CFI is 30% off both plans, and you don't need a code

CFI’s pricing page now shows 30% off without any coupon: Self-Study is $347.90 a year (from $497) and Full-Immersion $592.90 (from $847). Until now that price needed a code that changed every month. I checked it in a fresh browser with no cookies. The page gives no end date, so don’t assume it lasts.

Babbel's 12-month plan is half price

Babbel’s prices page shows the 12-month plan at $107.40 instead of $215.40, which works out to $8.95 a month across all 14 languages. The lifetime plan is $299.99. Babbel runs promotions often and printed no end date, so read the price on the page when you click.

DataCamp's half-price year 'ended' on 17 September. It restarted on 18 September

Same offer, new name: a “Flash Sale” with Premium at $165 for the year instead of $330. Its countdown on 18 September pointed to the evening of 24 September, US Eastern time. That’s the useful part: DataCamp’s deadlines keep coming back, so don’t let one rush you.

Udacity launched a Nanodegree built entirely on Claude

AI Engineering with Claude is 38 hours across four project-based courses: Claude’s Agent SDK, MCP tools, agent evaluation, and a multi-agent code-review system. Projects are reviewed by people, and it counts toward Udacity’s master’s. You’ll need Python and TypeScript first. New members get 40% off their first four months: $150 a month instead of $249.

A free Harvard course on tech ethics, with Michael Sandel and Michael B. Jordan

HarvardX’s Tech Ethics: Critical Thinking in the Age of Apps, Algorithms, and AI opened on 18 September. Sandel, the political philosopher, talks it through with Jordan and a panel of people who work in tech. It’s about 2–3 hours, self-paced, and free to audit. You only pay if you want the certificate.

Educative's AI series adds a course for backend engineers

AI for Backend Engineers is 5 hours, 34 lessons and 2 projects: adding LLM features to a Python backend and treating the model like any other unreliable dependency. That means prompts that return structured output, validating every response, and retries, fallbacks and caching so failures stay predictable. It follows the front-end and product courses from last week.