Python

The learning canon#

The courses and books that keep getting re-saved — the permanent shelf.

• Harvard CS50P — Introduction to Programming with Python and the full lecture playlist.
• Python for Everybody — Dr. Chuck’s full university course · MIT 6.100L on OCW · University of Helsinki’s Python MOOC — 14 parts, free certificate.
• The Al Sweigart thread: Automate the Boring Stuff, re-saved as the 3rd edition and its practice workbook — “the boring stuff is actually the USEFUL stuff,” which is very much your automation-first worldview.
• Think Python, 3rd edition (free online) · The Hitchhiker’s Guide to Python · Anthony Shaw’s “Learning Python from Scratch”.
• JPMorgan’s internal Python training — written for analysts and traders, works for any busy professional.
• A quiet pattern in this tag: you collect Python teaching in every language — Tsukuba University’s 300-page Japanese intro, Chinese visual-math repos, Spanish MOOC threads. Good pedagogy is good pedagogy.
• When you want structure and accountability: Trey Hunner’s 10-week jump start — one exercise a day for 50 days.

Python for AI — the transition shelf#

How much Python does the AI work actually need? These answer that directly.

• Python QuickStart for People Learning AI — the exact framing of your situation.
• “How much Python is enough to start with AI?” — the honest scoping answer.
• CS50’s AI with Python — full course — the canon course at the intersection.
• Deep Learning with Python, 3rd edition (Chollet) — free web edition too — and his step-by-step notebooks (“highest quality I’ve seen anywhere,” per the save).
• Practical Python AI Programming — a free 4–5-hour read, good weekend material.
• Siddhardhan’s Python-libraries-for-ML module — “the best Python tutor I’ve come across on YouTube,” says the note.
• Bernd Klein’s ML-with-Python book (PDF) · your collected LLM-Python resources doc.
• Learn AI-Assisted Python Programming — learning Python with Copilot/ChatGPT rather than despite them.
• Cisco’s Python Essentials for AgenticOps — the save whose framing (“Python is the glue that turns an AI thought into real action”) belongs in your architect talks.

The Pydantic thread#

Your densest single-vendor cluster — you tracked this ecosystem from validation library to full agent stack.

• “Pydantic is still all you need” — the talk that explains why structured outputs won.
• Why (and how) I am a Pydantic user — the everyday case, pre-AI.
• The PydanticAI arc: the launch — “an agent framework designed for production” → InfoQ’s coverage → V1 and why they built it → pydantic-ai-harness, one of the few projects that goes past prompt-plus-tools demos.
• Building a research agent with PydanticAI (video) — “I actually learned something — about Tavily search API,” you noted.
• Guido’s typeagent team using Pydantic AI + Logfire — validation-by-adoption at the highest level.
• The monty tangent: a durable code mode for agent-framework built on Pydantic’s monty · Drew Breunig on spec-driven development, where monty shows up again.
• Reflections on leaving OpenAI — “pretty much everything operates around FastAPI to create APIs and Pydantic for validation.” The stack you’re learning is the stack they run.

Building agents in Python#

Where this tag and AI Agents overlap — the hands-on, framework-level saves.

• Pamela Fox’s Python + Agents livestream series, saved session by session: building your first agent · adding context and memory · monitoring, evals, and red-teaming — plus the full series archive with slides and code.
• The same agent across Microsoft Agent Framework, LangChain v1, and PydanticAI — proof the frameworks are converging on “a prompt with a list of tools.”
• Google’s Agent Development Kit and the v1.18 visual agent builder · OpenAI’s Agents SDK — the other two corners of the framework triangle.
• Taking a deep agent to production — memory, execution environment, guardrails, durability; the checklist your agentic incident-management work needs.
• Temporal’s Python SDK running ADK agents in workflows — durable execution meets agents; catnip for a platform engineer.
• Multi-agent communication with the A2A Python SDK · Magentic-One, Microsoft’s generalist multi-agent system · smolagents + Browser Use — “OpenAI Operator, light”.
• Demystifiers: Build Your Own Coding Agent — “a while loop, an API call, and a few Python functions” · Build Your Own Openclaw, step by step · Build AI Agents with Python from Scratch.

MCP servers & sandboxed Python#

The 2025–2026 saves split into two practical questions: how do agents get tools, and where does their code run safely?

• “Build your first MCP server in Python” — the PyCon US tutorial — packed house, slides and code included.
• Microsoft’s Let’s Learn MCP Python tutorial · demystifying MCP with Python — beginner’s guide.
• Hosting a remote Python MCP server on Azure Functions — your cloud muscle applied directly.
• FastMCP’s built-in OpenTelemetry support — exporting to Aspire, App Insights, or Logfire; observability for MCP, which is exactly your SRE instinct asking the right question.
• Sandboxing: Simon Willison finally finds his Python-in-a-sandbox solution (MicroPython in WASM) · a Python agent executing coding tasks inside a Cloudflare Sandbox.

FastAPI — the serving layer#

Every AI thing you build eventually needs an API in front of it.

• freeCodeCamp’s FastAPI handbook — the full develop-and-deploy walkthrough.
• FastAPI for MLOps: project structure and API best practices — the bridge between your DevOps habits and ML serving.
• FastAPI for AI Engineers · server-sent events in FastAPI 0.135 — SSE being the transport for streaming agents.
• a conversational-agent FastAPI backend (LangChain/LangGraph + Qdrant, deployment included) and the Travel AI Agent — React + FastAPI + Cosmos DB vector store.
• library-skills — teach your coding agents to use FastAPI properly — run uvx library-skills in the repo; a very 2026 idea.

Tooling & craft#

The modern-Python hygiene shelf — uv, style, logging, and the internals worth understanding once.

• The uv moment: Sebastian Raschka’s uv + venv setup guide (“finally updated from conda + pip”) · the uv risk/benefit debate, curated by Simon Willison · the uv-back-to-pip migration guide — “nobody’s truly locked in.”
• how venv actually works, explained properly — worth it even in Japanese slides.
• Typer hits “Adopt” on the Thoughtworks Radar — CLIs are how agents and humans both drive your tools.
• Modern Good Practices for Python Development · Google’s Python style guide · Modern Python logging (video) · writing better unit tests.
• Internals, once: how async/await works in Python · Armin Ronacher on virtual threads — and his agentic coding recommendations, from Flask’s creator.
• At scale: how Kraken organizes a very large Python monolith — enterprise-architecture reading disguised as a Python post.
• The permanent repos: David Beazley’s Python Mastery course · awesome-python · TheAlgorithms/Python · memory_graph — draw your data structures · 10 ways fastcore makes Python more fun.
• skills.md — a tiny contract that tells the model how your codebase works; where Python craft and AI-generated code meet.

All links on this topic#