Real Python (@realpython) header
Real Python (@realpython) profile picture

Helping Python developers build with AI: apps, agents, models, and AI-assisted coding. Expert-reviewed tutorials and courses, read by 1M+ developers a month.

Joined August 2012
Photo by Real Python (@realpython)Photo by Real Python (@realpython)Photo by Real Python (@realpython)Photo by Real Python (@realpython)Photo by Real Python (@realpython)Photo by Real Python (@realpython)Photo by Real Python (@realpython)Photo by Real Python (@realpython)Photo by Real Python (@realpython)Photo by Real Python (@realpython)
Python puzzle: return the length of the shortest run of neighboring numbers that adds up to a target. [4, -2, 3, 1] with target 2 → 2, from [4, -2]. Negative numbers make it trickier than it looks. Can you solve it? realpython.com/courses/codin…
2
2
514
Meta's Muse Spark on "minimal" reasoning effort: latency down by more than half, cost down about 60% in one test. Same accuracy. The open-weights Muse Glimmer answered that test question for about one-eighth of Spark's cost. realpython.com/meta-muse-pyt…
1
2
704
Want to make games in Python but don't know where to start? Rock, Paper, Scissors in the terminal → turtle graphics → Tkinter board games → Pygame side-scrollers → Arcade platformers. One path, in that order 🎮 realpython.com/learning-path…
4
835
Your first Django contribution probably shouldn't be a ticket that's 15 years old. Tim Schilling says Djangonaut Space newcomers can pick exactly that, so Django now sets aside tickets for first-timers. It's changing Django sprints, too. realpython.com/podcasts/rpp/…
757
Got a project with a requirements.txt file? One command installs everything it lists: pip install -r requirements.txt The tutorial shows the rest of pip, from checking you have it to uninstalling. realpython.com/what-is-pip/?…
2
829
5. Once Python fundamentals feel comfortable, give about 20 to 30% of your weekly practice to coding challenges like LeetCode. Start with easy problems, then work up to medium. Learn the patterns behind solutions, like when recursion makes sense, instead of memorizing answers.
1
345
4. To get interviews for most entry-level jobs, you need solid Python fundamentals plus one area, such as web, data or automation, where you can troubleshoot problems on your own. Rate your skills from 1 to 5, then put most of your hours into the required ones you rated 1 or 2.
1
18
3. Make your first portfolio project small enough to finish. A tiny command-line tool that does one thing well teaches you more than an ambitious app you abandon halfway. It also gives you something concrete to talk about in interviews.
1
20
2. Open 5 to 10 current listings for the job you want. Highlight every skill that shows up in at least three. Frameworks vary by role. Nearly every listing shares the same core: Python fundamentals, Git, unit tests and systematic debugging.
1
33
1. Some companies, Meta among them, now let you use an AI assistant in coding interviews. The fastest way to fail is to prompt your way to a solution without writing code or reviewing what the AI gave you. Some interviewers now show you flawed AI code and ask what's wrong.
1
39
You know Python syntax but not what to learn next to get hired. Five things that take the guessing out of it: 1. How to fail an AI-enabled interview 2. The job-listing test 3. What your first project needs 4. How good is good enough 5. How much time LeetCode gets 🧵
3
1
3
1,979
A chatbot trained on your own WhatsApp chats makes a fun portfolio project. You clean the data with regular expressions, train ChatterBot on it, then add a local LLM through Ollama. realpython.com/build-a-chatb…
2
5
1,223
Claude Code hooks on one page: settings layout, lifecycle events, exit codes and copy-paste hook scripts. Free PDF cheat sheet 👇 realpython.com/bonus/claude-…
1
4
1,289
5. You can design a quantum circuit on your laptop with Qiskit. qc.h(0) puts a qubit into an equal superposition of 0 and 1, and qc.measure(0, 0) reads it into a classical bit. Run the circuit 10,000 times on a simulator and roughly half the results are 0, half 1.
1
351
4. Quantum advantage on broadly useful problems isn't here yet. Most of today's noisy intermediate-scale quantum (NISQ) machines have under 1,000 qubits. Qubits can lose their quantum state in microseconds, and control pulses, crosstalk and calibration drift add errors.
1
23
3. Grover's algorithm searches an unsorted list of N items in about √N steps. A classical search needs about N. On paper, a million items takes on the order of a thousand Grover steps instead of a million. That assumes a quantum computer big and reliable enough to run it.
1
18