NewFeatured
uv Quietly Won the Python Packaging War — Here's How to Switch in 10 Minutes
pip, venv, pyenv, pipx and poetry just got replaced by one Rust binary. Here's why uv won, and how to switch today without breaking anything.
Ali Jabbary
Latest writing
Latest Insights& Expert Guides
Working notes from 10+ years of teaching Python, machine learning, and data science. Real code, honest tradeoffs, no AI slop.
//browse_by
Pythonuv Quietly Won the Python Packaging War — Here's How to Switch in 10 Minutes
pip, venv, pyenv, pipx and poetry just got replaced by one Rust binary. Here's why uv won, and how to switch today without breaking anything.
Machine LearningBuild Your First Real AI Agent in 2026 (Without Drowning in Frameworks)
There are 120+ agent tools out there and most tutorials make agents sound like magic. Here's the boring, reliable way that actually ships.
LearningHow to Actually Land a Data or ML Role in 2026 (No Sugar-Coating)
Entry-level data roles got markedly scarcer and AI ate the easy tasks. Here's the honest map for breaking in when the front door got smaller.
Machine LearningRun a Genuinely Useful LLM on Your Own Laptop in 2026
Private, free, offline, and finally good enough. Here's exactly which small model to run, how, and where local hits its limits.
Data SciencePolars vs pandas in 2026: When the Speed Gap Actually Matters
Polars can be many times faster than pandas on large data, and the internet won't shut up about it. Here's when to actually switch — and when to stay put.
LearningHow to Learn With AI Without Quietly Rotting Your Brain
AI tutors make you feel 48% smarter and leave you 17% dumber on the real exam. Here's how to flip that — and actually keep what you learn.
Machine LearningContext Engineering: The Skill That Quietly Replaced Prompt Engineering
Your prompt isn't the bottleneck anymore. What the model can see is. Here's how to engineer context like a pro in 2026.
PythonThe GIL Is Finally Optional: A Hands-On Tour of Free-Threaded Python 3.14
Python 3.14 makes the GIL officially optional. Here's how to actually run no-GIL Python today, what gets 4x faster, and the traps nobody warns you about.
Data ScienceDuckDB: How I Replaced a Cloud Data Warehouse With One Tiny File
DuckDB runs SQL on gigabytes of Parquet from your laptop — no cluster, no bill. Here's why every data team quietly adopted it.
PythonRuff and ty: How One Company Quietly Rebuilt Python's Entire Dev Toolchain in Rust
Astral replaced flake8, black, isort and now mypy with Rust tools that are 10–100x faster. Here's the modern lint + format + type-check setup.
LearningEscaping Tutorial Hell: The 80/20 Rule That Finally Got Me Out
You don't have a knowledge problem — you have a reps problem. Here's the simple ratio that turns passive watching into a portfolio that gets you hired.
Data ScienceYour Jupyter Notebook Is Lying to You. marimo Fixes That.
96% of Jupyter notebooks don't reproduce. marimo's reactive, pure-Python notebooks kill hidden state for good — here's how.
Machine LearningHow to Actually Learn Machine Learning in 2026 (An Honest Roadmap)
A working engineer's honest roadmap for learning machine learning in 2026 — no PhD required, no "master all the math first" gatekeeping. The right order, the myths to ignore, and how to know you're actually progressing.
Data ScienceThe SQL You Actually Need for Data Science (and the 20% That's 80% of the Job)
SQL is the most underrated skill in data science, and you don't need all of it. Here's the honest short list of what matters day-to-day and in interviews, with runnable snippets and a clear marker for where "studied SQL" becomes "uses SQL."
Machine Learning1-on-1 ML Tutor vs an Online Course (Coursera, Udemy, DataCamp): An Honest Comparison
I tutor data science for a living, so I have an obvious bias. Here's the genuinely fair version: when a $30 course beats a tutor, when a tutor beats the course, and the hybrid path that usually wins on value.
CareerHow to Prepare for a Data Science Interview in 2026 (An Honest Guide)
Most data science interview prep gets the priorities backwards. Here's what actually gets you hired in 2026 — SQL, clear thinking, and reading messy data — versus the obscure algorithms everyone over-studies and nobody asks about.
PythonThe 7 pandas Errors That Trip Up Every Beginner (and How to Actually Fix Them)
SettingWithCopyWarning, KeyError, weird dtypes, exploding merges — almost every pandas beginner hits the same handful of errors. Here's what each one actually means, why it happens, and the real fix (not just a copy-paste that makes the warning go away).
Machine LearningThe scikit-learn pipeline mistake that quietly leaks your test set
You did everything right — train/test split, cross-validation, the works. Your model gets 94 percent accuracy in dev and 71 percent in production. The bug is almost always the same subtle scaling mistake. Here is how to spot it and the three-line fix.
CareerShould you hire a Python tutor in 2026? An honest answer from one
Free YouTube exists. So does Stack Overflow. So does Claude. Here is the honest case for and against hiring a 1-on-1 tutor in 2026 — written by someone who is one and gets paid to do it. With the four kinds of learner who actually benefit, and the three who do not.
Data ScienceP-values, demystified: what "statistically significant" actually means
A p-value is not the probability your hypothesis is true. It is not the probability your result is real. After 10 years of seeing students get this wrong in interviews, lab reports, and A/B tests, here is the explanation I wish someone had given me.
PythonNumPy vectorization vs Python loops: when "always vectorize" is wrong
Every performance tutorial says vectorize everything. Then they show a one-liner that beats a for-loop 100x. Real production code is messier — here is when vectorization actually helps, when it hurts, and the benchmarks to prove both.
PythonPandas groupby vs pivot_table: a working decision guide for 2026
Both APIs return the same answer in 90% of cases — so why does anyone teach both? A practical decision guide with benchmarks, three real failure modes, and the rule I use in 1-on-1 sessions.
GeneralAGI 2025 Update: Peeking Over The Horizon
A comprehensive look at the latest developments in artificial general intelligence and what to expect in 2025.
GeneralStudent AI Stack 2025: Essential Tools
The complete guide to AI tools every student needs to excel in their studies and future career.
Stuck on a specific problem?
These posts cover the patterns I see most often in 1-on-1 sessions. If you want help applying them to your code, the first session is free — no credit card required.
Book a free 30-minute session