm using MacOS so I was tempted to see if an installation through #brew was available.
Installing #Rust
The recommended way is to install rustup from the command line or terminal. I'm using MacOS so I was tempted to see if an installation through #brew was available.
June 9, 2025 at 5:02 PM
#LearningRust day 2
Installing #Rust
The recommended way is to install rustup from the command line or terminal. I'm using MacOS so I was tempted to see if an installation through #brew was available.
Installing #Rust
The recommended way is to install rustup from the command line or terminal. I'm using MacOS so I was tempted to see if an installation through #brew was available.
📊 formulaML Examples – Kickstart Your Projects with Regression, Classification & Clustering.
Don’t forget to join the beta testing waitlist! ➡️tally.so/r/mRlaWJ
Don’t forget to join the beta testing waitlist! ➡️tally.so/r/mRlaWJ
June 5, 2025 at 5:30 PM
📊 formulaML Examples – Kickstart Your Projects with Regression, Classification & Clustering.
Don’t forget to join the beta testing waitlist! ➡️tally.so/r/mRlaWJ
Don’t forget to join the beta testing waitlist! ➡️tally.so/r/mRlaWJ