Google: "Our productivity has increased 10% thanks to AI." Random Internet User: "Skill Issue! We are 100x faster!" This is a tough one: I'm not sure who to believe here.

I told a client the code behind the project I was selling them was 100% AI-free. He loved it. I'll do it again.

I've been uploading screenshots of my investment allocation to ChatGPT and asking where to invest a specific amount. I've been doing this for about a year. 10/10 recommend.

Writing tests is a superpower, especially now with agentic coding. Every time I ask Claude Code to build something, I get it to run my test suite: 1. It writes the code 2. Runs my test suite 3. Identifies regressions and fixes them 4. Repeats 1/2

You should stop treating agents like humans. This has probably been one of the biggest unlocks for me over the past few months in terms of building agents. I see people giving agents personalities, goals, and even titles. But agents aren't people — they are functions with context windows.

I'm working with a huge company on their data preparation pipeline, and they are running a 9-step data processing pipeline. This all happens before their data is ready for training. There's a lot of code here and considerable complexity. 1/4

Python is the best language in the world. Yes, Python is slower than other languages, but I don't care because most of my work doesn't require it to be faster. Python is fast enough for what I need, and fast enough for most people out there.

Yesterday, I posted that large language models are non-deterministic, and people went bonkers. People called me stupid, asked me to go back to school, and used so many other names I can't remember. Today, I want to elaborate. 1/9

Building software that works is very easy… as long as you ignore all the hard parts.

If someone presents themselves as a "Programmer" and somebody else does it as a "Software Engineer," would you consider them equally capable? Do you see a fundamental difference between a Programmer and a Software Engineer?

The final 20% of the work always takes 80% of the total effort.

This guy freaked out when he tried a model much dumber than GPT 3.5. He ended up losing his job. He was an engineer and thought the model was sentient. Imagine what will happen when 99.99% of the world population uses one of these models for the first time.

Anyone can learn to code, especially now with AI. But it's still hard to find good solutions to difficult problems. Stop obsessing about languages, libraries, frameworks, and code. They won't make a difference. Your brain will.

There are *zero* successful software engineers who have only used one language in their careers. Zero. Most have used a dozen languages or more. Programming languages don't matter. They have no bearing on how valuable you are. And more importantly, people who pay don't care.

It's been 24 months since people told me my job was dead for real. (Coincidentally, these have been the best 24 months of my professional career.) Stop paying attention to the hype. Keep investing in yourself.

Do you really think “Project Managers” will survive AI? Really? So many people in my replies seem to think Projeft Managers will be immune to AI. They are even recommending engineers to pivot.

With AI improving at lightning speed, I think education is more important than it's ever been.

Every developer should consider pivoting to AI/ML. There are infinite opportunities to start enhancing what you're building with AI. Don't stay stuck for too long. Be aggressive with this. Things are moving fast.

Slow unit tests are better than too many mocks.

This is a 10x move here. One of my favorite ways to leverage models.

I did not visit StackOverflow in 2024.

Almost every person I've heard complaining about the job market has nothing to show to demonstrate their qualifications. They have no past projects they can share, no one to vouch for them, no publications, no social presence. Nothing. Words are cheap. Show me what you've done.

My favorite all-time computer ever has been the MacBook Pro M1 Pro. I have an M3 Max and an M1 Pro. Of course, the M3 is better on every metric, but there's something I can't describe about the M1 Pro that makes it special for me.

The biggest challenge right now is not to prompt an AI model to produce code. Almost anyone can do that. The challenge is to validate that this code is actually doing what you want it to do. The challenge is to understand how to improve it, fix it, and update it.

There are *literally* infinite opportunities for AI and software builders. We aren’t building more and better because there aren’t enough qualified people. These aren’t zero sum. We aren’t fighting for a fixed number of jobs. There’s no cap on how much we can build. We need more builders.

In 2025, try to place a "Buy" button on the Internet. This is something that could change your life. It certainly changed mine.

Midwits with a chatbot will never build better software than you. These are two of my all-time favorite Software Engineering books: 1. The Pragmatic Programmer 2. Clean Code Read them in 2025. They will pay off for years to come.

Start worrying the minute you see that companies building AGI stop hiring people. Right now, all of these companies have hundreds of open positions. That should tell you everything you need to know.

The best code is the one you didn't write. The second best code is the one that solves the problem. The best tool is the one you already have. The best solution is the simplest one. Always make it work first. Make it better later.

Multi-tasking is the death of productivity.

Every “Programming is dead” take I’ve seen so far is coming from people who have no clue about programming.

This MIT class is still the best way to learn Linear Algebra. It's free. Gilbert Strang is one of those generational professors. The type of person that will leave a positive mark on you. ocw.mit.edu/courses/18-...

Current AI can’t design, build, evaluate, deploy, monitor, and maintain software systems of arbitrary complexity. That’s what Software Engineers do. Great progress, though!

“This is not merely incremental improvement, but a genuine breakthrough” François Chollet about the just released OpenAI o3 model.

You can opt out of Visual Studio Code's Copilot from using your data to train their models. Works for free and pro users. This option is enabled by default. (You'll find this in your Copilot settings.)

We need unlimited software, and we need unlimited increase in quality (most software today sucks.) I believe AI will help us build *more* and *better* software. I believe AI will help *more* people become developers, not fewer. AI should add, never subtract.

AI-powered coding assistants at the end of 2024: • Useful to professional developers • Useful for normies who want to learn • Useful to generate quick prototypes • Useful to keep Dunning-Kruger patients happy • Useless in most complex domains • Useless to replace humans in any meaningful way

Boomer developers are not going to make it.

Copilot is now free! Probably the biggest news for developers in 2024.

Google has completely eclipsed OpenAI's 12 days of Christmas: • Quantum (Willow) • Video generation (Veo 2) • Multimodal Live API with Gemini 2.0 • Deep Research • Computer agent (Mariner) They are back in the race in a huge way! OpenAI should be worried.

You can integrate uv with your shell to enable autocompletion on the terminal! The good stuff keeps getting better! Take 5 minutes and look into uv. I'm willing to bet you'll like it a lot.

In 1994, people told me programming was for nerds and that I should become a doctor or a lawyer instead. 10 years later, they told me that someone from India would take my job for $5/hour. Then, no code was going to doom my career. 1/2

More people should ditch their bosses and build something they own: 1. Have control over your time 2. Have unlimited economic potential 3. Have the freedom to decide how to live your life Not for everyone, but you will never know unless you try. 99% of people are too scared to even try.

A software engineer’s job has always been solving problems. Writing code is just a tool we’ve used to accomplish part of it. Some people have got this wrong. The process of writing code will continue evolving and it might even change forever, but the job will remain the same.

A cool feature of uv: You can run a Python script without activating a virtual environment first. $ uv run -- python training .py The command above will do the following automatically: 1. Install Python if not present 2. Create virtual env if not present 3. Install requirements

Code in notebooks doesn't change the world. Models in production do.

Game-changing upgrade: Start using an automatic formatting tool that's opinionated. I haven't found one tool for Python that's better than black. It's fast, and I love the way it formats my code.

Code has forever been the main focus when building software. But ML and AI change the equation: Code takes a secondary role, and "data" becomes the lead actor. Individuals that understand how to produce, collect, manage and interpret data will own the future.

Devin is back at $500/month. This makes it out of reach for any casual users. Only people who are truly committed to making it work will pay for this. I feel this price is, in a way, "hiding" the tool's shortcomings. Wouldn't they make this dirty cheap if Devin was truly great?

TypeScript is annoying as hell. I don't write TypeScript for a living, so my experience here is limited, but over the past few weeks, I've been writing a Visual Studio Code extension, and I'm using TypeScript. 1/3